Top Banner
A new volumetric and shear locking-free 3D enhanced strain element R.J. Alves de Sousa Departamento de Engenharia Meca Ãnica, Universidade de Aveiro, Portugal IDMEC ± Faculdade de Engenharia, Universidade do Porto, Portugal R.M. Natal Jorge IDMEC ± Faculdade de Engenharia, Universidade do Porto, Portugal R.A. Fontes Valente Departamento de Engenharia Meca Ãnica, Universidade de Aveiro, Portugal IDMEC ± Faculdade de Engenharia, Universidade do Porto, Portugal J.M.A. Ce Âsar de Sa  IDMEC ± Faculdade de Engenharia, Universidade do Porto, Portugal Keywords Shell structures, Strain measurement, Shear strength Abstract This paper focuses on the development of a new class of eight-node solid ®nite elements, suitable for the treatment of volumetric and transverse shear locking problems. Doing so, the proposed elements can be used ef®ciently for 3D and thin shell applications. The starting point of the work relies on the analysis of the subspace of incompressible deformations associated with the standard (displacement-based) fully integrated and reduced integrated hexahedral elements. Prediction capabilities for both formulations are de®ned related to nearly-incompressible problems and an enhanced strain approach is developed to improve the performance of the earlier formulation in this case. With the insight into volumetric locking gained and bene®ting from a recently proposed enhanced transverse shear strain procedure for shell applications, a new element conjugating both the capabilities of ef®cient solid and shell formulations is obtained. Numerical results attest the robustness and ef®ciency of the proposed approach, when compared to solid and shell elements well-established in the literature. Introduction Several problems of important physical meaning, such as incompressible elasticity, plasticity of metals or, additionally, the ¯ow of certain ¯uids, involve the inclusion and treatment of the incompressibility phenomenon. The Emerald Research Register for this journal is available at The current issue and full text archive of this journal is available at http://www.emeraldinsight.com/researchregister http://www.emeraldinsight.com/0264-4401.htm It is worth noting the funding by Ministe Ârio da Cie Ãncia e Ensino Superior, Fundac Ëa Äo para a Cie à ncia e Tecnologia, Portugal, under grants PRAXIS XXI/BD/21662/99 as well as FEDER under grant POCTI/33640/EME/2000. These supports are gratefully acknowledged. EC 20,7 896 Received September 2002 Revised May 2003 Accepted May 2003 Engineering Computations Vol. 20 No. 7, 2003 pp. 896-925 q MCB UP Limited 0264-4401 DOI 10.1108/02644400310502036
30

A new volumetric and shear locking-free 3D enhanced strain element

Feb 03, 2023

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: A new volumetric and shear locking-free 3D enhanced strain element

A new volumetric and shearlocking-free 3D enhanced

strain elementRJ Alves de Sousa

Departamento de Engenharia MecaAtilde nica Universidade de AveiroPortugal

IDMEC plusmn Faculdade de Engenharia Universidade do Porto Portugal

RM Natal JorgeIDMEC plusmn Faculdade de Engenharia Universidade do Porto Portugal

RA Fontes ValenteDepartamento de Engenharia MecaAtilde nica Universidade de Aveiro

PortugalIDMEC plusmn Faculdade de Engenharia Universidade do Porto Portugal

JMA CeAcircsar de SaAcircIDMEC plusmn Faculdade de Engenharia Universidade do Porto Portugal

Keywords Shell structures Strain measurement Shear strength

Abstract This paper focuses on the development of a new class of eight-node solid regnite elementssuitable for the treatment of volumetric and transverse shear locking problems Doing so theproposed elements can be used efregciently for 3D and thin shell applications The starting point ofthe work relies on the analysis of the subspace of incompressible deformations associated with thestandard (displacement-based) fully integrated and reduced integrated hexahedral elementsPrediction capabilities for both formulations are deregned related to nearly-incompressible problemsand an enhanced strain approach is developed to improve the performance of the earlierformulation in this case With the insight into volumetric locking gained and beneregting from arecently proposed enhanced transverse shear strain procedure for shell applications a new elementconjugating both the capabilities of efregcient solid and shell formulations is obtained Numericalresults attest the robustness and efregciency of the proposed approach when compared to solid andshell elements well-established in the literature

Introduction

Several problems of important physical meaning such as incompressible

elasticity plasticity of metals or additionally the macrow of certain macruids involve

the inclusion and treatment of the incompressibility phenomenon

The Emerald Research Register for this journal is available at The current issue and full text archive of this journal is available at

httpwwwemeraldinsightcomresearchregister httpwwwemeraldinsightcom0264-4401htm

It is worth noting the funding by MinisteAcircrio da CieAtildencia e Ensino Superior FundacEumlaAumlo para aCieAtildencia e Tecnologia Portugal under grants PRAXIS XXIBD2166299 as well as FEDER undergrant POCTI33640EME2000 These supports are gratefully acknowledged

EC207

896

Received September2002Revised May 2003Accepted May 2003

Engineering ComputationsVol 20 No 7 2003pp 896-925

q MCB UP Limited0264-4401DOI 10110802644400310502036

Mathematically incompressibility modelling relates to the imposition of aconstraint over the analysed continuum enforcing the volumetric part of thestrain regeld to be zero or very small when compared to its deviatoriccounterpart However the same treatment involving the frame of the regniteelement method (FEM) imposes some difregculties in particular when low-orderelementsrsquo formulations are employed

Low-order elements offer simpler implementation when compared withhigher-order formulations allowing at the same time straightforward meshoperations to be performed Nevertheless problems can arise in the presence ofincompressibility as stated before leading to the so-called volumetric lockingphenomenon In this case bilinear elements (in two-dimensional analysis) andtrilinear solid elements (in three-dimensional analysis) are not able to grant thecomplete nullity of the volumetric strain This leads to an overestimation ofstiffness values related to the volumetric deformation and in consequence tonear zero values of displacements obtained

Several approaches have been proposed in the latest decades to reduce oralleviate volumetric locking occurrence Reduced and selective reducedintegration (SRI) techniques were the regrst successful irreducible form ofsolutions for locking problems although in the beginning not directedspeciregcally to volumetric locking (Hughes et al 1978 Zienkiewicz et al 1971)For the particular case of the eight-node hexahedral regnite element bothtechniques correspond to the use of a lower quadrature rule (one Gauss pointcorresponding to the elementsrsquo center) in opposition to the so-called completequadrature rule (2 pound 2 pound 2 Gauss points) While the (total) reduced integrationappeared to lead to non-physical (spurious) deformation patterns selectivereduced integration proved quite successful being in some extent thepredecessor of the B-bar method (Hughes 1977) In the latter shape functionsderivatives related to the volumetric part of deformation were replaced byapproximations resulting from a mixed formulation without resorting toreduced integrations

All these methods proved to be effective proposals in alleviating speciregcallythe volumetric locking although their performance in bending dominatedsituations revealed some deregciencies For selective reduced integration andB-bar cases in addition their application is in some extent limited to materialswhose strain (stress) tensor can be decomposed into volumetric and deviatoricparts Other formulations succeeded in using an augmented functional whencompared to that obtained from displacement-based approaches incorporatingadditional regelds into the formulation and leading to the onset of general mixedmethods For the up mixed formulation displacements are interpolated withfunctions providing C 0 continuity requirements while the pressure regeld isintroduced via discontinuous functions between elements (Hughes 2000Zienkiewicz and Taylor 2000)

A new enhancedstrain element

897

In the 1990s the work of Simo and Rifai (1990) introduced the enhancedstrain method In this formulation the strain regeld is enlarged with the inclusionof an extra internal regeld of variables resulting therefore in additionaldeformation modes A particular choice of the extra modes of deformation ledto the QM6 element derived via the incompatible modes method of Wilson et al(1973) a formulation designed to improve the performance of quadrilateralelements in 2D bending problems (Taylor et al 1976)

Due to its versatility the enhanced strain method was successfully appliedin 2D 3D and shell formulations achieving good results even with coarsemeshes (Andelregnger and Ramm 1993 Armero and Dvorkin 2000 de Borstand Groen 1999 de Souza Neto et al 1996 Glaser and Armero 1997 Kasperand Taylor 2000 Korelc and Wriggers 1996 Piltner 2000 Rohel and Ramm1996 Simo and Armero 1992 Simo and Rifai 1990 Simo et al 1993) Thepossibility of respecting some conditions ordffreelyordm adding extra variables(deformation modes) represents an important matter being directly related to agiven elementrsquos performance

Concerning near incompressible situations and speciregcally for 2D planestrain problems the number of additional variables to use in reliablevolumetric locking-free elements can vary from two to four as in the proposalsof CeAcircsar de SaAcirc and Natal Jorge (1999) or Simo and Rifai (1990) among others In3D analysis this number is still a matter of discussion varying from 9 to 12 inaccordance to Korelc and Wriggers (1996) 12 according to Simo et al (1993) orstill 6 to 30 according to Andelregnger and Ramm (1993) only to quote somerelevant publications in the regeld

Still dealing with hexahedral regnite elements but departing from the exposedbefore an interesting application relates to the shell structures modelling Infact 3D elements were the starting point of some shell elements as in the onsetof the degenerated approach with the pioneer work of Ahmad et al (1970) Forthe proper reproduction of this particular kinematics either of Mindlin orKirchhoff-Koiter type some simpliregcations were taken leading to a class ofelements incorporating plane-stress assumptions with undesirable losses in thegeneral character provided by full 3D material laws of solid elements

In this sense a correct reproduction of shells (and plates) structuresbehaviour with the use of tridimensional elements is desirable Apart from theeasier formulation and material considerations solid elements provide astraightforward extension to geometrically non-linear problems in particularin the presence of large rotations once the only degrees of freedom involved are(additive type) translations Another difregculty in dealing with shell elements isthe treatment of corner-like zones (for example in reinforced shell structures)with the resulting drilling degrees of freedom and also the treatment ofconjunction between solids and shells in the same model

For a successful modelling of general shells with no limitations on thicknessvalues transverse shear locking effects must be accounted Transverse shear

EC207

898

locking phenomenon relates to the inability of a given element to reproduce anull transverse shear strain energy state in pure bending (CeAcircsar de SaAcirc et al2002) Backward in the earlier developments shear locking in degenerated shellelements were revealed as thickness values tended to small values The regrstsuccessful procedures to overcome this artiregcial behaviour were the citedselective numerical under-integration of the transverse shear strain termsThe appearance of spurious zero energy deformation modes required thedevelopment of stabilization or projection techniques in order to recover thecorrect rank of the stiffness matrices involved (Belytschko et al 1992) In casethe full quadrature is maintained some procedures proved to be efregcient inattenuating transverse shear locking Among the successful techniques werethe so-called mixed interpolation of tensorial (transverse shear strain)components (MITC) from the initial work of Dvorkin and Bathe (1984) andlater on Bathe and Dvorkin (1986) Bathe et al (2000) Chapelle and Bathe (2000)and Lee and Bathe (2002)

Additionally it is also worth noting the assumed natural strain (ANS)approach regrst introduced by Park (1986) and Park and Stanley (1986) Bothformulations employ a set of additional sampling points over a regnite element inorder to obtain a substitute or complementary strain regeld leading to thefulreglment of the Kirchhoff-Koiter hypothesis in the thin shell limit Relating tothe MITC this procedure is usually cast into a more general frame than theconventional degenerated formulation An example is the ordfsolid-shellordm class ofelements developed by Doll et al (2000) Harnau and Schweizerhof (2002)Hauptmann and Schweizerhof (1998) and Hauptmann et al (2001) In thisformulation mixed interpolation as described was used for the transverseshear locking while the EAS method is adopted in attenuating membranelocking Thickness volumetric and trapezoidal locking appearance is related tothese elements and proper treatments of the last locking effects are needed inorder to reach a successful and general regnite element (Harnau andSchweizerhof 2002)

In this paper based on the framework of subspace of deformations analysisalready successfully applied in 2D plane strain (CeAcircsar de SaAcirc and Natal Jorge1999) and shell problems (CeAcircsar de SaAcirc et al 2002) a new volumetric and shearlocking-free EAS element for 3D analyses is proposed The main idea is theimprovement of the original strain regeld in an additive way leading to elementswith higher performance in the presence of volumetric locking (3Dapplications) and transverse shear locking (shell structures applications)

Volumetric locking plusmn subspace analysisThe incompressibility problem can be formulated as a constrainedminimization of a functional (CeAcircsar de SaAcirc and Natal Jorge 1999) In simpleterms the goal is to obtain in the linear space of admissible solutions U a regeldof displacement u that minimizes the total energy of the system located in the

A new enhancedstrain element

899

subspace of the incompressible deformations I and simultaneously containedin the space of all the solutions (I U) This statement can be posed in theform

I = u [ U div(u) = 0 (1)

In an approach done by the FEM the linear space of the admissible solutions Uand the respective subspace I previously deregned are approximated by thespaces U h and I h respectively

A two regeld regnite element solution can be expressed for linear elasticityaccording to CeAcircsar de SaAcirc and Natal Jorge (1999) by

K Q

2 Q 0

uh

ph

=fhext

0

(2)

where fext is the vector of applied external forces p is the hydrostatic pressureand K is the stiffness matrix The superscript (h) means that the variable incause is a regnite element approximation

The incompressibility condition is given by the second group of equationsderegned in equation (2)

Quh = 0 (3)

which will deregne the subspace of the incompressible deformations I h as

I h = uh [ U h Quh = 0 (4)

In order to avoid the trivial solution (uh= 0) in equation (3) the regeld ofdisplacements uh should belong to the nullspace of Q that is to the subspaceof the incompressible deformations I h The approximated displacements uh

are contained in I h being therefore a linear combination of a given basis of I h

elementsIf the subspace I h is an approach to the original subspace I it is plausible to

admit that it cannot reproduce all the possible solutions contained in I In factdifferent formulations lead to better or worse approximations for the I subspaceThe volumetric locking phenomenon occurs when for a certain group ofboundary conditions and external forces applied in a near incompressiblesituation the expected solution or some of its components do not appearproperly represented in the subspace I h Different types of elements can offerdistinct subspaces of incompressible deformations originating thereforedifferent solutions (CeAcircsar de SaAcirc and Owen 1986)

To characterize the already deregned subspace I h consider a standardisoparametric eight-node hexahedral element with domain Oe (Figure 1)For small deformations the incompressibility constraint (e ii = 0) can takethe form

EC207

900

Z

Oe

div(u) dOe =Z

Oe

shy u

shy x+

shy v

shy Z+

shy w

shy z

sup3 acutedOe = 0 (5)

At the element level in any point of the domain the displacement regeld can beinterpolated using the standard linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000)

u lt uhe = N(x Z z )de (6)

One way to guarantee the incompressibility condition is to assure that theintegrand function in equation (5) is zero Substituting equation (6) inequation (5) results

shy u

shy x+

shy v

shy Z+

shy w

shy z= [N ix(x Z z ) N iZ(x Z z ) N iz(x Z z )] de = 0 (7)

for i = 1 nenodes Making use of a complete integration (2 pound 2 pound 2 eight Gauss

points Figure 1) the application of equation (7) leads to

Figure 1Standard eight-node

isoparametrichexahedral element with

2 pound 2 pound 2 Gauss pointsnumerical integration

rule

A new enhancedstrain element

901

2 a 2 a 2 a a 2 c 2 c c c 2 b 2 c a 2 c 2 c 2 c a c 2 b c b b b 2 b c c

2 a 2 c 2 c a 2 a 2 a c a 2 c 2 c c 2 b 2 c 2 b c c 2 c a b c c 2 b b b

2 c 2 c 2 b c 2 a 2 c a a 2 a 2 a c 2 c 2 b 2 b b b 2 c c c c a 2 c b c

2 c 2 a 2 c c 2 c 2 b a c 2 c 2 a a 2 a 2 b 2 c c b 2 b b c b c 2 c c a

2 c 2 c 2 a c 2 b 2 c b b 2 b 2 b c 2 c 2 a 2 a a a 2 c c c c b 2 c a c

2 c 2 b 2 c c 2 c 2 a b c 2 c 2 b b 2 b 2 a 2 c c a 2 a a c a c 2 c c b

2 b 2 b 2 b b 2 c 2 c c c 2 a 2 c b 2 c 2 c 2 c b c 2 a c a a a 2 a c c

2 b 2 c 2 c b 2 b 2 b c b 2 c 2 c c 2 a 2 c 2 a c c 2 c b a c c 2 a a a

2

666666666666666666666664

3

777777777777777777777775

pound de= 0 (8)

where the results for each Gauss point are grouped by rows and

a =1

8(1 + f )(1 + f ) b =

1

8(1 2 f )(1 2 f ) c =

1

8(1 + f )(1 2 f )

and f =

3

p

3

(9)

Equation (8) contains a matrix (Q) of rank seven Being 24 the total number ofdegrees of freedom ie the dimension of the subspace of admissible solutionsU h there will be a dimension 242 7 = 17 for the incompressible deformationssubspace

rank(Q) = 7 ^ nullity(Q) = 17 (10)

If for the same element a reduced numerical integration scheme with only oneGauss point is used (x = Z = z = 0) the imposition of condition (7) will lead toa subspace of dimension 23 when the maximum dimension of the space of theadmissible solutions is 24 The matrix in equation (8) is rewritten as follows

1

82 1 2 1 2 1 1 2 1 2 1 1 1 2 1 2 1 1 2 1 2 1 2 1 1 1 2 1 1 1 1 1 2 1 1 1

pound curren

pound de = 0(11)

and

EC207

902

rank(Q) = 1 ^ nullity(Q)= 23 (12)

Analysing these two possible bases for the subspace of the incompressibledeformations it can be inferred clearly that the use of the reduced integrationallows the reproduction of more six incompressible displacement modes thanthe case of complete integration Since the admissible incompressible solutionsare nothing more than a linear combination of a given I h basis this can be aquite reasonable explanation for the good performance of the reducedintegration techniques in volumetric locking problems and why the classicalcomplete integration numerical integration locks

For a clear illustration of the last statements a graphical representation ofthe linearly independent elements which form the basis of the incompressibledeformations subspace is shown in Figure 2 The six rigid body motions can beobtained linearly combining these elements The displacement regeld associatedwith each mode is plotted in Table I The modes 1-17 are reproduced both bythe complete and the reduced integration The modes 18-23 are reproduced onlyby the reduced integration being volumetric locking situations for the completeintegration They can be divided into four main groups

(1) simple edges translations in x y and z directions represented by themodes 1-12

(2) expansioncontraction of one face modes 13-17

(3) hourglass modes 18-20

(4) warping modes 21-23

In the following sections the proposed approaches to alleviate locking effects(using the enhanced strain method) are conveniently deregned within thesubspace analysis framework just described

Formulation of volumetric and transverse shear locking-freeelementsAs already referred the reduced integration selective reduced integration andsimilar techniques have proved to be efregcient methods on attenuating thevolumetric locking However the behaviour in bending dominated situations isnot the best and for the case of the SRI techniques the analysis is in somemeasure limited to materials where the stress tensor allows the decompositionin volumetric and deviatoric components

On the other hand the enhanced strain method is a powerful techniquein the sense that permits the inclusion of more or less additional variablesfor the enhanced strain regeld Thus it is possible to construct a formulationwith good behaviour both in near incompressible situation and bendingsituations

However the number of additional variables in the enhanced regeld is acrucial matter If increasing the number of additional variables and

A new enhancedstrain element

903

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 2: A new volumetric and shear locking-free 3D enhanced strain element

Mathematically incompressibility modelling relates to the imposition of aconstraint over the analysed continuum enforcing the volumetric part of thestrain regeld to be zero or very small when compared to its deviatoriccounterpart However the same treatment involving the frame of the regniteelement method (FEM) imposes some difregculties in particular when low-orderelementsrsquo formulations are employed

Low-order elements offer simpler implementation when compared withhigher-order formulations allowing at the same time straightforward meshoperations to be performed Nevertheless problems can arise in the presence ofincompressibility as stated before leading to the so-called volumetric lockingphenomenon In this case bilinear elements (in two-dimensional analysis) andtrilinear solid elements (in three-dimensional analysis) are not able to grant thecomplete nullity of the volumetric strain This leads to an overestimation ofstiffness values related to the volumetric deformation and in consequence tonear zero values of displacements obtained

Several approaches have been proposed in the latest decades to reduce oralleviate volumetric locking occurrence Reduced and selective reducedintegration (SRI) techniques were the regrst successful irreducible form ofsolutions for locking problems although in the beginning not directedspeciregcally to volumetric locking (Hughes et al 1978 Zienkiewicz et al 1971)For the particular case of the eight-node hexahedral regnite element bothtechniques correspond to the use of a lower quadrature rule (one Gauss pointcorresponding to the elementsrsquo center) in opposition to the so-called completequadrature rule (2 pound 2 pound 2 Gauss points) While the (total) reduced integrationappeared to lead to non-physical (spurious) deformation patterns selectivereduced integration proved quite successful being in some extent thepredecessor of the B-bar method (Hughes 1977) In the latter shape functionsderivatives related to the volumetric part of deformation were replaced byapproximations resulting from a mixed formulation without resorting toreduced integrations

All these methods proved to be effective proposals in alleviating speciregcallythe volumetric locking although their performance in bending dominatedsituations revealed some deregciencies For selective reduced integration andB-bar cases in addition their application is in some extent limited to materialswhose strain (stress) tensor can be decomposed into volumetric and deviatoricparts Other formulations succeeded in using an augmented functional whencompared to that obtained from displacement-based approaches incorporatingadditional regelds into the formulation and leading to the onset of general mixedmethods For the up mixed formulation displacements are interpolated withfunctions providing C 0 continuity requirements while the pressure regeld isintroduced via discontinuous functions between elements (Hughes 2000Zienkiewicz and Taylor 2000)

A new enhancedstrain element

897

In the 1990s the work of Simo and Rifai (1990) introduced the enhancedstrain method In this formulation the strain regeld is enlarged with the inclusionof an extra internal regeld of variables resulting therefore in additionaldeformation modes A particular choice of the extra modes of deformation ledto the QM6 element derived via the incompatible modes method of Wilson et al(1973) a formulation designed to improve the performance of quadrilateralelements in 2D bending problems (Taylor et al 1976)

Due to its versatility the enhanced strain method was successfully appliedin 2D 3D and shell formulations achieving good results even with coarsemeshes (Andelregnger and Ramm 1993 Armero and Dvorkin 2000 de Borstand Groen 1999 de Souza Neto et al 1996 Glaser and Armero 1997 Kasperand Taylor 2000 Korelc and Wriggers 1996 Piltner 2000 Rohel and Ramm1996 Simo and Armero 1992 Simo and Rifai 1990 Simo et al 1993) Thepossibility of respecting some conditions ordffreelyordm adding extra variables(deformation modes) represents an important matter being directly related to agiven elementrsquos performance

Concerning near incompressible situations and speciregcally for 2D planestrain problems the number of additional variables to use in reliablevolumetric locking-free elements can vary from two to four as in the proposalsof CeAcircsar de SaAcirc and Natal Jorge (1999) or Simo and Rifai (1990) among others In3D analysis this number is still a matter of discussion varying from 9 to 12 inaccordance to Korelc and Wriggers (1996) 12 according to Simo et al (1993) orstill 6 to 30 according to Andelregnger and Ramm (1993) only to quote somerelevant publications in the regeld

Still dealing with hexahedral regnite elements but departing from the exposedbefore an interesting application relates to the shell structures modelling Infact 3D elements were the starting point of some shell elements as in the onsetof the degenerated approach with the pioneer work of Ahmad et al (1970) Forthe proper reproduction of this particular kinematics either of Mindlin orKirchhoff-Koiter type some simpliregcations were taken leading to a class ofelements incorporating plane-stress assumptions with undesirable losses in thegeneral character provided by full 3D material laws of solid elements

In this sense a correct reproduction of shells (and plates) structuresbehaviour with the use of tridimensional elements is desirable Apart from theeasier formulation and material considerations solid elements provide astraightforward extension to geometrically non-linear problems in particularin the presence of large rotations once the only degrees of freedom involved are(additive type) translations Another difregculty in dealing with shell elements isthe treatment of corner-like zones (for example in reinforced shell structures)with the resulting drilling degrees of freedom and also the treatment ofconjunction between solids and shells in the same model

For a successful modelling of general shells with no limitations on thicknessvalues transverse shear locking effects must be accounted Transverse shear

EC207

898

locking phenomenon relates to the inability of a given element to reproduce anull transverse shear strain energy state in pure bending (CeAcircsar de SaAcirc et al2002) Backward in the earlier developments shear locking in degenerated shellelements were revealed as thickness values tended to small values The regrstsuccessful procedures to overcome this artiregcial behaviour were the citedselective numerical under-integration of the transverse shear strain termsThe appearance of spurious zero energy deformation modes required thedevelopment of stabilization or projection techniques in order to recover thecorrect rank of the stiffness matrices involved (Belytschko et al 1992) In casethe full quadrature is maintained some procedures proved to be efregcient inattenuating transverse shear locking Among the successful techniques werethe so-called mixed interpolation of tensorial (transverse shear strain)components (MITC) from the initial work of Dvorkin and Bathe (1984) andlater on Bathe and Dvorkin (1986) Bathe et al (2000) Chapelle and Bathe (2000)and Lee and Bathe (2002)

Additionally it is also worth noting the assumed natural strain (ANS)approach regrst introduced by Park (1986) and Park and Stanley (1986) Bothformulations employ a set of additional sampling points over a regnite element inorder to obtain a substitute or complementary strain regeld leading to thefulreglment of the Kirchhoff-Koiter hypothesis in the thin shell limit Relating tothe MITC this procedure is usually cast into a more general frame than theconventional degenerated formulation An example is the ordfsolid-shellordm class ofelements developed by Doll et al (2000) Harnau and Schweizerhof (2002)Hauptmann and Schweizerhof (1998) and Hauptmann et al (2001) In thisformulation mixed interpolation as described was used for the transverseshear locking while the EAS method is adopted in attenuating membranelocking Thickness volumetric and trapezoidal locking appearance is related tothese elements and proper treatments of the last locking effects are needed inorder to reach a successful and general regnite element (Harnau andSchweizerhof 2002)

In this paper based on the framework of subspace of deformations analysisalready successfully applied in 2D plane strain (CeAcircsar de SaAcirc and Natal Jorge1999) and shell problems (CeAcircsar de SaAcirc et al 2002) a new volumetric and shearlocking-free EAS element for 3D analyses is proposed The main idea is theimprovement of the original strain regeld in an additive way leading to elementswith higher performance in the presence of volumetric locking (3Dapplications) and transverse shear locking (shell structures applications)

Volumetric locking plusmn subspace analysisThe incompressibility problem can be formulated as a constrainedminimization of a functional (CeAcircsar de SaAcirc and Natal Jorge 1999) In simpleterms the goal is to obtain in the linear space of admissible solutions U a regeldof displacement u that minimizes the total energy of the system located in the

A new enhancedstrain element

899

subspace of the incompressible deformations I and simultaneously containedin the space of all the solutions (I U) This statement can be posed in theform

I = u [ U div(u) = 0 (1)

In an approach done by the FEM the linear space of the admissible solutions Uand the respective subspace I previously deregned are approximated by thespaces U h and I h respectively

A two regeld regnite element solution can be expressed for linear elasticityaccording to CeAcircsar de SaAcirc and Natal Jorge (1999) by

K Q

2 Q 0

uh

ph

=fhext

0

(2)

where fext is the vector of applied external forces p is the hydrostatic pressureand K is the stiffness matrix The superscript (h) means that the variable incause is a regnite element approximation

The incompressibility condition is given by the second group of equationsderegned in equation (2)

Quh = 0 (3)

which will deregne the subspace of the incompressible deformations I h as

I h = uh [ U h Quh = 0 (4)

In order to avoid the trivial solution (uh= 0) in equation (3) the regeld ofdisplacements uh should belong to the nullspace of Q that is to the subspaceof the incompressible deformations I h The approximated displacements uh

are contained in I h being therefore a linear combination of a given basis of I h

elementsIf the subspace I h is an approach to the original subspace I it is plausible to

admit that it cannot reproduce all the possible solutions contained in I In factdifferent formulations lead to better or worse approximations for the I subspaceThe volumetric locking phenomenon occurs when for a certain group ofboundary conditions and external forces applied in a near incompressiblesituation the expected solution or some of its components do not appearproperly represented in the subspace I h Different types of elements can offerdistinct subspaces of incompressible deformations originating thereforedifferent solutions (CeAcircsar de SaAcirc and Owen 1986)

To characterize the already deregned subspace I h consider a standardisoparametric eight-node hexahedral element with domain Oe (Figure 1)For small deformations the incompressibility constraint (e ii = 0) can takethe form

EC207

900

Z

Oe

div(u) dOe =Z

Oe

shy u

shy x+

shy v

shy Z+

shy w

shy z

sup3 acutedOe = 0 (5)

At the element level in any point of the domain the displacement regeld can beinterpolated using the standard linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000)

u lt uhe = N(x Z z )de (6)

One way to guarantee the incompressibility condition is to assure that theintegrand function in equation (5) is zero Substituting equation (6) inequation (5) results

shy u

shy x+

shy v

shy Z+

shy w

shy z= [N ix(x Z z ) N iZ(x Z z ) N iz(x Z z )] de = 0 (7)

for i = 1 nenodes Making use of a complete integration (2 pound 2 pound 2 eight Gauss

points Figure 1) the application of equation (7) leads to

Figure 1Standard eight-node

isoparametrichexahedral element with

2 pound 2 pound 2 Gauss pointsnumerical integration

rule

A new enhancedstrain element

901

2 a 2 a 2 a a 2 c 2 c c c 2 b 2 c a 2 c 2 c 2 c a c 2 b c b b b 2 b c c

2 a 2 c 2 c a 2 a 2 a c a 2 c 2 c c 2 b 2 c 2 b c c 2 c a b c c 2 b b b

2 c 2 c 2 b c 2 a 2 c a a 2 a 2 a c 2 c 2 b 2 b b b 2 c c c c a 2 c b c

2 c 2 a 2 c c 2 c 2 b a c 2 c 2 a a 2 a 2 b 2 c c b 2 b b c b c 2 c c a

2 c 2 c 2 a c 2 b 2 c b b 2 b 2 b c 2 c 2 a 2 a a a 2 c c c c b 2 c a c

2 c 2 b 2 c c 2 c 2 a b c 2 c 2 b b 2 b 2 a 2 c c a 2 a a c a c 2 c c b

2 b 2 b 2 b b 2 c 2 c c c 2 a 2 c b 2 c 2 c 2 c b c 2 a c a a a 2 a c c

2 b 2 c 2 c b 2 b 2 b c b 2 c 2 c c 2 a 2 c 2 a c c 2 c b a c c 2 a a a

2

666666666666666666666664

3

777777777777777777777775

pound de= 0 (8)

where the results for each Gauss point are grouped by rows and

a =1

8(1 + f )(1 + f ) b =

1

8(1 2 f )(1 2 f ) c =

1

8(1 + f )(1 2 f )

and f =

3

p

3

(9)

Equation (8) contains a matrix (Q) of rank seven Being 24 the total number ofdegrees of freedom ie the dimension of the subspace of admissible solutionsU h there will be a dimension 242 7 = 17 for the incompressible deformationssubspace

rank(Q) = 7 ^ nullity(Q) = 17 (10)

If for the same element a reduced numerical integration scheme with only oneGauss point is used (x = Z = z = 0) the imposition of condition (7) will lead toa subspace of dimension 23 when the maximum dimension of the space of theadmissible solutions is 24 The matrix in equation (8) is rewritten as follows

1

82 1 2 1 2 1 1 2 1 2 1 1 1 2 1 2 1 1 2 1 2 1 2 1 1 1 2 1 1 1 1 1 2 1 1 1

pound curren

pound de = 0(11)

and

EC207

902

rank(Q) = 1 ^ nullity(Q)= 23 (12)

Analysing these two possible bases for the subspace of the incompressibledeformations it can be inferred clearly that the use of the reduced integrationallows the reproduction of more six incompressible displacement modes thanthe case of complete integration Since the admissible incompressible solutionsare nothing more than a linear combination of a given I h basis this can be aquite reasonable explanation for the good performance of the reducedintegration techniques in volumetric locking problems and why the classicalcomplete integration numerical integration locks

For a clear illustration of the last statements a graphical representation ofthe linearly independent elements which form the basis of the incompressibledeformations subspace is shown in Figure 2 The six rigid body motions can beobtained linearly combining these elements The displacement regeld associatedwith each mode is plotted in Table I The modes 1-17 are reproduced both bythe complete and the reduced integration The modes 18-23 are reproduced onlyby the reduced integration being volumetric locking situations for the completeintegration They can be divided into four main groups

(1) simple edges translations in x y and z directions represented by themodes 1-12

(2) expansioncontraction of one face modes 13-17

(3) hourglass modes 18-20

(4) warping modes 21-23

In the following sections the proposed approaches to alleviate locking effects(using the enhanced strain method) are conveniently deregned within thesubspace analysis framework just described

Formulation of volumetric and transverse shear locking-freeelementsAs already referred the reduced integration selective reduced integration andsimilar techniques have proved to be efregcient methods on attenuating thevolumetric locking However the behaviour in bending dominated situations isnot the best and for the case of the SRI techniques the analysis is in somemeasure limited to materials where the stress tensor allows the decompositionin volumetric and deviatoric components

On the other hand the enhanced strain method is a powerful techniquein the sense that permits the inclusion of more or less additional variablesfor the enhanced strain regeld Thus it is possible to construct a formulationwith good behaviour both in near incompressible situation and bendingsituations

However the number of additional variables in the enhanced regeld is acrucial matter If increasing the number of additional variables and

A new enhancedstrain element

903

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 3: A new volumetric and shear locking-free 3D enhanced strain element

In the 1990s the work of Simo and Rifai (1990) introduced the enhancedstrain method In this formulation the strain regeld is enlarged with the inclusionof an extra internal regeld of variables resulting therefore in additionaldeformation modes A particular choice of the extra modes of deformation ledto the QM6 element derived via the incompatible modes method of Wilson et al(1973) a formulation designed to improve the performance of quadrilateralelements in 2D bending problems (Taylor et al 1976)

Due to its versatility the enhanced strain method was successfully appliedin 2D 3D and shell formulations achieving good results even with coarsemeshes (Andelregnger and Ramm 1993 Armero and Dvorkin 2000 de Borstand Groen 1999 de Souza Neto et al 1996 Glaser and Armero 1997 Kasperand Taylor 2000 Korelc and Wriggers 1996 Piltner 2000 Rohel and Ramm1996 Simo and Armero 1992 Simo and Rifai 1990 Simo et al 1993) Thepossibility of respecting some conditions ordffreelyordm adding extra variables(deformation modes) represents an important matter being directly related to agiven elementrsquos performance

Concerning near incompressible situations and speciregcally for 2D planestrain problems the number of additional variables to use in reliablevolumetric locking-free elements can vary from two to four as in the proposalsof CeAcircsar de SaAcirc and Natal Jorge (1999) or Simo and Rifai (1990) among others In3D analysis this number is still a matter of discussion varying from 9 to 12 inaccordance to Korelc and Wriggers (1996) 12 according to Simo et al (1993) orstill 6 to 30 according to Andelregnger and Ramm (1993) only to quote somerelevant publications in the regeld

Still dealing with hexahedral regnite elements but departing from the exposedbefore an interesting application relates to the shell structures modelling Infact 3D elements were the starting point of some shell elements as in the onsetof the degenerated approach with the pioneer work of Ahmad et al (1970) Forthe proper reproduction of this particular kinematics either of Mindlin orKirchhoff-Koiter type some simpliregcations were taken leading to a class ofelements incorporating plane-stress assumptions with undesirable losses in thegeneral character provided by full 3D material laws of solid elements

In this sense a correct reproduction of shells (and plates) structuresbehaviour with the use of tridimensional elements is desirable Apart from theeasier formulation and material considerations solid elements provide astraightforward extension to geometrically non-linear problems in particularin the presence of large rotations once the only degrees of freedom involved are(additive type) translations Another difregculty in dealing with shell elements isthe treatment of corner-like zones (for example in reinforced shell structures)with the resulting drilling degrees of freedom and also the treatment ofconjunction between solids and shells in the same model

For a successful modelling of general shells with no limitations on thicknessvalues transverse shear locking effects must be accounted Transverse shear

EC207

898

locking phenomenon relates to the inability of a given element to reproduce anull transverse shear strain energy state in pure bending (CeAcircsar de SaAcirc et al2002) Backward in the earlier developments shear locking in degenerated shellelements were revealed as thickness values tended to small values The regrstsuccessful procedures to overcome this artiregcial behaviour were the citedselective numerical under-integration of the transverse shear strain termsThe appearance of spurious zero energy deformation modes required thedevelopment of stabilization or projection techniques in order to recover thecorrect rank of the stiffness matrices involved (Belytschko et al 1992) In casethe full quadrature is maintained some procedures proved to be efregcient inattenuating transverse shear locking Among the successful techniques werethe so-called mixed interpolation of tensorial (transverse shear strain)components (MITC) from the initial work of Dvorkin and Bathe (1984) andlater on Bathe and Dvorkin (1986) Bathe et al (2000) Chapelle and Bathe (2000)and Lee and Bathe (2002)

Additionally it is also worth noting the assumed natural strain (ANS)approach regrst introduced by Park (1986) and Park and Stanley (1986) Bothformulations employ a set of additional sampling points over a regnite element inorder to obtain a substitute or complementary strain regeld leading to thefulreglment of the Kirchhoff-Koiter hypothesis in the thin shell limit Relating tothe MITC this procedure is usually cast into a more general frame than theconventional degenerated formulation An example is the ordfsolid-shellordm class ofelements developed by Doll et al (2000) Harnau and Schweizerhof (2002)Hauptmann and Schweizerhof (1998) and Hauptmann et al (2001) In thisformulation mixed interpolation as described was used for the transverseshear locking while the EAS method is adopted in attenuating membranelocking Thickness volumetric and trapezoidal locking appearance is related tothese elements and proper treatments of the last locking effects are needed inorder to reach a successful and general regnite element (Harnau andSchweizerhof 2002)

In this paper based on the framework of subspace of deformations analysisalready successfully applied in 2D plane strain (CeAcircsar de SaAcirc and Natal Jorge1999) and shell problems (CeAcircsar de SaAcirc et al 2002) a new volumetric and shearlocking-free EAS element for 3D analyses is proposed The main idea is theimprovement of the original strain regeld in an additive way leading to elementswith higher performance in the presence of volumetric locking (3Dapplications) and transverse shear locking (shell structures applications)

Volumetric locking plusmn subspace analysisThe incompressibility problem can be formulated as a constrainedminimization of a functional (CeAcircsar de SaAcirc and Natal Jorge 1999) In simpleterms the goal is to obtain in the linear space of admissible solutions U a regeldof displacement u that minimizes the total energy of the system located in the

A new enhancedstrain element

899

subspace of the incompressible deformations I and simultaneously containedin the space of all the solutions (I U) This statement can be posed in theform

I = u [ U div(u) = 0 (1)

In an approach done by the FEM the linear space of the admissible solutions Uand the respective subspace I previously deregned are approximated by thespaces U h and I h respectively

A two regeld regnite element solution can be expressed for linear elasticityaccording to CeAcircsar de SaAcirc and Natal Jorge (1999) by

K Q

2 Q 0

uh

ph

=fhext

0

(2)

where fext is the vector of applied external forces p is the hydrostatic pressureand K is the stiffness matrix The superscript (h) means that the variable incause is a regnite element approximation

The incompressibility condition is given by the second group of equationsderegned in equation (2)

Quh = 0 (3)

which will deregne the subspace of the incompressible deformations I h as

I h = uh [ U h Quh = 0 (4)

In order to avoid the trivial solution (uh= 0) in equation (3) the regeld ofdisplacements uh should belong to the nullspace of Q that is to the subspaceof the incompressible deformations I h The approximated displacements uh

are contained in I h being therefore a linear combination of a given basis of I h

elementsIf the subspace I h is an approach to the original subspace I it is plausible to

admit that it cannot reproduce all the possible solutions contained in I In factdifferent formulations lead to better or worse approximations for the I subspaceThe volumetric locking phenomenon occurs when for a certain group ofboundary conditions and external forces applied in a near incompressiblesituation the expected solution or some of its components do not appearproperly represented in the subspace I h Different types of elements can offerdistinct subspaces of incompressible deformations originating thereforedifferent solutions (CeAcircsar de SaAcirc and Owen 1986)

To characterize the already deregned subspace I h consider a standardisoparametric eight-node hexahedral element with domain Oe (Figure 1)For small deformations the incompressibility constraint (e ii = 0) can takethe form

EC207

900

Z

Oe

div(u) dOe =Z

Oe

shy u

shy x+

shy v

shy Z+

shy w

shy z

sup3 acutedOe = 0 (5)

At the element level in any point of the domain the displacement regeld can beinterpolated using the standard linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000)

u lt uhe = N(x Z z )de (6)

One way to guarantee the incompressibility condition is to assure that theintegrand function in equation (5) is zero Substituting equation (6) inequation (5) results

shy u

shy x+

shy v

shy Z+

shy w

shy z= [N ix(x Z z ) N iZ(x Z z ) N iz(x Z z )] de = 0 (7)

for i = 1 nenodes Making use of a complete integration (2 pound 2 pound 2 eight Gauss

points Figure 1) the application of equation (7) leads to

Figure 1Standard eight-node

isoparametrichexahedral element with

2 pound 2 pound 2 Gauss pointsnumerical integration

rule

A new enhancedstrain element

901

2 a 2 a 2 a a 2 c 2 c c c 2 b 2 c a 2 c 2 c 2 c a c 2 b c b b b 2 b c c

2 a 2 c 2 c a 2 a 2 a c a 2 c 2 c c 2 b 2 c 2 b c c 2 c a b c c 2 b b b

2 c 2 c 2 b c 2 a 2 c a a 2 a 2 a c 2 c 2 b 2 b b b 2 c c c c a 2 c b c

2 c 2 a 2 c c 2 c 2 b a c 2 c 2 a a 2 a 2 b 2 c c b 2 b b c b c 2 c c a

2 c 2 c 2 a c 2 b 2 c b b 2 b 2 b c 2 c 2 a 2 a a a 2 c c c c b 2 c a c

2 c 2 b 2 c c 2 c 2 a b c 2 c 2 b b 2 b 2 a 2 c c a 2 a a c a c 2 c c b

2 b 2 b 2 b b 2 c 2 c c c 2 a 2 c b 2 c 2 c 2 c b c 2 a c a a a 2 a c c

2 b 2 c 2 c b 2 b 2 b c b 2 c 2 c c 2 a 2 c 2 a c c 2 c b a c c 2 a a a

2

666666666666666666666664

3

777777777777777777777775

pound de= 0 (8)

where the results for each Gauss point are grouped by rows and

a =1

8(1 + f )(1 + f ) b =

1

8(1 2 f )(1 2 f ) c =

1

8(1 + f )(1 2 f )

and f =

3

p

3

(9)

Equation (8) contains a matrix (Q) of rank seven Being 24 the total number ofdegrees of freedom ie the dimension of the subspace of admissible solutionsU h there will be a dimension 242 7 = 17 for the incompressible deformationssubspace

rank(Q) = 7 ^ nullity(Q) = 17 (10)

If for the same element a reduced numerical integration scheme with only oneGauss point is used (x = Z = z = 0) the imposition of condition (7) will lead toa subspace of dimension 23 when the maximum dimension of the space of theadmissible solutions is 24 The matrix in equation (8) is rewritten as follows

1

82 1 2 1 2 1 1 2 1 2 1 1 1 2 1 2 1 1 2 1 2 1 2 1 1 1 2 1 1 1 1 1 2 1 1 1

pound curren

pound de = 0(11)

and

EC207

902

rank(Q) = 1 ^ nullity(Q)= 23 (12)

Analysing these two possible bases for the subspace of the incompressibledeformations it can be inferred clearly that the use of the reduced integrationallows the reproduction of more six incompressible displacement modes thanthe case of complete integration Since the admissible incompressible solutionsare nothing more than a linear combination of a given I h basis this can be aquite reasonable explanation for the good performance of the reducedintegration techniques in volumetric locking problems and why the classicalcomplete integration numerical integration locks

For a clear illustration of the last statements a graphical representation ofthe linearly independent elements which form the basis of the incompressibledeformations subspace is shown in Figure 2 The six rigid body motions can beobtained linearly combining these elements The displacement regeld associatedwith each mode is plotted in Table I The modes 1-17 are reproduced both bythe complete and the reduced integration The modes 18-23 are reproduced onlyby the reduced integration being volumetric locking situations for the completeintegration They can be divided into four main groups

(1) simple edges translations in x y and z directions represented by themodes 1-12

(2) expansioncontraction of one face modes 13-17

(3) hourglass modes 18-20

(4) warping modes 21-23

In the following sections the proposed approaches to alleviate locking effects(using the enhanced strain method) are conveniently deregned within thesubspace analysis framework just described

Formulation of volumetric and transverse shear locking-freeelementsAs already referred the reduced integration selective reduced integration andsimilar techniques have proved to be efregcient methods on attenuating thevolumetric locking However the behaviour in bending dominated situations isnot the best and for the case of the SRI techniques the analysis is in somemeasure limited to materials where the stress tensor allows the decompositionin volumetric and deviatoric components

On the other hand the enhanced strain method is a powerful techniquein the sense that permits the inclusion of more or less additional variablesfor the enhanced strain regeld Thus it is possible to construct a formulationwith good behaviour both in near incompressible situation and bendingsituations

However the number of additional variables in the enhanced regeld is acrucial matter If increasing the number of additional variables and

A new enhancedstrain element

903

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 4: A new volumetric and shear locking-free 3D enhanced strain element

locking phenomenon relates to the inability of a given element to reproduce anull transverse shear strain energy state in pure bending (CeAcircsar de SaAcirc et al2002) Backward in the earlier developments shear locking in degenerated shellelements were revealed as thickness values tended to small values The regrstsuccessful procedures to overcome this artiregcial behaviour were the citedselective numerical under-integration of the transverse shear strain termsThe appearance of spurious zero energy deformation modes required thedevelopment of stabilization or projection techniques in order to recover thecorrect rank of the stiffness matrices involved (Belytschko et al 1992) In casethe full quadrature is maintained some procedures proved to be efregcient inattenuating transverse shear locking Among the successful techniques werethe so-called mixed interpolation of tensorial (transverse shear strain)components (MITC) from the initial work of Dvorkin and Bathe (1984) andlater on Bathe and Dvorkin (1986) Bathe et al (2000) Chapelle and Bathe (2000)and Lee and Bathe (2002)

Additionally it is also worth noting the assumed natural strain (ANS)approach regrst introduced by Park (1986) and Park and Stanley (1986) Bothformulations employ a set of additional sampling points over a regnite element inorder to obtain a substitute or complementary strain regeld leading to thefulreglment of the Kirchhoff-Koiter hypothesis in the thin shell limit Relating tothe MITC this procedure is usually cast into a more general frame than theconventional degenerated formulation An example is the ordfsolid-shellordm class ofelements developed by Doll et al (2000) Harnau and Schweizerhof (2002)Hauptmann and Schweizerhof (1998) and Hauptmann et al (2001) In thisformulation mixed interpolation as described was used for the transverseshear locking while the EAS method is adopted in attenuating membranelocking Thickness volumetric and trapezoidal locking appearance is related tothese elements and proper treatments of the last locking effects are needed inorder to reach a successful and general regnite element (Harnau andSchweizerhof 2002)

In this paper based on the framework of subspace of deformations analysisalready successfully applied in 2D plane strain (CeAcircsar de SaAcirc and Natal Jorge1999) and shell problems (CeAcircsar de SaAcirc et al 2002) a new volumetric and shearlocking-free EAS element for 3D analyses is proposed The main idea is theimprovement of the original strain regeld in an additive way leading to elementswith higher performance in the presence of volumetric locking (3Dapplications) and transverse shear locking (shell structures applications)

Volumetric locking plusmn subspace analysisThe incompressibility problem can be formulated as a constrainedminimization of a functional (CeAcircsar de SaAcirc and Natal Jorge 1999) In simpleterms the goal is to obtain in the linear space of admissible solutions U a regeldof displacement u that minimizes the total energy of the system located in the

A new enhancedstrain element

899

subspace of the incompressible deformations I and simultaneously containedin the space of all the solutions (I U) This statement can be posed in theform

I = u [ U div(u) = 0 (1)

In an approach done by the FEM the linear space of the admissible solutions Uand the respective subspace I previously deregned are approximated by thespaces U h and I h respectively

A two regeld regnite element solution can be expressed for linear elasticityaccording to CeAcircsar de SaAcirc and Natal Jorge (1999) by

K Q

2 Q 0

uh

ph

=fhext

0

(2)

where fext is the vector of applied external forces p is the hydrostatic pressureand K is the stiffness matrix The superscript (h) means that the variable incause is a regnite element approximation

The incompressibility condition is given by the second group of equationsderegned in equation (2)

Quh = 0 (3)

which will deregne the subspace of the incompressible deformations I h as

I h = uh [ U h Quh = 0 (4)

In order to avoid the trivial solution (uh= 0) in equation (3) the regeld ofdisplacements uh should belong to the nullspace of Q that is to the subspaceof the incompressible deformations I h The approximated displacements uh

are contained in I h being therefore a linear combination of a given basis of I h

elementsIf the subspace I h is an approach to the original subspace I it is plausible to

admit that it cannot reproduce all the possible solutions contained in I In factdifferent formulations lead to better or worse approximations for the I subspaceThe volumetric locking phenomenon occurs when for a certain group ofboundary conditions and external forces applied in a near incompressiblesituation the expected solution or some of its components do not appearproperly represented in the subspace I h Different types of elements can offerdistinct subspaces of incompressible deformations originating thereforedifferent solutions (CeAcircsar de SaAcirc and Owen 1986)

To characterize the already deregned subspace I h consider a standardisoparametric eight-node hexahedral element with domain Oe (Figure 1)For small deformations the incompressibility constraint (e ii = 0) can takethe form

EC207

900

Z

Oe

div(u) dOe =Z

Oe

shy u

shy x+

shy v

shy Z+

shy w

shy z

sup3 acutedOe = 0 (5)

At the element level in any point of the domain the displacement regeld can beinterpolated using the standard linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000)

u lt uhe = N(x Z z )de (6)

One way to guarantee the incompressibility condition is to assure that theintegrand function in equation (5) is zero Substituting equation (6) inequation (5) results

shy u

shy x+

shy v

shy Z+

shy w

shy z= [N ix(x Z z ) N iZ(x Z z ) N iz(x Z z )] de = 0 (7)

for i = 1 nenodes Making use of a complete integration (2 pound 2 pound 2 eight Gauss

points Figure 1) the application of equation (7) leads to

Figure 1Standard eight-node

isoparametrichexahedral element with

2 pound 2 pound 2 Gauss pointsnumerical integration

rule

A new enhancedstrain element

901

2 a 2 a 2 a a 2 c 2 c c c 2 b 2 c a 2 c 2 c 2 c a c 2 b c b b b 2 b c c

2 a 2 c 2 c a 2 a 2 a c a 2 c 2 c c 2 b 2 c 2 b c c 2 c a b c c 2 b b b

2 c 2 c 2 b c 2 a 2 c a a 2 a 2 a c 2 c 2 b 2 b b b 2 c c c c a 2 c b c

2 c 2 a 2 c c 2 c 2 b a c 2 c 2 a a 2 a 2 b 2 c c b 2 b b c b c 2 c c a

2 c 2 c 2 a c 2 b 2 c b b 2 b 2 b c 2 c 2 a 2 a a a 2 c c c c b 2 c a c

2 c 2 b 2 c c 2 c 2 a b c 2 c 2 b b 2 b 2 a 2 c c a 2 a a c a c 2 c c b

2 b 2 b 2 b b 2 c 2 c c c 2 a 2 c b 2 c 2 c 2 c b c 2 a c a a a 2 a c c

2 b 2 c 2 c b 2 b 2 b c b 2 c 2 c c 2 a 2 c 2 a c c 2 c b a c c 2 a a a

2

666666666666666666666664

3

777777777777777777777775

pound de= 0 (8)

where the results for each Gauss point are grouped by rows and

a =1

8(1 + f )(1 + f ) b =

1

8(1 2 f )(1 2 f ) c =

1

8(1 + f )(1 2 f )

and f =

3

p

3

(9)

Equation (8) contains a matrix (Q) of rank seven Being 24 the total number ofdegrees of freedom ie the dimension of the subspace of admissible solutionsU h there will be a dimension 242 7 = 17 for the incompressible deformationssubspace

rank(Q) = 7 ^ nullity(Q) = 17 (10)

If for the same element a reduced numerical integration scheme with only oneGauss point is used (x = Z = z = 0) the imposition of condition (7) will lead toa subspace of dimension 23 when the maximum dimension of the space of theadmissible solutions is 24 The matrix in equation (8) is rewritten as follows

1

82 1 2 1 2 1 1 2 1 2 1 1 1 2 1 2 1 1 2 1 2 1 2 1 1 1 2 1 1 1 1 1 2 1 1 1

pound curren

pound de = 0(11)

and

EC207

902

rank(Q) = 1 ^ nullity(Q)= 23 (12)

Analysing these two possible bases for the subspace of the incompressibledeformations it can be inferred clearly that the use of the reduced integrationallows the reproduction of more six incompressible displacement modes thanthe case of complete integration Since the admissible incompressible solutionsare nothing more than a linear combination of a given I h basis this can be aquite reasonable explanation for the good performance of the reducedintegration techniques in volumetric locking problems and why the classicalcomplete integration numerical integration locks

For a clear illustration of the last statements a graphical representation ofthe linearly independent elements which form the basis of the incompressibledeformations subspace is shown in Figure 2 The six rigid body motions can beobtained linearly combining these elements The displacement regeld associatedwith each mode is plotted in Table I The modes 1-17 are reproduced both bythe complete and the reduced integration The modes 18-23 are reproduced onlyby the reduced integration being volumetric locking situations for the completeintegration They can be divided into four main groups

(1) simple edges translations in x y and z directions represented by themodes 1-12

(2) expansioncontraction of one face modes 13-17

(3) hourglass modes 18-20

(4) warping modes 21-23

In the following sections the proposed approaches to alleviate locking effects(using the enhanced strain method) are conveniently deregned within thesubspace analysis framework just described

Formulation of volumetric and transverse shear locking-freeelementsAs already referred the reduced integration selective reduced integration andsimilar techniques have proved to be efregcient methods on attenuating thevolumetric locking However the behaviour in bending dominated situations isnot the best and for the case of the SRI techniques the analysis is in somemeasure limited to materials where the stress tensor allows the decompositionin volumetric and deviatoric components

On the other hand the enhanced strain method is a powerful techniquein the sense that permits the inclusion of more or less additional variablesfor the enhanced strain regeld Thus it is possible to construct a formulationwith good behaviour both in near incompressible situation and bendingsituations

However the number of additional variables in the enhanced regeld is acrucial matter If increasing the number of additional variables and

A new enhancedstrain element

903

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 5: A new volumetric and shear locking-free 3D enhanced strain element

subspace of the incompressible deformations I and simultaneously containedin the space of all the solutions (I U) This statement can be posed in theform

I = u [ U div(u) = 0 (1)

In an approach done by the FEM the linear space of the admissible solutions Uand the respective subspace I previously deregned are approximated by thespaces U h and I h respectively

A two regeld regnite element solution can be expressed for linear elasticityaccording to CeAcircsar de SaAcirc and Natal Jorge (1999) by

K Q

2 Q 0

uh

ph

=fhext

0

(2)

where fext is the vector of applied external forces p is the hydrostatic pressureand K is the stiffness matrix The superscript (h) means that the variable incause is a regnite element approximation

The incompressibility condition is given by the second group of equationsderegned in equation (2)

Quh = 0 (3)

which will deregne the subspace of the incompressible deformations I h as

I h = uh [ U h Quh = 0 (4)

In order to avoid the trivial solution (uh= 0) in equation (3) the regeld ofdisplacements uh should belong to the nullspace of Q that is to the subspaceof the incompressible deformations I h The approximated displacements uh

are contained in I h being therefore a linear combination of a given basis of I h

elementsIf the subspace I h is an approach to the original subspace I it is plausible to

admit that it cannot reproduce all the possible solutions contained in I In factdifferent formulations lead to better or worse approximations for the I subspaceThe volumetric locking phenomenon occurs when for a certain group ofboundary conditions and external forces applied in a near incompressiblesituation the expected solution or some of its components do not appearproperly represented in the subspace I h Different types of elements can offerdistinct subspaces of incompressible deformations originating thereforedifferent solutions (CeAcircsar de SaAcirc and Owen 1986)

To characterize the already deregned subspace I h consider a standardisoparametric eight-node hexahedral element with domain Oe (Figure 1)For small deformations the incompressibility constraint (e ii = 0) can takethe form

EC207

900

Z

Oe

div(u) dOe =Z

Oe

shy u

shy x+

shy v

shy Z+

shy w

shy z

sup3 acutedOe = 0 (5)

At the element level in any point of the domain the displacement regeld can beinterpolated using the standard linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000)

u lt uhe = N(x Z z )de (6)

One way to guarantee the incompressibility condition is to assure that theintegrand function in equation (5) is zero Substituting equation (6) inequation (5) results

shy u

shy x+

shy v

shy Z+

shy w

shy z= [N ix(x Z z ) N iZ(x Z z ) N iz(x Z z )] de = 0 (7)

for i = 1 nenodes Making use of a complete integration (2 pound 2 pound 2 eight Gauss

points Figure 1) the application of equation (7) leads to

Figure 1Standard eight-node

isoparametrichexahedral element with

2 pound 2 pound 2 Gauss pointsnumerical integration

rule

A new enhancedstrain element

901

2 a 2 a 2 a a 2 c 2 c c c 2 b 2 c a 2 c 2 c 2 c a c 2 b c b b b 2 b c c

2 a 2 c 2 c a 2 a 2 a c a 2 c 2 c c 2 b 2 c 2 b c c 2 c a b c c 2 b b b

2 c 2 c 2 b c 2 a 2 c a a 2 a 2 a c 2 c 2 b 2 b b b 2 c c c c a 2 c b c

2 c 2 a 2 c c 2 c 2 b a c 2 c 2 a a 2 a 2 b 2 c c b 2 b b c b c 2 c c a

2 c 2 c 2 a c 2 b 2 c b b 2 b 2 b c 2 c 2 a 2 a a a 2 c c c c b 2 c a c

2 c 2 b 2 c c 2 c 2 a b c 2 c 2 b b 2 b 2 a 2 c c a 2 a a c a c 2 c c b

2 b 2 b 2 b b 2 c 2 c c c 2 a 2 c b 2 c 2 c 2 c b c 2 a c a a a 2 a c c

2 b 2 c 2 c b 2 b 2 b c b 2 c 2 c c 2 a 2 c 2 a c c 2 c b a c c 2 a a a

2

666666666666666666666664

3

777777777777777777777775

pound de= 0 (8)

where the results for each Gauss point are grouped by rows and

a =1

8(1 + f )(1 + f ) b =

1

8(1 2 f )(1 2 f ) c =

1

8(1 + f )(1 2 f )

and f =

3

p

3

(9)

Equation (8) contains a matrix (Q) of rank seven Being 24 the total number ofdegrees of freedom ie the dimension of the subspace of admissible solutionsU h there will be a dimension 242 7 = 17 for the incompressible deformationssubspace

rank(Q) = 7 ^ nullity(Q) = 17 (10)

If for the same element a reduced numerical integration scheme with only oneGauss point is used (x = Z = z = 0) the imposition of condition (7) will lead toa subspace of dimension 23 when the maximum dimension of the space of theadmissible solutions is 24 The matrix in equation (8) is rewritten as follows

1

82 1 2 1 2 1 1 2 1 2 1 1 1 2 1 2 1 1 2 1 2 1 2 1 1 1 2 1 1 1 1 1 2 1 1 1

pound curren

pound de = 0(11)

and

EC207

902

rank(Q) = 1 ^ nullity(Q)= 23 (12)

Analysing these two possible bases for the subspace of the incompressibledeformations it can be inferred clearly that the use of the reduced integrationallows the reproduction of more six incompressible displacement modes thanthe case of complete integration Since the admissible incompressible solutionsare nothing more than a linear combination of a given I h basis this can be aquite reasonable explanation for the good performance of the reducedintegration techniques in volumetric locking problems and why the classicalcomplete integration numerical integration locks

For a clear illustration of the last statements a graphical representation ofthe linearly independent elements which form the basis of the incompressibledeformations subspace is shown in Figure 2 The six rigid body motions can beobtained linearly combining these elements The displacement regeld associatedwith each mode is plotted in Table I The modes 1-17 are reproduced both bythe complete and the reduced integration The modes 18-23 are reproduced onlyby the reduced integration being volumetric locking situations for the completeintegration They can be divided into four main groups

(1) simple edges translations in x y and z directions represented by themodes 1-12

(2) expansioncontraction of one face modes 13-17

(3) hourglass modes 18-20

(4) warping modes 21-23

In the following sections the proposed approaches to alleviate locking effects(using the enhanced strain method) are conveniently deregned within thesubspace analysis framework just described

Formulation of volumetric and transverse shear locking-freeelementsAs already referred the reduced integration selective reduced integration andsimilar techniques have proved to be efregcient methods on attenuating thevolumetric locking However the behaviour in bending dominated situations isnot the best and for the case of the SRI techniques the analysis is in somemeasure limited to materials where the stress tensor allows the decompositionin volumetric and deviatoric components

On the other hand the enhanced strain method is a powerful techniquein the sense that permits the inclusion of more or less additional variablesfor the enhanced strain regeld Thus it is possible to construct a formulationwith good behaviour both in near incompressible situation and bendingsituations

However the number of additional variables in the enhanced regeld is acrucial matter If increasing the number of additional variables and

A new enhancedstrain element

903

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 6: A new volumetric and shear locking-free 3D enhanced strain element

Z

Oe

div(u) dOe =Z

Oe

shy u

shy x+

shy v

shy Z+

shy w

shy z

sup3 acutedOe = 0 (5)

At the element level in any point of the domain the displacement regeld can beinterpolated using the standard linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000)

u lt uhe = N(x Z z )de (6)

One way to guarantee the incompressibility condition is to assure that theintegrand function in equation (5) is zero Substituting equation (6) inequation (5) results

shy u

shy x+

shy v

shy Z+

shy w

shy z= [N ix(x Z z ) N iZ(x Z z ) N iz(x Z z )] de = 0 (7)

for i = 1 nenodes Making use of a complete integration (2 pound 2 pound 2 eight Gauss

points Figure 1) the application of equation (7) leads to

Figure 1Standard eight-node

isoparametrichexahedral element with

2 pound 2 pound 2 Gauss pointsnumerical integration

rule

A new enhancedstrain element

901

2 a 2 a 2 a a 2 c 2 c c c 2 b 2 c a 2 c 2 c 2 c a c 2 b c b b b 2 b c c

2 a 2 c 2 c a 2 a 2 a c a 2 c 2 c c 2 b 2 c 2 b c c 2 c a b c c 2 b b b

2 c 2 c 2 b c 2 a 2 c a a 2 a 2 a c 2 c 2 b 2 b b b 2 c c c c a 2 c b c

2 c 2 a 2 c c 2 c 2 b a c 2 c 2 a a 2 a 2 b 2 c c b 2 b b c b c 2 c c a

2 c 2 c 2 a c 2 b 2 c b b 2 b 2 b c 2 c 2 a 2 a a a 2 c c c c b 2 c a c

2 c 2 b 2 c c 2 c 2 a b c 2 c 2 b b 2 b 2 a 2 c c a 2 a a c a c 2 c c b

2 b 2 b 2 b b 2 c 2 c c c 2 a 2 c b 2 c 2 c 2 c b c 2 a c a a a 2 a c c

2 b 2 c 2 c b 2 b 2 b c b 2 c 2 c c 2 a 2 c 2 a c c 2 c b a c c 2 a a a

2

666666666666666666666664

3

777777777777777777777775

pound de= 0 (8)

where the results for each Gauss point are grouped by rows and

a =1

8(1 + f )(1 + f ) b =

1

8(1 2 f )(1 2 f ) c =

1

8(1 + f )(1 2 f )

and f =

3

p

3

(9)

Equation (8) contains a matrix (Q) of rank seven Being 24 the total number ofdegrees of freedom ie the dimension of the subspace of admissible solutionsU h there will be a dimension 242 7 = 17 for the incompressible deformationssubspace

rank(Q) = 7 ^ nullity(Q) = 17 (10)

If for the same element a reduced numerical integration scheme with only oneGauss point is used (x = Z = z = 0) the imposition of condition (7) will lead toa subspace of dimension 23 when the maximum dimension of the space of theadmissible solutions is 24 The matrix in equation (8) is rewritten as follows

1

82 1 2 1 2 1 1 2 1 2 1 1 1 2 1 2 1 1 2 1 2 1 2 1 1 1 2 1 1 1 1 1 2 1 1 1

pound curren

pound de = 0(11)

and

EC207

902

rank(Q) = 1 ^ nullity(Q)= 23 (12)

Analysing these two possible bases for the subspace of the incompressibledeformations it can be inferred clearly that the use of the reduced integrationallows the reproduction of more six incompressible displacement modes thanthe case of complete integration Since the admissible incompressible solutionsare nothing more than a linear combination of a given I h basis this can be aquite reasonable explanation for the good performance of the reducedintegration techniques in volumetric locking problems and why the classicalcomplete integration numerical integration locks

For a clear illustration of the last statements a graphical representation ofthe linearly independent elements which form the basis of the incompressibledeformations subspace is shown in Figure 2 The six rigid body motions can beobtained linearly combining these elements The displacement regeld associatedwith each mode is plotted in Table I The modes 1-17 are reproduced both bythe complete and the reduced integration The modes 18-23 are reproduced onlyby the reduced integration being volumetric locking situations for the completeintegration They can be divided into four main groups

(1) simple edges translations in x y and z directions represented by themodes 1-12

(2) expansioncontraction of one face modes 13-17

(3) hourglass modes 18-20

(4) warping modes 21-23

In the following sections the proposed approaches to alleviate locking effects(using the enhanced strain method) are conveniently deregned within thesubspace analysis framework just described

Formulation of volumetric and transverse shear locking-freeelementsAs already referred the reduced integration selective reduced integration andsimilar techniques have proved to be efregcient methods on attenuating thevolumetric locking However the behaviour in bending dominated situations isnot the best and for the case of the SRI techniques the analysis is in somemeasure limited to materials where the stress tensor allows the decompositionin volumetric and deviatoric components

On the other hand the enhanced strain method is a powerful techniquein the sense that permits the inclusion of more or less additional variablesfor the enhanced strain regeld Thus it is possible to construct a formulationwith good behaviour both in near incompressible situation and bendingsituations

However the number of additional variables in the enhanced regeld is acrucial matter If increasing the number of additional variables and

A new enhancedstrain element

903

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 7: A new volumetric and shear locking-free 3D enhanced strain element

2 a 2 a 2 a a 2 c 2 c c c 2 b 2 c a 2 c 2 c 2 c a c 2 b c b b b 2 b c c

2 a 2 c 2 c a 2 a 2 a c a 2 c 2 c c 2 b 2 c 2 b c c 2 c a b c c 2 b b b

2 c 2 c 2 b c 2 a 2 c a a 2 a 2 a c 2 c 2 b 2 b b b 2 c c c c a 2 c b c

2 c 2 a 2 c c 2 c 2 b a c 2 c 2 a a 2 a 2 b 2 c c b 2 b b c b c 2 c c a

2 c 2 c 2 a c 2 b 2 c b b 2 b 2 b c 2 c 2 a 2 a a a 2 c c c c b 2 c a c

2 c 2 b 2 c c 2 c 2 a b c 2 c 2 b b 2 b 2 a 2 c c a 2 a a c a c 2 c c b

2 b 2 b 2 b b 2 c 2 c c c 2 a 2 c b 2 c 2 c 2 c b c 2 a c a a a 2 a c c

2 b 2 c 2 c b 2 b 2 b c b 2 c 2 c c 2 a 2 c 2 a c c 2 c b a c c 2 a a a

2

666666666666666666666664

3

777777777777777777777775

pound de= 0 (8)

where the results for each Gauss point are grouped by rows and

a =1

8(1 + f )(1 + f ) b =

1

8(1 2 f )(1 2 f ) c =

1

8(1 + f )(1 2 f )

and f =

3

p

3

(9)

Equation (8) contains a matrix (Q) of rank seven Being 24 the total number ofdegrees of freedom ie the dimension of the subspace of admissible solutionsU h there will be a dimension 242 7 = 17 for the incompressible deformationssubspace

rank(Q) = 7 ^ nullity(Q) = 17 (10)

If for the same element a reduced numerical integration scheme with only oneGauss point is used (x = Z = z = 0) the imposition of condition (7) will lead toa subspace of dimension 23 when the maximum dimension of the space of theadmissible solutions is 24 The matrix in equation (8) is rewritten as follows

1

82 1 2 1 2 1 1 2 1 2 1 1 1 2 1 2 1 1 2 1 2 1 2 1 1 1 2 1 1 1 1 1 2 1 1 1

pound curren

pound de = 0(11)

and

EC207

902

rank(Q) = 1 ^ nullity(Q)= 23 (12)

Analysing these two possible bases for the subspace of the incompressibledeformations it can be inferred clearly that the use of the reduced integrationallows the reproduction of more six incompressible displacement modes thanthe case of complete integration Since the admissible incompressible solutionsare nothing more than a linear combination of a given I h basis this can be aquite reasonable explanation for the good performance of the reducedintegration techniques in volumetric locking problems and why the classicalcomplete integration numerical integration locks

For a clear illustration of the last statements a graphical representation ofthe linearly independent elements which form the basis of the incompressibledeformations subspace is shown in Figure 2 The six rigid body motions can beobtained linearly combining these elements The displacement regeld associatedwith each mode is plotted in Table I The modes 1-17 are reproduced both bythe complete and the reduced integration The modes 18-23 are reproduced onlyby the reduced integration being volumetric locking situations for the completeintegration They can be divided into four main groups

(1) simple edges translations in x y and z directions represented by themodes 1-12

(2) expansioncontraction of one face modes 13-17

(3) hourglass modes 18-20

(4) warping modes 21-23

In the following sections the proposed approaches to alleviate locking effects(using the enhanced strain method) are conveniently deregned within thesubspace analysis framework just described

Formulation of volumetric and transverse shear locking-freeelementsAs already referred the reduced integration selective reduced integration andsimilar techniques have proved to be efregcient methods on attenuating thevolumetric locking However the behaviour in bending dominated situations isnot the best and for the case of the SRI techniques the analysis is in somemeasure limited to materials where the stress tensor allows the decompositionin volumetric and deviatoric components

On the other hand the enhanced strain method is a powerful techniquein the sense that permits the inclusion of more or less additional variablesfor the enhanced strain regeld Thus it is possible to construct a formulationwith good behaviour both in near incompressible situation and bendingsituations

However the number of additional variables in the enhanced regeld is acrucial matter If increasing the number of additional variables and

A new enhancedstrain element

903

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 8: A new volumetric and shear locking-free 3D enhanced strain element

rank(Q) = 1 ^ nullity(Q)= 23 (12)

Analysing these two possible bases for the subspace of the incompressibledeformations it can be inferred clearly that the use of the reduced integrationallows the reproduction of more six incompressible displacement modes thanthe case of complete integration Since the admissible incompressible solutionsare nothing more than a linear combination of a given I h basis this can be aquite reasonable explanation for the good performance of the reducedintegration techniques in volumetric locking problems and why the classicalcomplete integration numerical integration locks

For a clear illustration of the last statements a graphical representation ofthe linearly independent elements which form the basis of the incompressibledeformations subspace is shown in Figure 2 The six rigid body motions can beobtained linearly combining these elements The displacement regeld associatedwith each mode is plotted in Table I The modes 1-17 are reproduced both bythe complete and the reduced integration The modes 18-23 are reproduced onlyby the reduced integration being volumetric locking situations for the completeintegration They can be divided into four main groups

(1) simple edges translations in x y and z directions represented by themodes 1-12

(2) expansioncontraction of one face modes 13-17

(3) hourglass modes 18-20

(4) warping modes 21-23

In the following sections the proposed approaches to alleviate locking effects(using the enhanced strain method) are conveniently deregned within thesubspace analysis framework just described

Formulation of volumetric and transverse shear locking-freeelementsAs already referred the reduced integration selective reduced integration andsimilar techniques have proved to be efregcient methods on attenuating thevolumetric locking However the behaviour in bending dominated situations isnot the best and for the case of the SRI techniques the analysis is in somemeasure limited to materials where the stress tensor allows the decompositionin volumetric and deviatoric components

On the other hand the enhanced strain method is a powerful techniquein the sense that permits the inclusion of more or less additional variablesfor the enhanced strain regeld Thus it is possible to construct a formulationwith good behaviour both in near incompressible situation and bendingsituations

However the number of additional variables in the enhanced regeld is acrucial matter If increasing the number of additional variables and

A new enhancedstrain element

903

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 9: A new volumetric and shear locking-free 3D enhanced strain element

consequently modes of deformation normally improves the elementrsquosperformance it can also lead to instabilities numerical inefregciency and largeCPU costs Using the classical complete (eightw Gauss points) numericalintegration the objective turns therefore to reach the dimension 23for the subspace of incompressible deformations It was already shown

Figure 2Basic components of theincompressibledeformations subspace

EC207

904

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 10: A new volumetric and shear locking-free 3D enhanced strain element

(Alves de Sousa et al 2002) that this subspace dimension is related to a good

performance in 3D volumetric locking situations

At the element level the interpolation of the strain regeld follows the usualmethodology (Simo and Rifai 1990)

Figure 2

A new enhancedstrain element

905

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 11: A new volumetric and shear locking-free 3D enhanced strain element

m=

u

=1

v 1w

1u

2v 2

w2

u3

v 3w

3u

4v 4

w4

u5

v 5w

5u

6v 6

w6

u7

v 7w

7u

8v 8

w8T

m1

=1

00

10

00

00

00

00

00

00

00

00

00

0T

m2

=0

00

00

00

00

00

01

00

10

00

00

00

0T

m3

=0

00

00

01

00

10

00

00

00

00

00

00

0T

m4

=0

00

00

00

00

00

00

00

00

01

00

10

0T

m5

=0

10

00

00

00

01

00

00

00

00

00

00

0T

m6

=0

00

01

00

10

00

00

00

00

00

00

00

0T

m7

=0

00

00

00

00

00

00

10

00

00

00

01

0T

m8

=0

00

00

00

00

00

00

00

01

00

10

00

0T

m9

=0

01

00

00

00

00

00

01

00

00

00

00

0T

m10=

00

00

01

00

00

00

00

00

01

00

00

00

T

m11=

00

00

00

00

00

01

00

00

00

00

00

01

T

m12=

00

00

00

00

10

00

00

00

00

00

10

00

T

m13=

21

01

10

10

00

00

02

10

21

10

21

00

00

00

T

m14=

00

00

00

10

12

10

10

00

00

01

02

12

10

21

T

m15=

00

00

00

00

00

00

12

10

21

21

02

11

01

10

T

m16=

12

10

21

21

02

11

01

10

00

00

00

00

00

00

T

m17=

02

11

00

00

00

01

10

21

21

00

00

00

01

21

T

m18=

00

10

02

10

02

10

01

00

21

00

10

01

00

21

T

m19=

10

02

10

02

10

01

00

21

00

10

01

00

21

00

T

m20=

21

00

00

02

10

00

00

21

00

00

00

00

10

0T

m21=

00

00

00

00

00

00

00

10

02

10

01

00

21

T

m22=

00

01

00

21

00

00

00

00

21

00

10

00

00

T

m23=

00

00

10

00

00

10

01

00

21

00

00

00

0T

Table IBasic components of theincompressibledeformations subspace

EC207

906

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 12: A new volumetric and shear locking-free 3D enhanced strain element

e = ed + ea = Bed Be

a

h i de

a e

= Bede

(13)

where Bed is the standard FEM strain-displacement differential operator

function of the standard FEM linear shape functions (Hughes 2000Zienkiewicz and Taylor 2000) while Be

a is the one used for the enhancedvariables regeld (a)

At the element level the operator Bea is previously formulated in the

isoparametric space leading to the Mea matrix The transition from the

isoparametric space to the global reference frame must be carried out verycarefully in order to satisfy the patch test for arbitrary conreggurations asstated by Taylor et al (1976)

Z

OMa dO = 0 (14)

Condition (14) can also be obtained from the orthogonality condition betweenthe stress regeld and the enhanced strain regeld (Simo and Rifai 1990)

In problems involving distorted meshes the transition of the differentialoperator for the enhanced regeld is also a crucial matter and can have a relevantinmacruence in the elementrsquos performance Following the work of Andelregnger andRamm (1993) and Simo and Rifai (1990) the transformation provided byequation (15) performed at the center of each regnite element allows thefulregllment of equation (14) by the elements proposed and can be stated asfollows

Bea =

j J0 jj J j

T0Mea (15)

and as usual

J =

shy x

shy xshy y

shy xshy z

shy x

shy x

shy Zshy y

shy Zshy z

shy Z

shy x

shy zshy y

shy zshy z

shy z

2

666666664

3

777777775

(16)

j J0 j and j J j being Jacobian matrix determinants and a the number of additionalvariables The subscript (0) refers to evaluations in the centre of the standardelement (x = Z = z = 0) T is the transformation tensor relating theisoparametric space and the global reference frame deregned by

A new enhancedstrain element

907

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 13: A new volumetric and shear locking-free 3D enhanced strain element

T =

J 2 111 J 2 1

11 J 2 112 J 2 1

12 J 2 113 J 2 1

13 J 2 111 J 2 1

12 J 2 111 J 2 1

13 J 2 112 J 2 1

13

J 2 121 J 2 1

21 J 2 122 J 2 1

22 J 2 123 J 2 1

23 J 2 121 J 2 1

22 J 2 121 J 2 1

23 J 2 122 J 2 1

23

J 2 131 J 2 1

31 J 2 132 J 2 1

32 J 2 133 J 2 1

33 J 2 131 J 2 1

32 J 2 131 J 2 1

33 J 2 132 J 2 1

33

2J 2 111 J 2 1

21 2J 2 112 J 2 1

22 2J 2 113 J 2 1

23 ( J 2 111 J 2 1

22 )+ ( J 2 112 J 2 1

21 ) ( J 2 111 J 2 1

23 )+ ( J 2 121 J 2 1

13 ) ( J 2 112 J 2 1

23 )+ ( J 2 122 J 2 1

13 )

2J 2 111 J 2 1

31 2J 2 112 J 2 1

32 2J 2 113 J 2 1

33 ( J 2 111 J 2 1

32 )+ ( J 2 112 J 2 1

31 ) ( J 2 111 J 2 1

33 )+ ( J 2 131 J 2 1

13 ) ( J 2 112 J 2 1

33 )+ ( J 2 132 J 2 1

13 )

2J 2 121 J 2 1

31 2J 2 122 J 2 1

32 2J 2 123 J 2 1

33 ( J 2 121 J 2 1

32 )+ ( J 2 122 J 2 1

31 ) ( J 2 121 J 2 1

33 )+ ( J 2 131 J 2 1

23 ) ( J 2 122 J 2 1

33 )+ ( J 2 132 J 2 1

23 )

2

66666666666666666666664

3

77777777777777777777775

(17)

where J 2 1ij relates to the ij components of the inverse Jacobian matrix This

expression comes from considering a transformation between two generalcurvilinear (convective) referentials (Bathe 1996)

The starting point for the following developments is to include nineadditional variables in the enhanced strain regeld each one associated with thespace derivatives of the displacement regeld H Thus for a 3D problem thedisplacement gradient matrix is augmented as follows

H =

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

+

shy u

shy x

shy u

shy y

shy u

shy z

shy v

shy x

shy v

shy y

shy v

shy z

shy z

shy x

shy z

shy y

shy z

shy z

2

6666666664

3

7777777775

(18)

Making use of a bubble function Na

N a =1

2(1 2 x2)(1 2 Z2)(1 2 z 2) (19)

extra compatible modes of deformation are added with the goal of reachingdimension 23 for the subspace I hand resulting in the enhanced strain regeldinterpolation matrix denoted by Me

9 deregned in the local referential Thisincreases the total elementary degrees of freedom number from 24 to 33

EC207

908

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 14: A new volumetric and shear locking-free 3D enhanced strain element

Me9 =

shy N ashy x 0 0 0 0 0 0 0 0

0shy N a

shy Z0 0 0 0 0 0 0

0 0shy N a

shy z0 0 0 0 0 0

0 0 0shy Na

shy xshy N a

shy Z0 0 0 0

0 0 0 0 0shy N a

shy xshy N a

shy z0 0

0 0 0 0 0 0 0shy Na

shy Zshy N a

shy z

2

6666666666666664

3

7777777777777775

(20)

The subsequent transformation to the global frame follows the rule stated inequation (15) This time the application of equation (7) employs the matrix Be

Be = [ Bed Be

9 ] (21)

and results in the following matrix

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

66666666666664

3

77777777777775

de = 0 (22)

where

d = f (1 2 f 2)2 and f =

3

p

3(23)

and

rank(Q) = 7 ^ nullity(Q) = 26 (24)

Following the methodology applied in the subspace analysis for the completeand reduced integration schemes of the previous section it is possible toconclude that expression (22) will implicitly deregne a basis for the subspace ofincompressible deformations with dimension 20 This value comes fromconsidering a number of 24 + 9= 33 degrees of freedom subtracting the rank ofQ and the number of null displacement modes associated with the null columnsin equation (22) (resulting in a dimension of 33 2 7 2 6 = 20) Note that these

A new enhancedstrain element

909

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 15: A new volumetric and shear locking-free 3D enhanced strain element

six neglected modes respect the incompressibility condition but concern onlynon-zero values for the enhanced variables regeld not introducing additionaldisplacement modes This approximated basis can be represented by the regrst20 elements represented in Figure 2

However the previous approach still achieve less three displacement modesthan the use of the reduced integration which may be the cause of volumetriclocking problems for this element (Alves de Sousa et al 2002)

Hence three more new internal variables will be added aiming to be able toreproduce all the modes of Figure 2 which is effectively achieved At the sametime these new variables in this case bilinear terms coming from the doublederivatives in equation (25) assure the condition of incompressibility as statedby de Borst and Groen (1999) Korelc and Wriggers (1996) and Simo et al (1993)

Focusing only on the volumetric components of the strain regeld themodiregcation is performed in the volumetric components of the enhanced regeldinterpolation matrix this time called Me

12

Me12 = Me

9

pound curren

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z

shy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy zshy 2Na

shy xshy Zshy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0

0 0 0

0 0 0

2

6666666666666664

3

7777777777777775

(25)

Replacing Me9 by Me

12 in equation (21) and again applying equation (7) results in

[equation (8)] acute acute acute

d d d

2 d d d

2 d 2 d d

d 2 d d

d d 2 d

2 d d 2 d

2 d 2 d 2 d

d 2 d 2 d

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

e e e

2 e e 2 e

e 2 e 2 e

2 e 2 e e

e 2 e 2 e

2 e 2 e e

e e e

2 e e 2 e

2

666666666666664

3

777777777777775

pound ^de = 0

(26)

where

EC207

910

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 16: A new volumetric and shear locking-free 3D enhanced strain element

e = f 2 + 1 and f =

3

p

3(27)

This will generate an approximation for the subspace of incompressibledeformations with dimension 23 (comes from 362 7 2 6) again neglecting themodes associated with the six internal degrees of freedom in equation (26) aswell in equation (22) which have zero nodal displacements This fulregls thenecessary subspace dimension for a volumetric locking-free element

The HCiS18 elementOnce the requirements for a volumetric locking-free formulation are satisregedwe turn back to the transverse shear locking Indeed the two last elementsshow strong sensitivity to the shear locking phenomenon in problemsinvolving thin shells or plates with low length to thickness ratio (Alves deSousa et al 2002)

Working only in the shear part of the strain regeld to guarantee the samevolumetric locking-free properties already achieved by the use of Me

12 anoptimal choice of six new enhanced shear modes is added resulting in the Me

18interpolation matrix once again function of the bubble function (19)

Me18 = Me

12

pound curren

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy 2N a

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy 2Na

shy xshy Zshy 2N a

shy xshy z

2

666666666666666664

3

777777777777777775

(28)

This element satisreged the patch test (14) and showed a good behaviour involumetric and transverse shear locking-free related problems even withdistorted meshes (Alves de Sousa et al 2002)

A new approach the HCiS12 elementIn an alternative approach it is possible to split the already deregned Me

18(equation (28)) matrix into two different ones The Me

VL matrix grouping onlythe enhanced modes related to the volumetric part and the Me

TSL matrixreferring to the transverse shear enhanced modes

A new enhancedstrain element

911

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 17: A new volumetric and shear locking-free 3D enhanced strain element

MeVL =

shy Na

shy x0 0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0shy N a

shy Z0

shy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0shy N a

shy zshy 2Na

shy xshy Zshy 2N a

shy xshy zshy 2N a

shy Zshy z

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

2

666666666666666664

3

777777777777777775

(29)

MeTSL =

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0

shy Na

shy xshy N a

shy Z0 0 0 0

shy 2Na

shy xshy zshy 2N a

shy Zshy z0 0 0 0

0 0shy Na

shy xshy N a

shy z0 0 0 0

shy 2Na

shy xshy Zshy 2Na

shy Zshy z0 0

0 0 0 0shy N a

shy Zshy Na

shy z0 0 0 0

shy 2N a

shy xshy Zshy 2Na

shy xshy z

2

666666666666666664

3

777777777777777775

(30)

The use of the MeVL matrix is sufregcient to reach the stated dimension for the

incompressible deformations subspace in a volumetric locking-free element ie 23In fact comparing with equation (26) the only difference when applying equation(30) in equation (7) is the elimination of the six null columns in equations (22) and(26) resulting in this case

rank(Q) = 7 ^ nullity(Q) = 23 (31)

Consequently the number of elements in this basis of the incompressibledeformations subspace is equal to the nullity of Q matrix 23 since there is no needto neglect the six displacement modes related to null nodal displacements

Referring to the work of CeAcircsar de SaAcirc et al (2002) and Fontes Valente et al(2002) the framework of the subspace analysis can be used to achieve a shearlocking-free element In fact it is possible to enforce the nullity of thetransverse shear strain energy by including in an additive way an enhancedstrain regeld directly over the transverse shear strain terms of the original

EC207

912

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 18: A new volumetric and shear locking-free 3D enhanced strain element

bilinear degenerated element This inclusion involves the use of six additionalvariables per regnite element and relies on the analysis of the components of thenull transverse shear strain subspace This new strain-displacementdifferential operator (Me

TSL ) involves derivatives of bubble function (19)being deregned as follows

Me

TSL =

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

shy Na

shy xshy Na

shy Zshy 2N a

shy xshy Z0 0 0

0 0 0shy N a

shy xshy Na

shy Zshy 2Na

shy xshy Z

2

6666666666666664

3

7777777777777775

(32)

Thus the crucial idea of the HCiS12 element consists of replacing the enhancedstrain regeld interpolation matrix Me

TSL for the Me

TSL matrix This is performedmaintaining the volumetric enhanced strain interpolation matrix Me

VL in orderto assure the fulreglment of the incompressible subspace requirements Theresult is a new element with 12 additional variables

MeHCiS12 = Me

VL Me

TSL

h i(33)

In summary the proposed HCiS12 element consists of an improvement of theHCiS18 element In fact as will be conregrmed in the next section theirperformances are very close However the HCiS12 has the clear advantage ofusing less six variables in its enhanced strain regeld which means morenumerical stability and less CPU costs

AssessmentsThe goal of the following linear elastic assessments is to test the proposedformulation (HCiS12) performance in situations involving the volumetriclocking andor the transverse shear locking The sensitivity to mesh distortionsis also studied Comparisons with other formulations are carried out including

Solids Q1 plusmn classical FEM eight-node hexahedral element with complete

integration QM1E12 plusmn 3D eight-node hexahedral element with 12 internal variables

proposed by Simo et al (1993)

A new enhancedstrain element

913

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 19: A new volumetric and shear locking-free 3D enhanced strain element

QSE9 plusmn 3D eight-node hexahedral element with nine internal variablesand stabilization proposed by Korelc and Wriggers (1996)

3DEAS-21 plusmn 3D eight-node hexahedral element with 21 internalvariables proposed by Andelregnger and Ramm (1993)

HEXDS plusmn multiple quadrature under-integrated 3D eight-nodehexahedral element proposed by Liu et al (1998)

H1ME9 plusmn mixed-enhanced eight-node hexahedral element with nineenhanced modes and complete quadrature rule proposed by Kasper andTaylor (2000) and

HCiS18 plusmn 3D eight-node hexahedral element with 18 internal variablesproposed by Alves de Sousa et al (2002)

Shells SIMO_89 plusmn bilinear shell mixed element for membrane and bending

stresses proposed by Simo et al (1989) S4E6P7 plusmn shell element using enhanced transverse shear strain terms

proposed by CeAcircsar de SaAcirc et al (2002) MITC4 plusmn fully integrated and mixed interpolated bilinear shell element

derived by Dvorkin and Bathe (1984) EAS7-ANS plusmn in plane bilinear enhanced shell element proposed by

Andelregnger and Ramm (1993) QPH plusmn quadrilateral shell element with physical hourglass control

proposed by Belytschko and Leviathan (1994)

Clamped square plate with concentrated loadFigure 3 shows a clamped square plate loaded by a concentrated load F Onequarter of the geometry needs to be analysed using meshes of 2 pound 2 and 4 pound 4elements The plotted results for vertical displacement in the loaded point arenormalized against the analytical solution (Table II) of Kirchhoff plate theory

Figure 3Clamped square plateunder concentrated load(2 pound 2 mesh)

EC207

914

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 20: A new volumetric and shear locking-free 3D enhanced strain element

Poissonrsquos coefregcient values for compressible and near-incompressiblebehaviours as well as low length to thickness ratios (1100 and 11000) arestudied

The proposed solid elements perform well even in situations of nearincompressibility allied to low length to thickness ratios where volumetric andtransverse shear locking are expected The accuracy is as good as for the shellelement

Block under compressionThis well-known test aims to assess the performance of the elements in nearincompressibility situations allied to mesh distortions (Andelregnger andRamm 1993) Figure 4 shows a block of dimensions 100 pound 100 pound 50 withnear incompressibility properties subjected to a distributed load in itscentral area of 20 pound 20 By symmetry only one quarter of the model isstudied using two different meshes of 5 pound 5 pound 5 elements a regular one

03 04999b

Mesh HCiS12 HCiS18 S4E6P7 MITC4a EAS7-ANS HCiS12 HCiS18 S4E6P7

tL = 1100 2 pound 2 0869 0884 0869 0868 0889 0875 0885 08754 pound 4 0970 0974 0971 0969 0976 0973 0976 0974

tL = 11000 2 pound 2 0866 0882 0866 NA NA 0868 0879 08704 pound 4 0966 0972 0966 NA NA 0968 0972 0968

Notes aFrom Andelregnger and Ramm (1993) bnot available for the MITC4 and EAS7-ANSelements NA plusmn not available

Table IINormalized

displacements for theclamped square plate

Figure 4Block with nearincompressible

properties (a) Regularmesh (b) distorted mesh

A new enhancedstrain element

915

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 21: A new volumetric and shear locking-free 3D enhanced strain element

(Figure 4(a)) and a distorted one (Figure 4(b)) The loaded area is equal forboth meshes The vertical displacement of the central point of the block (A)and the relative error between the solutions for regular and distorted meshesis analysed (Table III)

From the results it can be inferred that the HCiS12 element is volumetriclocking-free allowing good behaviour in near incompressibility situations andalso having low sensitivity to mesh distortions

Thick wall sphere problemThis assessment proposed by Kasper and Taylor (2000) aims to demonstratethe ability of the presented formulations in near incompressible stateOne-eighth of a thick walled sphere (Figure 5) with inner radius Ri= 75 andouter radius Re= 10 is subjected to an internal pressure Q = 1 The elasticmodule is E = 250 and the normalized radial displacements for R = Re isdepicted for several Poisson coefregcient (n) values The HCiS12 and HCiS18elements results (Table IV) are average values for all nodes with R= Re Thevolumetric locking is evident for the Q1 element and the other formulationsbehave well

Figure 5Thick walled sphere

Element Regular mesh Distorted mesh Relative error (per cent)

Q1 1604 pound 10 2 3 1322 pound 102 3 1315HCiS12 1931 pound 10 2 2 1927 pound 102 2 021HCiS18 1913 pound 10 2 2 1903 pound 102 2 052QM1E12a 1892 pound 10 2 2 1840 pound 102 2 275QSE9 1910 pound 10 2 2 1887 pound 102 2 1203DEAS-21 1905 pound 10 2 2 1830 pound 102 2 394

Note aFrom Korelc and Wriggers (1996)

Table IIIVertical displacementresults in the centre ofthe block

EC207

916

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 22: A new volumetric and shear locking-free 3D enhanced strain element

Morleyrsquos 30 8 skew PlateThis example originally proposed by Morley (1963) is analysed in order to testthe sensitivity of the formulations to mesh distortions and the ability to avoidthe shear locking present in problems with low length to thickness ratio in thiscase 1100 The plate is simply supported and subjected to a uniform pressureQ The data of the problem shown in Figure 6 is based on the work ofAndelregnger and Ramm (1993)

The Kirchhoff reference solution of 4455 for the demacrection in the center ofthe plate (C) presented by Morley (1963) is replaced by the value of 4640(Andelregnger and Ramm 1993) since for this thickness to length to ratio of1100 the shear deformation effects cannot be neglected

As can be inferred from Table V and Figure 7 the HCiS12 element showsexcellent results even for coarse meshes being as accurate as the best shellelement performance and showing convergence to the reference solution

Scordelis-Lo roof problemThe proposed original problem from Scordelis and Lo (1969) reports to acylindrical shell with radius R length L and thickness t supported by rigiddiaphragms in the curved edges and subjected to dead load (Figure 8) Due toits symmetry only one quarter of the model is studied and the midpoint freeedgersquos vertical displacement (D) is assessed comparing with the referencesolution of 03024 from Belytschko et al (1985) and Simo et al (1989)

Figure 6Morleyrsquos 30 8 skew plate(four elements per side)

n Q1 HCiS12 HciS18 H1ME9

03 0998 0999 0999 NA049 0971 0997 0998 10190499 0782 0997 0998 100104999 0268 0997 0998 0999049999 0004 0997 0998 09990499999 0006 0997 0998 NA04999999 0001 0997 0997 NA

Note NA plusmn Not available

Table IVNormalized

displacements at R = Re

for the thick walledsphere problem

A new enhancedstrain element

917

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 23: A new volumetric and shear locking-free 3D enhanced strain element

The presented results are normalized against the reference solution andrepresented in Figure 9 and Table VI

From the results it can be inferred that all the elements perform well evenfor coarse meshes Besides that it is important to remark that the solidelements show as good accuracy as the shell elements

Pinched cylinder with end diaphragmsFigure 10 shows one-eighth of a pinched cylinder with radius R length Lthickness t and rigid diaphragms at the end edges The material data arealso provided The structure is subjected to two concentrated loads F = 1This is a classical test for shell elements and also known for causing poorconvergence results in 3D elements The normalized results for the verticaldisplacement in the loaded point against the reference solution of 182488 pound102 5 are presented in Table VII for several distinct meshes based on theones used by the authors here in comparison when testing their formulations

Elements per side 4 5 8 9 16 17 32 33

Q1 00958 NA 03187 NA 08165 NA 15760 NAHCiS12 45060 NA 44210 NA 44750 NA 45740 NAHCiS18 41000 NA 42240 NA 43930 NA 45340 NAS4P6E7 45090 NA 44380 NA 44820 NA 45610 NAMITC4a 39182 NA 38991 NA 41875 NA 44098 NAEAS7-ANS 42122 NA 42239 NA 43738 NA 44827 NASIMO_89 NA 42820 NA 42640 NA 43870 NA 44960Reference 46400 46400 46400 46400 46400 46400 46400 46400

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VMorleyrsquos skew platecentral point demacrection

Figure 7Central point demacrectionfor Morleyrsquos skew plate

EC207

918

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 24: A new volumetric and shear locking-free 3D enhanced strain element

Figure 9Scordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

Figure 8Scordelis-Lo roof

geometry

Elements per side 4 5 8 9 16 17 32

HCiS12 0937 NA 0974 NA 0990 NA 0995HCiS18 1028 NA 1005 NA 0997 NA 0996S4E6P7 1001 NA 1002 NA 0992 NA 0993MITC4a 0937 NA 0973 NA 0993 NA NAEAS7-ANS 1041 NA 1006 NA 1002 NA NAHEXDS NA NA 1157 NA 1137 NA 1132QPH 0940 NA 0980 NA 1010 NA NASIMO_89 NA 1083 NA 1015 NA 1000 NA

Notes aFrom Andelregnger and Ramm (1993) NA plusmn not available

Table VIScordelis-Lo roof

normalized demacrection atfree edgersquos midpoint (D)

A new enhancedstrain element

919

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 25: A new volumetric and shear locking-free 3D enhanced strain element

for this assessment The HCiS12 element shows good results andconvergence properties although showing some sensitivity associated withan increase in thickness of elements

Partly clamped hyperbolic paraboloidThis interesting bending-dominated test was introduced in the work ofChapelle and Bathe (2000) and further developed by Bathe et al (2000) Thistype of bending-dominated tests can be a hard test for standard FEMformulations The problem consists of a self-weighted hyperbolic paraboloidshell structure clamped in one edge and free in the others as shown inFigure 11 By symmetry only one half of the geometry needs to be consideredThe geometry details and the problem data can be found in Chapelle and Bathe(2000)

Figure 10Pinched cylinder withend diaphragms

Mesh HCiS12 HCiS18 QSE9 HEXDS H1ME9 QPH S4E6P7 MITC4a SIMO_89

4 pound 4 pound 1 0104 0106 NA NA 0107 0370 0392 0370 NA5 pound 5 pound 1 0188 0193 0154 NA NA NA 0514 NA 03998 pound 8 pound 1 0494 0498 NA NA 0496 0740 0746 0740 NA9 pound 9 pound 1 0593 0597 0506 NA NA NA 0790 NA 076310 pound 10 pound 2 0673 0677 NA 0801 NA NA 0823 NA NA16 pound 16 pound 1 0912 0912 NA NA 0914 0930 0923 0930 NA16 pound 16 pound 4 0906 0905 NA 0945 NA NA 0923 NA NA17 pound 17 pound 1 0928 0927 0864 NA NA NA 0932 NA 093520 pound 20 pound 4 0950 0947 NA 0978 NA NA 0969 NA NA30 pound 30 pound 1 0993 0987 0971 NA NA NA 0982 NA NA32 pound 32 pound 1 0995 0989 NA NA 0992 NA 0982 NA NA

Notes aFrom Belytschko and Leviathan (1994) NA plusmn not available

Table VIICentral verticaldisplacement for thepinched cylinder

EC207

920

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 26: A new volumetric and shear locking-free 3D enhanced strain element

For the FEM meshes sequences of N pound N2 elements (N = 8 16 32 64) wereconsidered with thicknesslength ratios of 1100 11000 and 110000following the proposal of Bathe et al (2000)

Since there is no analytical solution for this problem for comparisonpurposes the reference values for the vertical displacement of point P and thetotal strain energy are the ones obtained by Bathe et al (2000) using a highorder shell element with a reregned mesh The results present graphically thestrain energy error (Er) of the FEM solutions (Eh) against the reference solution(E) (Figures 12-14) ie

Er = 1 2Eh

E(34)

The reference solution is obtained by the solid elements HCiS12 and HCiS18 inall cases although needing more reregned meshes in the last two situations(11000 and 110000) which denotes sensitivity for diminutions in thethicknesslength ratio

Figure 11Hyperbolic paraboloid

shell structure

Figure 12Relative strain energy

error for the hyperbolicparaboloid problem

A new enhancedstrain element

921

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 27: A new volumetric and shear locking-free 3D enhanced strain element

Concluding remarksBased on the EAS method it was possible to design a new eight-nodedhexahedral element suited for near incompressible situations andor thin shellor plate problems Using a quite simple formulation with no stabilization orunder-integration techniques and a minimal set of additional variablesjustireged by the framework of subspace analysis the HCiS12 element performedsuccessfully in all the examples presented

In the block under compression and thick walled sphere examples novolumetric locking and minimal sensitivity to mesh distortions were seen

In the clamped square plate performed well with low thickness to lengthratios allied to near incompressibility

Figure 13Relative strain energyerror for the hyperbolicparaboloid problem

Figure 14Relative strain energyerror for the hyperbolicparaboloid problem

EC207

922

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 28: A new volumetric and shear locking-free 3D enhanced strain element

In Morleyrsquos skew plate pinched cylinder and Scordelis-Lo roofassessments the results were excellent for a solid element showing noshear locking effects

In the hyperbolic paraboloid shell the convergence to correct solutionswas achieved although showing sensitivity related to decrease of thelength to thickness ratio from 1100 to 110000

Comparing with other well known EAS solid elements a great improvementwas done The extension of the proposed formulation to geometrically andormaterial non-linear problems is currently being carried out

References

Ahmad S Irons BM and Zienkiewicz OC (1970) ordfAnalysis of thick and thin shells structuresby curved regnite elementsordm International Journal for Numerical Methods in EngineeringVol 2 pp 419-51

Alves de Sousa RJ Natal Jorge RM Areias PMA Fontes Valente RA and CeAcircsar de SaAcircJMA (2002) ordfLow order elements for 3D analysisordm in Mang HA Rammerstorfer FGand Eberhardsteiner J (Eds) Proceedings of the Fifth World Congress on ComputationalMechanics (WCCM V) 7-12 July Vienna Austria Vienna University of TechnologyAustria ISBN 3-9501554-0-6 httpwccmtuwienacat

Andelregnger U and Ramm E (1993) ordfEAS-Elements for 2D 3D plate and shell structures andtheir equivalence to HR-elementsordm International Journal for Numerical Methods inEngineering Vol 36 pp 1311-37

Armero F and Dvorkin EN (2000) ordfOn regnite elements for nonlinear solid mechanicsordmComputers and Structures Vol 75 No 3 Special Issue

Bathe KJ (1996) Finite Element Procedures 2nd ed Prentice-Hall New Jersey

Bathe KJ and Dvorkin E (1986) ordfA formulation of general shell elements plusmn the use of mixedinterpolation of tensorial componentsordm International Journal for Numerical Methods inEngineering Vol 22 pp 697-722

Bathe KJ Iosilevich A and Chapelle D (2000) ordfAn evaluation of the MITC shell elementsordmComputers and Structures Vol 75 pp 1-30

Belytschko T and Leviathan I (1994) ordfPhysical stabilization of the 4-node shell element withone point quadratureordm Computer Methods in Applied Mechanics and Engineering Vol 113pp 321-50

Belytschko T Wong BL and Chiang HY (1992) ordfAdvances in one-point quadrature shellelementsordm Computer Methods in Applied Mechanics and Engineering Vol 96 pp 93-108

Belytschko T Stolarski H Liu WK Carpenter N and Ong JSJ (1985) ordfStress projection formembrane and shear locking in shell regnite elementsordm Computer Methods in AppliedMechanics and Engineering Vol 51 pp 221-58

CeAcircsar de SaAcirc JMA and Natal Jorge RM (1999) ordfNew enhanced strain elements forincompressible problemsordm International Journal for Numerical Methods in EngineeringVol 44 pp 229-48

CeAcircsar de SaAcirc JMA and Owen DRJ (1986) ordfThe imposition of the incompressibility constraintin regnite elements plusmn a review of methods with a new insight to the locking phenomenaordm inTaylor C et al (Eds) Proceedings of the III Internacional Conference on NumericalMethods for Non-Linear Problems Dubrovnik Pineridge Press Swansea UK

A new enhancedstrain element

923

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 29: A new volumetric and shear locking-free 3D enhanced strain element

CeAcircsar de SaAcirc JMA Natal Jorge RM Fontes Valente RA and Areias PMA (2002)ordfDevelopment of shear locking-free shell elements using an enhanced assumed strainformulationordm International Journal for Numerical Methods in Engineering Vol 53pp 1721-50

Chapelle D and Bathe KJ (2000) ordfThe mathematical shell model underlying general shellelementsordm International Journal for Numerical Methods in Engineering Vol 48pp 289-313

de Borst R and Groen AE (1999) ordfTowards efregcient and robust elements for 3D-soilplasticityordm Computers and Structures Vol 70 pp 23-34

de Souza Neto EA Peric D Dutko M and Owen DRJ (1996) ordfDesign of simple low orderregnite elements for large strain analysis of nearly incompressible solidsordm InternationalJournal of Solids and Structures Vol 33 pp 3277-96

Doll S Schweizerhof K Hauptmann R and Freischlager C (2000) ordfOn volumetric locking oflow-order solid and solid-shell elements for regnite elastoviscoplastic deformations andselective reduced integrationordm Engineering Computations Vol 17 pp 874-902

Dvorkin EN and Bathe KJ (1984) ordfA continuum mechanics based four-node shell element forgeneral nonlinear analysisordm Engineering Computations Vol 1 pp 77-88

Fontes Valente RA Natal Jorge RM CeAcircsar de SaAcirc JMA and Areias PMA (2002)ordfApplication of the enhanced assumed strain concept towards the development of shearlocking-free shell elementsordm in Mang HA Rammerstorfer FG and Eberhardsteiner J(Eds) Proceedings of the Fifth World Congress on Computational Mechanics (WCCM V)7-12 July 2002 Vienna Austria Vienna University of Technology AustriaISBN 3-9501554-0-6 httpwccmtuwienacat

Glaser S and Armero F (1997) ordfOn the formulation of enhanced strain regnite elements in regnitedeformationsordm Engineering Computations Vol 14 No 7 pp 759-91

Harnau M and Schweizerhof K (2002) ordfAbout linear and quadratic `solid-shellrsquo elements atlarge deformationsordm Computers and Structures Vol 80 pp 805-17

Hauptmann R and Schweizerhof K (1998) ordfA systematic development of `solid-shellrsquo elementformulations for linear and non-linear analyses employing only displacement degrees offreedomordm International Journal for Numerical Methods in Engineering Vol 42 pp 49-69

Hauptmann R Doll S Harnau M and SchweizerhofK (2001) ordfSolid-shell elements with linearand quadratic shape functions at large deformations with nearly incompressiblematerialsordm Computers and Structures Vol 79 pp 1671-85

Hughes TJR (1977) ordfEquivalence of regnite elements for nearly incompressible elasticityordmJournal of Applied Mechanics Vol 44 pp 181-3

Hughes TJR (2000) The Finite Element Method Linear Static and Dynamic Finite ElementAnalysis 2nd ed Dover Editions New Jersey

Hughes TJR Cohen M and Haroun M (1978) ordfReduced and selective integration techniquesin regnite element analysis of platesordm Nuclear Engineering Design Vol 46 pp 203-22

Kasper EP and Taylor RL (2000) ordfA mixed-enhanced strain method Part I plusmn linearproblemsordm Computer and Structures Vol 75 pp 237-50

Korelc J and Wriggers P (1996) ordfAn efregcient 3D enhanced strain element with Taylorexpansion of the shape functionsordm Computational Mechanics Vol 19 pp 30-40

Lee PS and Bathe KJ (2002) ordfOn the asymptotic behaviour of shell structures and theevaluation of regnite element solutionsordm Computers and Structures Vol 80 pp 235-55

EC207

924

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925

Page 30: A new volumetric and shear locking-free 3D enhanced strain element

Liu WK Guo Y Tang S and Belytschko T (1998) ordfA multiple-quadrature eight-nodehexahedral regnite element for large deformation elastoplastic analysisordm Computer Methodsin Applied Mechanics and Engineering Vol 154 pp 69-132

Morley LSD (1963) Skew Plates and Structures Pergamon Press Oxford

Park KC (1986) ordfImproved strain interpolation for curved C0 elementsordm International Journalfor Numerical Methods in Engineering Vol 22 pp 281-8

Park KC and Stanley G (1986) ordfA curved C0 shell element based on assumed naturalcoordinate strainsordm Journal of Applied Mechanics Vol 53 pp 278-90

Piltner R (2000) ordfAn implementation of mixed enhanced regnite elements with strains assumed inCartesian and natural element coordinates using sparse B-matricesordm EngineeringComputations Vol 17 No 8 pp 933-49

Rohel D and Ramm E (1996) ordfLarge elasto-plastic regnite element analysis of solids and shellswith the enhanced assumed strain conceptordm International Journal of Solids and StructuresVol 33 pp 3215-37

Scordelis AC and Lo KS (1969) ordfComputer analysis of cylindrical shellsordm Journal of AmericanConcrete Institute Vol 61 pp 539-61

Simo JC and Armero F (1992) ordfGeometrically non-linear enhanced strain mixed methods andthe method of incompatible modesordm International Journal for Numerical Methods inEngineering Vol 33 pp 1413-49

Simo JC and Rifai MS (1990) ordfA class of mixed assumed strain methods and the method ofincompatible modesordm International Journal for Numerical Methods in Engineering Vol 29pp 1595-638

Simo JC Armero F and Taylor RL (1993) ordfImproved versions of assumed enhanced straintri-linear elements for 3D regnite deformation problemsordm Computer Methods in AppliedMechanics and Engineering Vol 110 pp 359-86

Simo JC Fox DD and Rifai MS (1989) ordfOn a stress resultant geometrically exact shellmodelordm Computer Methods in Applied Mechanics and Engineering Vol 73 pp 53-92

Taylor RL Beresford PJ and Wilson EL (1976) ordfA non-conforming element for stressanalysisordm International Journal for Numerical Methods in Engineering Vol 10 pp 1211-9

Wilson EL Taylor RL Doherty WP and Ghaboussi J (1973) ordfIncompatible displacementmodelsordm in Fenves SJ et al (Eds) Numerical Computational Models in StructuralMechanics Academic Press New York

Zienkiewicz OC and Taylor RL (2000) The Finite Element Method 5th edButterworth-Heinemann London

Zienkiewicz OC Taylor RL and Too JM (1971) ordfReduced integration technique in generalanalysis of plates and shellsordm International Journal for Numerical Methods in EngineeringVol 3 pp 275-90

Further reading

CeAcircsar de SaAcirc JMA Areias PMA and Natal Jorge RM (2001) ordfQuadrilateral elements for thesolution of elasto-plastic regnite strain problemsordm International Journal for NumericalMethods in Engineering Vol 51 pp 883-917

Hughes TJR (1980) ordfGeneralization of selective integration procedures to anisotropic andnonlinear mediaordm International Journal for Numerical Methods in Engineering Vol 15pp 1413-8

A new enhancedstrain element

925