Top Banner
u u y y x u v u v y y u v fi‘ u w v fi‘ u u v y y u v u u é x u v u u u y v y u u x x x u u v y u y u u u y u y u x u u x u wy w y x v v v u u v yu v w w y y u y w ~ y u v w x x x x u u w x y u v u y w u y w w z y u w u w x v uu y u —v v x u y
6

ExternalUniformPressure - ovgu.de · At the inner boundaries of the shell (x=x1, x=x2)conditions of joining must be posed. It means, that dis— placements, axial and shear stresses

Feb 09, 2019

Download

Documents

doantram
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: ExternalUniformPressure - ovgu.de · At the inner boundaries of the shell (x=x1, x=x2)conditions of joining must be posed. It means, that dis— placements, axial and shear stresses

TECHNISCHE MECHANIK, Band 20, Heft 4, (2000), 349-354

Manuskripteingang: 06. Mai 1999

Buckling of Cylindrical Shells of Variable Thickness, Loaded by

External Uniform Pressure

I.V. Andrianov, B.G. Ismagulov, M.V. Matyash

This paper is dedicated to the memory of Professor N.A. Alfutov.

From the mathematical standpoint one has a partial difi‘erential equation with variable coefi‘icients. Perturba-

tion procedure gives the possibilityfor an analytical solution of this eigenvalue problem. Self-adjoint equations

and Padé approximants are usedfor improving the obtained results.

l Introduction

The problem under consideration is important from an engineering standpoint. It may be solved by numerical

procedures, for example finite element or finite differences method, matrix method, etc. On the other hand,

changing of thickness in practice does not exceed 20-40 %, so, the perturbation procedure is very natural in this

case for an analytical solution.

Let us consider a circular cylindrical shell (Figure 1), loaded by uniform external pressure. It is assumed that the

external pressure keeps always the direction towards the cylinder principal axis (dead loading). In the linear case

one obtains a conservative eigenvalue problem (Alfutov, 1978; Grigolyuk and Kabanov, 1978). It is worth not-

ing two essentially different limiting cases. For Ll << L one may suppose an isotropic ring-stiffened shell, for

(L— L!) << L one has practically a shell of constant thickness. Here we are interested in cases L] ~ L , then the

shell may be subdivided into three parts (or two parts, if xl = i0, x2 :t L or x2 = L, x] at 0 ).

2 Perturbation Procedure

In the framework of the semi-inextensional theory (Alfutov, 1978; Reissner, 1964) (or the quasimembrane the-

ory, if we use the terminology of (Awrejcewicz et a1., 1998)) for each part of the shell one may use the following

equations:

34w 1) a4 32 2 q a“ a2BI- 4 —2+l —2+1 W=0 i=1,2

ax R atp acp R" atp Btp

Ehi Eh? „ ' . . . . . . .Here B,- = 2 ; D,- = 2 ‚ l = 1, 2; E, v - modulus of elastrcrty and Pmsson ratio; (p — Circumferential

l—v 1211-v

coordinate.

x

A

‘_ 11

"'— LI

<— L

co— h]

0 l

>I

2R

Figure 1. Cylindrical Shell of Variable Thickness

349

Page 2: ExternalUniformPressure - ovgu.de · At the inner boundaries of the shell (x=x1, x=x2)conditions of joining must be posed. It means, that dis— placements, axial and shear stresses

At the inner boundaries of the shell (x = x1, x= x2)conditions of joining must be posed. It means, that dis—

placements, axial and shear stresses must be equal. Using distributions, one may write the stability equation for

the whole shell in the following form

a4w D a4 a2 2 a4 a2B +———+1W+iz——74——2+1W=0 (2)

8x4 R6 8(p4 E)th R" acp 8(p

Here B=E—h’{1+£[H(x—xl)—H(x—x2)]} D = Eh? h+ee[H(x_x,)_H(x_x2)]}i—v2 12l1—v2l

3

s=—h2_h‘ soc: £2— —1

h]

H(x) is the Heaviside function.

We use the linear theory of stability and don’t take into account initial imperfections and moments of the critical

state. These factors are not important in the case under consideration. We suppose that the shell is simply sup-

ported, then

2

W=av¥=0 for x=0,L (3)

3x

Satisfying conditions of periodicity in circumferential direction one may write the unknown eigenvalue function

W(x‚ (p) in the following form

W = f(x)sin mp

Then the function f (x) may be obtained from the ordinary differential equation with variable coefficients:

d4 4 2_1 D

:3 [er-MMNow we have the following possibilities for a first approximation of the eigenvalue problem. If Ll << 0.5L, in

equation (4) one can suppose B = BZ, D = D2. For L1 ~ 0.5Lit will be more suitable to suppose

B = 0.5(31 +132), D = 0.5(D1 +D2). Further we shall deal with the case 0 << Ll << 0.5L‚ then as first ap-

proximation one may use equation (4) with 8 = O.

A solution of the eigenvalue problem equations (4) and (3) may be easily obtained:

. TEX

= s1n~——f0 L

fl 4 R3 + Eh? ( 2 l)

(L) n4ln2—1l 12R3ll—v2ln _

Minimization of expression (5) with respect to n gives us a known formula (Alfutov, 1978; Grigolyuk and Ka-

banov, 1978), and for v = 0.3 and L > 2R one may write

5/2

R h= 0.92E— —1

c10 L f R j

(5)

(10

It is worth noting that further approximations may change the value of n, so, one must use expression (5) and the

minimization procedure with respect to n in each approximation.

The solution of the governing eigenvalue problem we represent in the form of formal expansions:

350

Page 3: ExternalUniformPressure - ovgu.de · At the inner boundaries of the shell (x=x1, x=x2)conditions of joining must be posed. It means, that dis— placements, axial and shear stresses

q = qo + 891+ 8242 + (6.1)

f = f0 + sf1+ 52102 +... . (6.2)

Substituting series (6) in equation (4) and boundary conditions (3) and splitting it with respect to powers of 8,

one obtains the recurrent sequence of eigenvalue problems:

M (f0) = quU'o) <7)

M(f1)+ M1(f0)= qmm) + slow) <8)

M(f2)+M1(f1)=q2N(f0)+01N(.f1)+QON(f2) (9)

M03) + M102) = 6/3N(fo) + qum) + q1N(f2)+ qu(f3) (10>

f- = dzfi = 0 i=0, 1,2, (11)

l dx2

Here

d4f _ _ 4f

M(f)=B1d 4 + A'f N(f>—C<f M1(f)— 816M +A~oc-f [H(x~x1>—H(x—x2)]X

A : n4(n2;l)D1 C I n (mg—1)

R R

Since the nonperturbative eigenvalue problem is self—adjoint, one can easily obtain ql from equation (8), using

scalar product by f0 (Nayfen, 1973)

2E4 3 E 2—1 L 2% . nL41: 2 7:5 2 (hZ—h1)+ n (113—1113) ~ ~Ll———cos srn—1

l—v L‘h h —1 2 27': L L

where x* = (xI +x2)/2‚

Minimization with respect to n gives us in the first approximation

. 8 6R“2 — . . 2 1 . L_ 439 7(h2 hl)+ 1 1458_02 h;_hP)Äh_iCOS nxsmh (12)

Lh](7.29R3/-—Lh}/3) 1115/2 L111”2 R“2 2 21c L L

Neglecting small terms in formula (12) one obtains:

2 3

1. Lw: 4.976&+4.374 93 +1.458 A—h i—icoszm smh

h1 h1 h1 L

Here Ah = hz — hl . The coefficient (0 takes into account the influence of the parameters L1, Ah and xi; on

the critical pressure (Figure 1). The expression in square brackets gives us the dependence of c0 on Ah, the

expression in parentheses contains parameters L1 and x* respectively. One may write

0) = V’(Ah/hl) V”(x*,A)

351

Page 4: ExternalUniformPressure - ovgu.de · At the inner boundaries of the shell (x=x1, x=x2)conditions of joining must be posed. It means, that dis— placements, axial and shear stresses

/

/

o.1_ /

0 1.0 2.0 V'

Figure 2. Dependence of Nondimensional Critical Pressure from Variation of Thickness

As one can see from Figure 2, the dependence of w from Ah/hl is almost linear for small values of AM h1 .

Some numerical results are shown in Figure 3, where we use the following geometrical parameters:

L1 = 1.5 m,L=15mand R = 10.5m . Maximal influence of the inclusion takes place for x. = L/2.

‚AV

n

/ \\

:2: / \0.04 \\

0.02 '/ X“

„x? \\0 k

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 x./L

Figure 3. Dependence of Nondimensional Critical Pressure from Coordinate x*

In the first approximation one obtains the eigenform as follows

fl(x)= iAI-sinnax (l3)

j=2

Here a : 1t/ L.

Coefficients of expansion (13) may be written in the following form:

-= a4+0L __—Sin(l_j)ax _ 5m(1+j)ax X2

A} ( (l—j)2a (l+j)2a J I

A. = _ _ (Bla4+D,a) {sinU—jhx _ Sln(l+j)a,x]

' 7Ill'ilb'a)“+A—quJ (l—j) (1+ j)

Using solution (13), one obtains term q2 in the expansion (6.1)

352

Page 5: ExternalUniformPressure - ovgu.de · At the inner boundaries of the shell (x=x1, x=x2)conditions of joining must be posed. It means, that dis— placements, axial and shear stresses

L (q0C+A(a—l))j2:2Aj

A—qu °" A' _ _ A— C °° 14‘ _ . . .Z—{cos Jaxs1nax+i Z—ékos Jaxsmax+cosaxsm jax)+

a j=2 J ‘1 i=2]

4 .

+[2131a3 +fl—q—1CJ-(cosax-sm axi

a a

q =_2_[ w (sin(l—j)ax sin(l+j)axJ+

2 C

+2

x2

XI

3 Using Adjoint Equations

For the coefficient q} one may write

_ ((Ml(f2)—41N(f2)):fo)‘42(N(f|)‚f1)

q} ‘ (mm “4’

Since the function f2 is unknown as well, we transform the first term in the numerator of the ratio (Marchuk et

al., 1996) as follows

((Ml(f2)"qlN(f2))rf0) = ((Ml(fO)—qlN(f0))’f2) (15)

From equation (8) we have:

M|(fo)—q1N(fo)=“M(f1)+40N(f1) (16)

and

((Ml(f0)_qlN(f0))er) *((M(f1)—QON(f1))sf2) =

"((M(f2)—CI0N(f2))af|) = ‘(l‘Mllfi)‘l'LIIler)+q2N(fo)+Q2N(fo))’fI)=

((M1(f1)—41N(fr))vfr) = ((qu(fo))‚fl)

Finally we obtain

((M,(f1)—q1N(f.)),fi)—((2q2N(fo)),fl) (17)

(N(fo)‚fo)

4 Error Estimation and Padé Approximants

(13:

To estimate the accuracy of the obtained solution we also solved the governing eigenvalue problem by the finite

difference method. Some numerical results are shown in Figure 4 for UR = 4; R/h. = 500, x* : 0.5L; L1 = 0.3L.

le

3.0

4

1(4”’—--

/,’ / 4

Ö

/’ /

l0 \

\

\2

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Zh/h,

Figure 4. Comparison of Analytical and Numerical Results

353

Page 6: ExternalUniformPressure - ovgu.de · At the inner boundaries of the shell (x=x1, x=x2)conditions of joining must be posed. It means, that dis— placements, axial and shear stresses

Results obtained by the perturbation method are presented by curves 1 (Q, : qo + Sq] ) , 2 (Q2 = Q1 +82q2 )and

3 (Q3 2 Q2 +83q3), the finite difference solution (EDS) is presented by the dotted line.

The perturbation series is divergent. Really, for Ah/hl = 0.3 the discrepancy between FDS and Ql is 8.5 %, but

between FDS and Q3 the discrepancy is 38 %.

For overcoming this drawback we use the analytical continuation by Fade approximants (Awrejcewicz, et a1.

1998; Baker and Graves—Morris, 1996). The Padé approximant for Q2 has the following form

qu; _ l+a€

_ (18)

go l+b€QPA E

b=—q2/q1; a=q1/q0+b.

Numerical values of function QPA are depicted in Figure 4 by curve 4. Evidently, formula (18) gives good re—

sults even for a large change of the thickness.

5 Concluding Remarks

The presented approach allows to obtain the closed analytical formula for the critical pressure of the cylindrical

shell of variable thickness.

Acknowledgment

The authors thank the anonymous reviewer for helpful comments.

Literature

1. Alfutov, N.A.: Foundation of Stability Calculations of Elastic Systems. Moscow, Mashinostroyenie, in Rus—

sian, (1978).

2. Awrejcewicz, 1.; Andrianov, I.V.; Manevitch, L.I.: Asymptotic Approaches in Nonlinear Dynamics: New

Trends and Applications. Heidelberg, Springer Verlag, (1998).

3. Baker, G. A.; Graves-Morris, P.: Pade Approximants. Cambridge, Cambridge VP, (1996).

4. Grigolyuk, E.I.; Kabanov, V.V.: Stability of Shells. Moscow, Nauka, in Russian, (1978).

5. Marchuk, G.I.; Agoshkov, V.1.; Shutyaev, V.P.: Adjoint Equations and Perturbation Algorithms in Nonlin—

ear Problems. Boca Raton, CRC, (1996).

6. Nayfeh, A.H.: Perturbation Methods. New York, Wiley Interscience, (1973).

7. Reissner, E.: An asymptotic expansions for circular cylindrical shells. Trans. ASME, J. Appl. Mechs, 31,

(1964), 245-252.

Addresses: Professor Dr. Igor V. Andrianov, Department of Mathematics, Pridneprovskaya State Academy of

Civil Engineering and Architecture (PGASA), 24 a Cheryshevskogo, UA—49005 Dnepropetrovsk; Senior Re-

search Scientist Dr. Bulat G. Ismagulov, Department of Metallic Constructions, PGSA; Associate Professor Dr.

Marina V. Matyash, Department of Finantial-Economical Information Computer Processing, Dnepropetrovsk

State University, 13 Nauchny line, UA—49625 Dnepropetrovsk.

354