Multiscale approach for the design of composite sandwich structures for train application

Multiscale approach for the design of composite sandwich structuresfor train application

A. Zinno *, E. Fusco, A. Prota, G. ManfrediDepartment of Structural Engineering, University of Napoli Federico II, Via Claudio 21, 80125 Napoli, Italy

a r t i c l e i n f o

Article history:Available online 1 September 2009

Keywords:Sandwich structuresRailway vehicleMechanical characterizationAdhesive joints

a b s t r a c t

In the present work a multiscale approach is considered for the design of composite sandwich structuresfor a roof of railway vehicle. The procedure consists in different steps that start from cost/benefit analysison materials and their manufacturing process and cycle up to analysis of sub-components and entirestructures. Each step is characterized by experimental, theoretical and numerical studies. The designactivities herein presented count experimental campaigns able to characterize both the properties ofnovel sandwich material, manufactured expressly for transportation industry, the sandwich and jointbehaviors. Analytical and numerical approaches have been used to validate and optimize the structurallayout. Finite element analysis has been also performed on a test article to verify the ‘‘new” sandwich roofin regard to structural requirements suggested by European Code.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Sandwich composite structures consist of two thin, stiff andstrong fiber reinforced composite face sheets (skins) separatedby a thick layer of low density material (core) which may bemuch less stiff and strong. The role of the face sheets, due tothe higher elastic properties, is to withstand bending and in-planeactions, while the transverse shear loads are sustained by thecore. The bending stiffness of this type of structural arrangementis much larger than that of a single solid plate of same totalweight made of the same materials as the faces. For this reason,composite sandwich structures are widely used in high-perfor-mance applications where weight must be kept to a minimum,for example aerospace structures, high-speed marine craft andtrains, and racing cars.

Common materials for the sandwich skins are composite orwood laminates and thin aluminum sheets. Polymeric expandedfoams are frequently used for the core which, for more demandingapplications, can alternatively be made of aluminum or aramidcomposite honeycomb. It is quite difficult or impossible to gener-ally define the best combination of sandwich constituents becausethe choice of materials depends not only on strength and stiffnessrequirements but also on process and cost considerations. More-over, other interesting properties of the constituents can influencethe design choices, such as fire and environmental resistance, ther-mal and acoustic insulation, vibration damping and damagetolerance.

The peculiar morphology of a sandwich panel—the layeredand multimaterial structure—requires special attention duringthe design phase. Reliable stiffness and strength predictionscan be made only by using suitable, accurate methodologiesaccounting for the intrinsic structural complexity and the severalfailure modes that a panel can experience. The theoretical anal-ysis of sandwich panels is summarized by Allen [1] and morerecently by Zenkert [2] and Vinson [3], including a systematicdesign strategy for stiffness and strength. It has been recognizedthat sandwich beams could fail by a number of competing mech-anisms. Numerous investigators [4–6] have used the ‘‘failuremode map” concept for sandwich beams in bending to showthe dependence of failure mode upon the geometry and the rel-ative strength of both skins and core. The concept of failuremode map is extended to give a useful design tools for sandwichstructures that can be optimized by minimizing an objectivefunction such as weight or cost against a set of constraints suchas structural stiffness or strength. Frosting and Baruch [7–9]used variational principles to develop a high-order sandwichpanel theory, which includes the transverse flexibility of the corethat is capable to model the local effects at the load points.‘High-order’ refers to the non-linear way in which the in-planeand vertical displacements are allowed to vary through theheight of the core, in contrast to simple beam theory wherethe core in-plane displacements are assumed to vary in a linearway through the depth, and the out-of-plane displacements areassumed to be constant.

Recently sandwich structures are investigated for structural ele-ment of railway vehicle body. Belingardi et al. [10] analyze glassfiber composite–foam sandwich structures for the structural

0263-8223/$ - see front matter � 2009 Elsevier Ltd. All rights reserved.doi:10.1016/j.compstruct.2009.08.044

* Corresponding author. Tel.: +39 0817686336.E-mail address: [email protected] (A. Zinno).

Composite Structures 92 (2010) 2208–2219

Contents lists available at ScienceDirect

Composite Structures

journal homepage: www.elsevier .com/locate /compstruct

design of the front shield of a high speed train, Shin et al. [11] con-sider sandwich panels for low floor of a Korean bus, and analyzethe impact response of the composite solution, while Kim et al.[12,13] underline the optimal fire resistance of composite sand-wich panels on large scale test on a train.

A multiscale approach is presented as optimum tool for thestructural design of composite sandwich structures for a roof ofrailway vehicle (Fig. 1). Fig. 2 shows a developed flow chart thatincludes design and optimization procedure. It starts from thechoice of materials and manufacturing processes as functions ofstructural requirements, cost considerations, and technologiesavailable in the area. In particular the manufacturing process rep-resents a main variable when materials are cured; the curing pro-cess could influence both the mechanical properties of thematerials and the structural behavior of the entire configuration,especially in regard to the failure modes. As result of the previousconsiderations, the derived flow chart underlines the importanceof experimental activities to determinate the mechanical proper-ties of materials, especially for composite skins there is very oftena lack of reliable data, and to analyze the behavior of the sand-wich configurations in order to verify the productive processand to validate analytical and numerical tools. After the validationstep, the better configurations and geometry can be defined usingdifferent strategies available in literature. Grow up on the scale,

structural details, such as joint configurations, should be ana-lyzed. The last step consists to verify the design solutions on asubcomponent and finally on the entire structure. The validationof the engineering product can be developed by full-scale testsand/or by finite element analysis [14].

2. Materials and production

The main structural requirements in the roof design are flexuralstiffness, fire and environmental resistance and fatigue life ofjoints. For this reasons glass and carbon fabric in a phenolic resinsupplied as pre-impregnates sheets have been investigated assandwich skin for their inherently fire-retardant properties thatevolve low levels of smoke and combustion products in a fire.The skins have been combined with either a 130 kg/m3 high-per-formance expanded polymer foam or a 48 kg/m3 aramid fiber rein-forced phenolic (Nomex) honeycomb with a nominal cell size of3.18 mm, to define the better sandwich configuration. All the de-scribed materials are certificated according to AFNOR NF F 16-101 standard [15] in regards to their fire, smoke and toxicityproperties.

The composite solutions have been manufactured throughvacuum bagging process cured in autoclave. The selected basiccure cycle for the composite laminates is shown in Fig. 3. It

Fig. 1. Type of train investigated: (a) vehicle body and (b) roof made of composite sandwich panels.

Fig. 2. Multiscale procedure for the design and optimization of sandwich structures.

A. Zinno et al. / Composite Structures 92 (2010) 2208–2219 2209

counts a vacuum pressure of 2.5 bar, a ramping temperature upto 135 �C at 2 �C/min and maintaining this temperature for90 min.

In the present step an experimental activities was carried out onboth the skin and core materials. In particular mechanical charac-terizations of the skins were developed to define the in-plane prop-

erties. The specimen geometries (Fig. 4) and staking sequences arereported in Table 1. Tensile tests were run in one series with thewarp fibers parallel to the load (Fig. 5a and b) and in a second ser-ies with warp fibers perpendicular to the load (fill direction)(Fig. 5c and d). These tests were performed in accordance withthe ASTM D3039M standard [16]. Ultimate tensile stress andstrain, elastic modulus, and Poisson ratio have been derived forboth wrap and fill directions.

Shear tests (Fig. 5e and f) were run on five glass and fivecarbon coupons. These tests were performed in accordance withthe ASTM D3518 M standard [17]. Shear modules, ultimate shearstress and strain have been derived by these tests. For both in-plane tensile and shear tests, three strain gauges were appliedto each coupon, in order to monitor the longitudinal and trans-verse strain and the possible bending due to misalignment of thespecimens.

Interlaminar shear stress have been derived by ‘‘short-beam”tests (Fig. 6a and b) using a three-point bending set-up in accor-dance with the ASTM D2344 M standard [18].

All laminate tests were run on a 10 kN universal test frame con-trolled by an electronic control unit which allows monitoring theapplied load and the stroke of the top cross head. Strain signalswere acquired by a digital data acquisition system. Tests were con-ducted at a constant cross head velocity of 2 for tensile and sheartests and 1 mm/min for the short-beam tests. Table 2 shows all themechanical properties derived by the described tests.

Experimental activity on core materials consists in out-planetests, while their in-plane properties are deduced by the manufac-

0

20

40

60

80

100

120

140

160

0 50 100 150 200 250

Time [min]

Tem

pera

ture

[°C

]

0

0.5

1

1.5

2

2.5

3

3.5

Pres

sure

[bar

]

TemperaturePressure

Fig. 3. Manufacturing process parameters of composite sandwich element.

x

y12

(Loading direction)

Fiber orientations θ

w

LtLgLt

t(a)

(b)

Fig. 4. Shape geometry of material coupons: (a) laminates and (b) core materials.

Table 1Coupon geometries for laminate tests.

No. coupon Lt (mm) Lg (mm) L (mm) w (mm) t (mm) h (�)

Tensile || warp direction Glass fabric 6 60 190 250 15 1 All 0�Carbon fabric 5 60 190 250 25 0.8 All 0�

Tensile \ warp direction Glass fabric 6 60 190 250 15 1 All 90�Carbon fabric 5 60 190 250 25 0.8 All 90�

Shear Glass fabric 5 60 190 250 25 2 [+45�/�45�]2sCarbon fabric 5 60 190 250 25 1.6 [+45�/�45�]2s

Short-beam Glass fabric 6 – – 36 12 6 All 0�Carbon fabric 6 – – 36 13 5 All 0�

2210 A. Zinno et al. / Composite Structures 92 (2010) 2208–2219

turer’s data [19,20] and are showed in Table 3. The out-of-planecompressive properties, especially the non-linear behavior, havegreat influence on the response of a sandwich structures underconcentrated load, as indentation or impact. Uniaxial compressivetests (Fig. 7) were run on four 100 � 60 � 10.5 mm and100 � 60 � 11 mm coupons for Nomex and foam material respec-tively according to ASTM C365-05 [21] standard. The tests wereconducted at a constant cross head velocity of 0.5 mm/min. Ulti-mate compressive strength and compressive modulus have beenderived, they result equal to the manufacturer’s value.

3. Sandwich analysis

The selected sandwich beams are obtained combining the samematerials described in the previous section. Table 4 reports all thespecimen material combinations and geometries. Three-pointbending tests were run in accordance the ASTM C393 M standard[22]. Three electrical resistance strain gauges were applied onspecimens, one on the midspan of the bottom skin and one at25 mm distance of the midspan both on the top and bottom skin,while the displacement of the midspan has monitored using a

0

100

200

300

400

500

600

700

800

0 0.005 0.01 0.015 0.02

Strain [-]

Stre

ss [M

Pa]

0

100

200

300

400

500

600

700

800

0 0.005 0.01 0.015 0.02

Strain [-]

Stre

ss [M

Pa]

)b()a(

0

100

200

300

400

500

600

700

800

0 0.005 0.01 0.015 0.02

Strain [-]

Stre

ss [M

Pa]

0

100

200

300

400

500

600

700

800

0 0.005 0.01 0.015 0.02

Strain [-]

Stre

ss [M

Pa]

)d()c(

0

10

20

30

40

50

60

70

80

90

100

0 0.01 0.02 0.03 0.04 0.05

Strain [-]

Stre

ss [M

Pa]

0

10

20

30

40

50

60

70

80

90

100

0 0.01 0.02 0.03 0.04 0.05

Strain [-]

Stre

ss [M

Pa]

)f()e(

Fig. 5. Tests on laminates made of phenolic matrix: (a) tensile tests on glass sheets in wrap direction; (b) tensile tests on carbon sheets in wrap direction; (c) tensile tests onglass sheets in fill direction; (d) tensile tests on carbon sheets in fill direction; (e) shear tests on glass sheets and (f) shear tests on carbon sheets.

A. Zinno et al. / Composite Structures 92 (2010) 2208–2219 2211

LVDT. Each type of sandwich structure was tested with two differ-ent support spans: L1 = 325 mm and L2 = 265 mm. All tests wererun on a 10 kN universal test frame as described in the previoussection. Beam tests were conducted in stroke control with a crosshead speed of 6 mm/min. The load was applied by a 25 mm wideflat steel block. Flexural D and shear U stiffness of each sandwichconfiguration have been derived by comparing the results of thetwo tests performed using different support span according toASTM D7250M standard [23] (Table 5).

The derived results have showed that the manufacturing seemsnot to influence the mechanical behavior of the sandwich elementsand no premature failures (such as debonding or delamination)have occurred. Moreover closed-form analytical and numericaltools have been analyzed in order to capture the experimentallybehavior both in term of stiffness and in regard to failure modes.Theoretical analysis have been based on both ordinary bendingtheory and higher-order sandwich beam theory (HOSBT). For bothapproaches, the analysis is elastic, which is appropriate to describe

0

5

10

15

20

25

30

35

40

45

50

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Crosshead displacement [mm]

Stre

ss [M

Pa]

0

5

10

15

20

25

30

35

40

45

50

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Crosshead displacement [mm]

Stre

ss [M

Pa]

)b()a(

Fig. 6. Flexural tests on laminates made of phenolic matrix: (a) short-beam tests on glass sheets and (b) short-beam tests on carbon sheets.

Table 2Mechanical properties of composite skins derived by experimental tests.

Ex (GPa) Ey (Gpa) Gxy (GPa) rx (MPa) ry (MPa) sxy (MPa) rsb (MPa) ex (�) ey (�) cxy (�) mxy (�)

Glass fabric 25.54 22.97 3.41 325.77 288.21 43.30 21.34 1.53 1.56 2.47 0.15Carbon fabric 46.65 44.32 4.44 579.59 546.90 91.19 45.09 1.22 1.23 4.12 0.10

Table 3Mechanical properties of Nomex and foam materials according to manufacturer data.

Density Compressive Shear

q r E s (MPa) G (MPa)

(kg/m3) (MPa) (MPa) L direction W direction L direction W direction

Nomex HRC-10 1/8 48 2.24 137.95 1.21 0.69 44.83 24.14Klegecell TR 130 130 2.5 86 1.95 50

0

0.5

1

1.5

2

2.5

3

3.5

4

0 0.1 0.2 0.3 0.4 0.5 0.6

Strain [-]

Stre

ss [M

Pa]

0

0.5

1

1.5

2

2.5

3

3.5

4

0 0.1 0.2 0.3 0.4 0.5 0.6

Strain [-]

Stre

ss [M

Pa]

)b()a(

Fig. 7. Out-of-plane compressive tests on the core: (a) Nomex coupons and (b) foam coupons.

2212 A. Zinno et al. / Composite Structures 92 (2010) 2208–2219

the beam response up to peak load for the possible materialcombinations.

The ordinary theory have been adopted in order to evaluate thestresses in the core or skins and hence the applied loads corre-sponding to various failure mechanisms. Based on the summary gi-ven in. [24], the failure loads depend on properties of the skin andsolid core material, relative density of the core, thickness of bothskins and core and beam span and loading details. The maximumstress in the face sheets can be used to predict the beam failure

due to the skin failure modes—i.e., face ultimate strength, facewrinkling and intra-cellular buckling. In a symmetric beam, thestress is the same in the compression and tension faces. For com-posite faces, the critical face is generally the compressive. The fail-ure occurs when the axial stress in either of the skins reaches thein-plane ultimate strength. In the case of wrinkling of the compres-sion face, the wavelength of the buckled face is of the same orderas the thickness of the core. This problem may be analyzed asthe buckling of a beam (the face sheet) supported transversely byan elastic foundation (the core). With three-point bending, wrin-kling of the top skin occurs in the vicinity of the central load. Allen[1] gives the critical compressive stress that results in wrinkling.

A sandwich with a honeycomb core may fail by buckling of theface in a small region where it is unsupported by the walls of thehoneycomb. The following expression has been proposed [25] forthe in-plane stress rfi in the skin at which intra-cellular bucklingoccurs:

rfi ¼ 2Ef

ð1� m2f ÞtR

� �2

; ð1Þ

Table 4Specimen geometries of composite sandwich beams.

No. coupon Length L (mm) Width w (mm) Thickness

Skin Core

t (mm) h c (mm)

Glass/Nomex 4 470 100 1.00 [0/90]s 11.00Glass/foam 2 470 100 1.00 [0/90]s 10.50Carbon/Nomex 3 470 100 0.20 [0] 11.00Carbon/foam 4 470 100 0.20 [0] 10.50

Table 5Specimen geometries of composite sandwich beams.

Experimental Theoretical

D (kN mm2) U (kN mm) D (kN mm2) U (kN mm)

Glass/Nomex 189905.37 59.05 183864.96 58.69Glass/foam 168578.79 62.55 168862.09 62.98Carbon/Nomex 57082.05 51.26 58522.43 51.12Carbon/foam 53648.66 54.35 53413.84 54.52

1.E-02

1.E-01

1.E+00

1.E-04 1.E-03 1.E-02 1.E-0Skin thickness/support span

Rel

ativ

e co

re d

ensi

ty

Exp. data

Core Shear

Wrinkling

IndentationSkin failure

(a)

1.E-02

1.E-01

1.E+00

1.E-04 1.E-03 1.E-02 1.E-0Skin thickness/support span

Rel

ativ

e co

re d

ensi

ty

Exp. data

Core ShearWrinkling

Indentation

Skin failure

(b)

Fig. 8. Failure mode map for the composite sandwich beam under three-point bending configuration: (a) glass–Nomex beam and (b) glass–foam beam.

A. Zinno et al. / Composite Structures 92 (2010) 2208–2219 2213

where R is the cell size of the honeycomb core, while Ef, t and mf arethe elastic modulus, thickness and Poisson ratio of the skinrespectively.

Moreover sandwich structures loaded in bending can fail due tocore failure. Pertinent modes are shear failure or indentation bylocal crushing in the vicinity of the loads application. Shear failureoccurs when the applied shear stress equals the shear strength ofthe core. Indentation failure is predicted when the out-of-planecompressive stress equals the out-of-plane compressive strengthof the core. To evaluate the core failure mechanism, stiffness andstrength properties for the Nomex honeycomb and foam core arerequired. The out-of-plane Poisson’s ratio, required for the failureanalysis, can be taken, to a first approximation, as that of the solidmaterials, while the out-of-plane Young’s modulus of the Nomexhoneycomb can be reached by the rule of mixture expression.

For a honeycomb with regular hexagonal cells, Wierzbicki [26]gives the expression for the ultimate out-of-plane compressivestrength, while Petras and Sutcliffe [27] derive the expressionsfor the out-of-plane shear strengths of regular hexagonal honey-comb core. Based on an analysis of the manufacturer’s modulusand strength data, the following expressions are derived for theanalyzed foam core:

Ec

Es¼ 1:16

qc

qs

� �1:28

; ð2Þ

rcc

rsc¼ 1:10

qc

qs

� �1:63

; ð3Þ

scrsc

¼ 0:78qc

qs

� �1:28

: ð4Þ

where qc and qs, Es and Es and rcc and rsc are the density, Young’smodulus and compressive strength of the foam core and of the solidfrom which the core is made respectively.

The actual behavior is then governed by the mode with the min-imum failure load that can be reached combining all the describedmechanisms. Failure load surfaces and the relative failure modemaps have been drawn, using the Matlab programming language,for the tested sandwich structures (Figs. 8 and 9). The failuremodes and loads are plotted as a function of core relative densityand skin thickness to span ratio, at fixed core thickness to span ra-tio. When plotting the experimental measurements for each sand-wich configuration, it is observed that the experimental failuremodes are consistent with the analytical predictions in each case.It is interesting to observe that, for a honeycomb core, the intra-cellular buckling mode represents less severe condition than thewrinkling failure due to the small hexagonal cell size. However,the wrinkling surface represents a small area of the failure modemap since the honeycomb sandwich has a high modulus in theout-of-plane direction, thus limiting the wrinkling phenomenon.

The basic assumptions of the HOSBT approach are:

� the shear stresses in the core are uniform through the thicknessof the core;

� the core vertical displacement variation is a quadratic polyno-mial in z, allowing the core to distort and its height to change;

� the core is considered as a 3D elastic isotropic medium, whichhas significant out-of-plane compressive and shear rigidity,but negligible in-plane normal and shear rigidity;

� the skins are thin, elastic and, isotropic plates.

1.E-02

1.E-01

1.E+00

1.E-04 1.E-03 1.E-02 1.E-0Skin thickness/support span

Rel

ativ

e co

re d

ensi

ty

Exp. data

Core Shear

Wrinkling

IndentationSkin failure

(a)

1.E-02

1.E-01

1.E+00

1.E-04 1.E-03 1.E-02 1.E-0Skin thickness/support span

Rel

ativ

e co

re d

ensi

ty

Exp. data

Core ShearWrinkling

Indentation

Skin failure

(b)

Fig. 9. Failure mode map for the composite sandwich beam under three-point bending configuration: (a) carbon–Nomex beam and (b) carbon–foam beam.

2214 A. Zinno et al. / Composite Structures 92 (2010) 2208–2219

The 3D problem of [9] have been reduced to a 2D problem [28],which described equations have been implemented using the Mat-lab programming language to calculate the beam response in termof load–displacement curves.

Moreover finite element modeling has been developed to per-form reliable simulations of structural behavior of sandwich beamsto be compared with the experimental data. The analysis was per-formed on 3D-models using Nastran� finite element codes. Theskins were meshed using 4-node shell elements, while the corewas meshed using 8-node chexa solid elements. A 2D-orthotropicmaterial was used to define the composite fabric prepreg, andthe composite function was used to create the stacking sequenceof the face sheets. Isotropic and 3D-orthotropic materials wereused for foam and honeycomb core respectively. Skins and corematerial properties were defined only in the linear elastic range.Fig. 10 shows that the numerical and analytical behaviors are per-fectly matched with the experiments. After the validations step,the configuration of the sandwich roof can be define. The sandwichlayout has been chosen considering stiffness and geometrical con-strains. In particular the same flexural rigidity of the previous roofconfiguration have been guaranteed considering a sandwich struc-ture which skins, made by both carbon and glass prepreg, havebeen bonded to Nomex core of 12.5 thickness. In this way the totalthickness of the composite sandwich panel is in the same tolerancerange of the previous configuration, so that all non-structural com-ponents can be placed at the same way. The Nomex solution hasbeen preferred to the foam one in order to obtain a grater reduc-tion of structural weight. In fact the composite solution allows to

reduce by approximately 50% the total weight of the roof (512 kgvs. 1080 kg).

4. Analysis of structural details

In the present section adhesive joint configurations have beeninvestigated to connect the new composite roof with the aluminum

0

1

2

3

4

5

0 5 10 15 20 25Displacement [mm]

Forc

e [K

N]

Exp. S=325Analy. S=325FEM S=325Exp. S=265Analy. S=265FEM S=265

0

1

2

3

4

5

0 5 10 15 20 25Displacement [mm]

Forc

e [K

N]

Exp. S=325Analy. S=325FEM S=325Exp. S=265Analy. S=265FEM S=265

)b()a(

0

0.5

1

1.5

2

0 1 2 3 4 5 6 7 8Displacement [mm]

Forc

e [K

N]

Exp. S=325 Analy. S=325FEM S=325Exp. S=265Analy. S=265FEM S=265

0

0.5

1

1.5

2

0 1 2 3 4 5 6 7 8Displacement [mm]

Forc

e [K

N]

Exp. S=325Analy. S=325FEM S=325Exp. S=265Analy. S=265FEM S=265

)d()c(

Fig. 10. Force–displacement curves of sandwich beams: (a) glass–Nomex sandwich beams; (b) glass–foam sandwich beams; (c) carbon–Nomex sandwich beams and (d)carbon–foam sandwich beams.

Table 6Specimen geometries for adhesive tests.

Adherents No.coupon

Overlap surface(mm2)

Static shear Aluminum–Aluminum

8 500

Aluminum–Composite

13

Composite–Composite

14

Staticcleavage

Aluminum–Aluminum

2 2000

Aluminum–Composite

8

Composite–Composite

10

Fatigue shear Aluminum–Aluminum

8 375

Aluminum–Composite

13

Composite–Composite

14

A. Zinno et al. / Composite Structures 92 (2010) 2208–2219 2215

frame. Static and fatigue characterizations were run in order to ana-lyze the adherent–adhesive behavior and their total life.

The experimental activities were carried out on three combina-tions of adherents (aluminum–aluminum, composite–composite,and aluminum–composite) and on one bi-component adhesivechosen for its capacity to resist to extreme dynamic stresses, highstrength and outdoor environmental, such as UV rays, temperatureand moisture. The considered alloy aluminum is classified as theEN-AW6065A, while the composite laminates are made by pheno-lic/glass prepreg described in the previous sections.

Static shear (ASTM D1002 [29]) and peel (ASTM D3807 [30])properties of each adherent–adhesive–adherent combination havebeen derived on single lap joint and cleavage configurations bytension loading respectively. Table 6 reports sampling and testspecimen for all the possible combinations. Shear and cleavagetests were run on a 5 kN and 10 kN universal test frame, con-trolled by an electronic control unit which allows monitoringthe applied load and the stroke of the top cross head, respec-tively. In particular shear and cleavage tests were conducted instroke control with a cross head speed of 1.27 and 12.7 mm/min respectively.

Moreover fatigue properties of the joint configurations, accord-ing to ASTM D3166 [31], have been derived by cyclic tests usingthe S–N approach. All the specimens were loaded between theminimum and maximum axial force at 5 Hz frequency. In orderto obtain statistical analysis of fatigue data, replicate tests wereperformed with different load amplitude (Table 7). For each loadlevel, the number of cycle at specimen failure was derived. In thisway, the S–N relationship have been derived using the Basquinequation (Fig. 11). The design values for each material combinationhave been obtained as the strength at 107 cycles, as suggested byEN 12663 code [32]. Table 8 reports static and fatigue propertiesobtained by the experimental activities.

5. Subcomponent validation

A numerical simulation was carried out on the new designsolution. Fig. 12a shows a test article considered into the finite

Table 7Load history of fatigue tests.

n. coupon Load Amplitude (kN)

Aluminum–Aluminum 2 0.602 0.801 0.952 1.00

Aluminum–Composite 4 0.452 0.552 0.653 0.702 1.00

Aluminum–Composite 1 0.302 0.401 0.456 0.504 0.55

0.1

1

10

1.0E+00 1.0E+02 1.0E+04 1.0E+06Number of Cycles

Stre

ss A

mpl

itude

[MPa

]

Exp. dataiS-N curve

0.1

1

10

1.0E+00 1.0E+02 1.0E+04 1.0E+06Number of Cycles

Stre

ss A

mpl

itude

[MPa

]

Exp. dataS-N curve

)b()a(

0.1

1

10

1.0E+00 1.0E+02 1.0E+04 1.0E+06Number of Cycles

Stre

ss A

mpl

itude

[MPa

]

Exp. dataS-N curve

)c(

Fig. 11. Fatigue tests on single lap configuration: (a) aluminium–aluminium; (b) aluminium–composite and (c) composite–composite.

2216 A. Zinno et al. / Composite Structures 92 (2010) 2208–2219

element analysis. The model has been performed using MSC Nas-tran code�. The sandwich roof has been modeled using the sameelements and materials described in Section 3, the aluminumframe using shell elements and the adhesive joint using themethod proposed by NASA counting three different springs (onefor the axial and two for the shear stiffness of the glue) andtwo rigid elements to connect the spring to the nodes(Fig. 12b). A concentrated load was applied to simulate the max-imum load due to the heavier gear can be placed on the roof. The

stress field in the sandwich, derived by the static simulation, hasbeen verified using Hill and maximum stress criteria for skins andcore respectively (Fig. 13). Static and fatigue simulations havebeen also performed to validate the adhesive joint comparingthe data derived by the output of the spring with the experimen-tal values.

Moreover both free and constrained modal analysis on the testarticle have been performed (Fig. 14). In particular the free analysisverify that the first six modes are rigid, while the seven is different

Fig. 12. FE test article model of the roof of the vehicle: (a) test article description and (b) joint detail.

Table 8Tests results on adhesive joint configurations.

Adherents Static Fatigue Shear

Shear Cleavage 107 cycle(MPa)

Av. (MPa) Cv. (%) Av. (kN/mm) Cv. (%)

Aluminum–Aluminum 5.95 8.37 0.30 0.95 1.26Aluminum–Composite 4.48 10.37 0.21 6.35 0.96Composite–Composite 3.30 19.51 0.15 5.53 0.90

Fig. 13. FE quasi static analysis of test article: (a) ply failure index and (b) max shear distribution in the core.

A. Zinno et al. / Composite Structures 92 (2010) 2208–2219 2217

to zero as suggested by the EN 12663 code, which define also thefrequency range of the first ten vibration modes that have beensatisfied by the constrained analysis.

6. Conclusion

The objective of the work presented in this paper was to de-velop a multiscale procedure for the design problems involvingsandwich composites for train applications. The extensivemechanical characterization activity was carried out to assesmechanical data of novel phenolic sandwich materials manufac-tured expressly for the transportation accepted validate manufac-turing process and calibrate analytical and numerical models. Thestudy underlines the capacity of failure mode map to be a goodtool to predict the failure loads and modes as function of geome-try and materials and the capacity of analytical and numericalmodel to capture the elastic flexural stiffness of sandwichstructures.

Experimental activity was also carried out to assess mechanicalproperties of adhesive joint that can be used to connect structuralcomponents of a railway vehicle. This technology is increasinglybeing used due to their improved mechanical performance, a bet-ter understanding of the mechanics of failure and benefits in termof costs and time compared to traditional techniques. Finallynumerical simulation have been developed on a test article to val-idate the design procedure in term of structural elements anddetails.

The proposed multiscale approach allows to optimize the de-sign phase using and combining experimental, numerical and ana-lytical tools that integrate in each described step generate a virtualtesting approach. In this way, if material and manufacturing char-acteristics are known, composite sandwich structure will be designto substitute other structural elements, such as flanks or frontshield, of railway vehicle and then validate trough (few) bench-mark tests.

Acknowledgements

The activities presented in the paper have been conductedwithin a project in collaboration between the Department of Struc-tural Engineering at University of Naples Federico II and FiremaTrasporti whose financial support is kindly acknowledged.

References

[1] Allen HG. Analysis and design of structural sandwich panels. Pergamon Press;1969.

[2] Zenkert D. An introduction to sandwich construction. Engineering MaterialsAdvis Serv; 1997.

[3] Vinson JR. The behavior of sandwich structures of isotropic and compositematerials. Technomic Publishing; 1999.

[4] Kuenzi EW. Sandwich design criteria. Forest Products Laboratory; 1959.[5] Triantafillou TC, Gibson LJ. Failure mode maps for foam core sandwich beams.

Mater Sci and Eng 1987;95:37–53.[6] Shenhar Y, Frosting Y, Altus E. Stresses and failure patterns in the bending of

sandwich beams with transversely flexible cores and laminates compositeskins. Compos Struct 1996;35:143–52.

[7] Frosting Y, Baruch M. Bending of sandwich panels with transversely flexiblecore. AIAA Journal 1990;28:523–31.

[8] Frosting Y, Baruch M. Localized load effects in high-order bending of sandwichpanels with flexible core. J Eng Mech 1996;122:1069–76.

[9] Frosting Y, Baruch M. High-order theory for sandwich-beam behavior withtransversely flexible core. J Eng Mech 1992;118:1026–43. ASCE.

[10] Belingardi G, Cavarota MP, Duella R. Material characterization of a composite-foam sandwich for the front structure of high speed train. Compos Struct2003;61:13–25.

[11] Shin KB, Lee JY, Cho SH. An experimental study of low-velocity impactresponses of sandwich panels for Korean low floor bus. Compos Struct2008;84:228–40.

[12] Kim JS, Lee SJ, Shin KB. Manufacturing and structural safety evaluation of acomposite train carbody. Compos Struct 2008;78:468–76.

[13] Kim JS, Chung SK. A study on the low-velocity impact response of laminates forcomposite railway bodyshells. Compos Struct 2007;77:484–92.

[14] NAFEMS. Quality system supplement to ISO 9001 relating to finite elementanalysis in the design and validation of engineering products. National Agencyfor Finite Element Methods and Standards; 1993 [Issue 1.3].

[15] AFNOR NF F16-101. Matériel roulant ferroviaire – Comportement au feu –Choix des matériaux, October 1988.

[16] ASTM D3039M. Standard test method for tensile properties of polymer matrixcomposite materials; 2000.

[17] ASTM D3518M. In-plane shear response of polymer matrix compositematerials by tensile test of a ±45� laminate; 1994.

[18] ASTM D2344M. Standard test method for short-beam strength of polymermatrix composite materials and their laminates; 2000.

[19] Hexcel. Mechanical properties of Hexcel honeycomb materials.[20] DIAB. Klegecell TR grade technical data.[21] ASTM C365M. Flatwise compressive properties of sandwich cores; 2005.[22] ASTM C393M. Core shear properties of sandwich constructions by beam

flexure; 2006.[23] ASTM D7250M. Determining sandwich beam flexural and shear stiffness;

2006.[24] Zinno A, Bakis CE, Prota A. Mechanical characterization and structural behavior

of composite sandwich structures for train applications. In: Proceedings of the8th international conference on sandwich structures (ICSS8), Berlin: Springe;vol. 1, May 6–8; 2008. p. 570–581.

[25] Anonymous. Composite construction for flight vehicles. US Military handbook;1959.

[26] Wierzbicki T. Crushing analysis of metal honeycombs. Int J Imp Eng1983;12:157.

Fig. 14. FE modal analysis of test article: (a) free analysis and (b) constrained analysis.

2218 A. Zinno et al. / Composite Structures 92 (2010) 2208–2219

[27] Petras A, Sutcliffe MPF. Failure mode maps for honeycomb sandwich panels.Compos Struct 1999;44:237–52.

[28] Zinno A, Bakis CE, Prota A. Numerical estimates of composite sandwichresponse under quasi static indentation. In: Proceedings of the 13th Europeanconference on composite materials (ECCM 13), June 2–5, Stockholm, Sweden;2008 [Paper Id. 2624].

[29] ASTM D1002. Apparent shear strength of single-lap-joint adhesively bondedmetal specimens by tension loading; 2005.

[30] ASTM D3807. Strength properties of adhesives in cleavage peel by tensionloading; 2004.

[31] ASTM D3166. Fatigue properties of adhesives in shear by tension loading;2004.

[32] EN 12663. Railway applications – structural requirements of railway vehiclebodies; November 2002.

A. Zinno et al. / Composite Structures 92 (2010) 2208–2219 2219