You are on page 1of 11

Composite Structures 76 (2006) 303–313

www.elsevier.com/locate/compstruct

Modal coupled instabilities of thin-walled


composite plate and shell structures
Z. Kolakowski *, M. Krolak
Department of Strength of Materials and Structures (K12), Technical University of Lodz, PL-90-924 Lodz, ul. Stefanowskiego 1/15, Poland

Available online 8 May 2006

Abstract

The problem of buckling and initial post-buckling equilibrium paths of thin-walled structures built of plate and/or shell elements
subjected to compression and bending has been solved. Plate and shell elements can be made of multi-layer orthotropic materials. A
method of the modal solution to the coupled buckling problem within the first-order approximation of KoiterÕs asymptotic theory,
using the transition matrix method, has been presented. In the solution obtained, the effect of cross-sectional distortions and a shear
lag phenomenon is included. The calculations are carried out for a few thin-walled structures.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Composite structures; Membrane–flexural coupling; Buckling; Coupled instabilities

1. Introduction the same in all directions. Composites are strong and


more fail-safe in their nature than other materials. For
Since their appearance in the 1960s, modern compos- instance, homogeneous materials such as metals crack
ite materials have augured the new, bright and, at the easily, whereas in non-homogeneous composites such
same time, light and reliable future. Scientists have cracks have simply no chances to enlarge. Composites
started to design composite materials whose mechanical can alter their properties considerably, but despite this
properties can be determined in advance. Composites fact they can guarantee safety. Still, an unpredicted load
have been lately commonly used in aircraft industry, variation may be sometimes followed by disaster.
army, automotive industry and marine structures, as Nevertheless, during the last fifteen years composites
well as in medicine and sport. have been widely used as structural elements of aero-
A limited application of these materials in our every- planes, helicopters and space shuttles. Thorough re-
day life results mainly from their complexity. They are search aimed at obtaining the best materials at the
made of carbon or glass fibres, embedded in a polymer, lowest production costs has been conducted all that
ceramic or metallic warp. Their complexity brings into time. As a result of these investigations, numerous new
question their design, mass production and application composite materials, characterised by properties that of-
by traditionally educated engineers. Composite materials fer new possibilities to designers, have appeared.
carry mainly stresses along the direction of the fibre align- The increasing popularity of fibrous composites as
ment. Consequently, they require more attention while structural materials can be attributed to the following
designing than metals whose strength is most frequently reasons:

• high strength-to-weight and stiffness-to-weight ratios;


*
Corresponding author. Tel.: +48 42 631 2223; fax: +48 42 631 2217. • high resistance of some fibrous composites to aggres-
E-mail address: kola@orion.p.lodz.pl (Z. Kolakowski). sive chemical compounds and to corrosion;

0263-8223/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2005.02.022
304 Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313

• broad spectrum of thermal properties, varying The most important advantage of this method is that
according to the sort of a matrix and fibres used it enables us to describe a complete range of behaviour
(thermal insulating power, thermal conductivity, high of thin-walled structures from all global (flexural, flex-
mechanical strength at elevated, high or very low ural–torsional, lateral, distortional buckling and their
temperatures). combinations) to the local stability [9].
The consideration of displacements and load compo-
Other positive factors promoting the use of fibrous nents in the middle surface of walls within the first-order
composites in the construction of load carrying elements approximation, as well as precise geometrical relation-
include good wear (e.g. friction) resistance, fatigue life ships have enabled the analysis of all possible buckling
(under time-variable loads), sound insulation properties modes, including a mixed buckling mode [8–11,13]. See
or electromagnetic transparency. also [12,21].
Composite materials have a great advantage, which
consists in the fact that their material properties can
be easily modelled in selected directions and regions. It 2. Formulation of the problem
concerns especially orthotropic composite beams, plates
and shells. Let us consider thin-walled girders built of plate and
In order to apply composites, however, a comprehen- shell elements (in the form of cylindrical shell segments)
sive and thorough analysis of the structure behaviour is re- with closed or open cross-sections. Both the plate and
quired. In engineering practice a development of new shell elements under consideration can be multi-layer
methods of the analysis that account for special properties walls made of orthotropic materials. The classical lami-
of composite materials is needed. To fulfil these require- nated panel theory [6] is used in the theoretical analysis,
ments, models based on analytical and numerical methods the effects of shear deformation through the thickness of
that describe a complex structure of composites are built. the laminate are neglected and the results given are those
The problem is more complicated as composite elements for thin laminated panels. The materials they are made
are subjected to complex loading in modern structures. of are subjected to HookeÕs law.
With progress in the technology of composite materi- Long prismatic thin-walled structures built of panels
als, composite plates made of solid laminates, sandwich connected on longitudinal edges have been considered.
laminates, and laminates reinforced with stiffeners are In order to account for all modes of global, local and
widely used in cars, wagons, ships, aeroplanes, aero- coupled buckling, a plate model of thin-walled struc-
space and marine structures. The analysis of buckling tures has been assumed.
of composite plates is crucial to the safe design of these For each panel component, precise geometrical rela-
structures. Composite plates are inherently anisotropic tionships are assumed in order to enable the consider-
and non-homogeneous, and their buckling deformations ation of both out-of-plane and in-plane bending of the
are more complicated. plate:
A more comprehensive review of the literature can be
e1 ¼ u1;1 þ 0:5um;1 um;1
found in [2–4,6,8,13–16,18,22–25].
Mathematical models tend to higher precision and e2 ¼ u2;2 þ 0:5um;2 um;2  ku3
closer approximation of real structures, which enables e3 ¼ u1;2 þ u2;1 þ um;1 um;2
one to analyse more and more exactly the phenomena ð1Þ
occurring during and after the loss of stability. e4 ¼ hu3;11
As a result of the present investigations, an analytical- e5 ¼ hu3;22
numerical model that makes it possible to account for
e6 ¼ 2hu3;12
many properties of real structures, such as different char-
acteristics of orthotropic materials, beam-columns with where h is the thickness of the plate, k = 1/R is the cur-
open and closed sections, longitudinal stiffeners, a band vature of the cylindrical shell segment; u1  u, u2  v,
thickness variation, edge reinforcement, various boundary u3  w—the components of the displacement vector in
conditions, a wall interaction, a shear lag phenomenon, an the x1  x, x2  y, x3  z axis direction, respectively,
effect of the interaction of buckling modes and also an ef- and 1 = x, 2 = y, 3 = 2xy = cxy, 4 = hjx, 5 = hjy,
fect of cross-sectional distortions, has been developed. 6 = hjxy. The summation with respect to the factor m
Thin-walled composite structures in the elastic range, is from 1 to 3 (m = 1, 2, 3).
being under axial compression or/and a bending mo- In the majority of publications devoted to stability of
ment, are examined on the basis of Byskov and Hutch- structures, the terms ðu21;1 þ u22;1 Þ; ðu21;2 þ u22;2 Þ, (u1,1u1,2 +
insonÕs method. The study is based on the numerical u2,1u2,2) are neglected for e1, e2, e3, respectively, in the strain
method of the transition matrix using GodunovÕs tensor components (1).
orthogonalization. Instead of the finite strip method, Other shells have more complicated strain-displace-
the exact transition matrix method is used in this case. ment relations [5] and they have been disregarded here.
Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313 305

Using the classical plate theory [6], the constitutive impossible to pull on a laminate that has Bij terms with-
equation for the laminate is taken as follows: out bending and/or twisting the laminate at the same
  time. That is, an extensional force results in not only
½A ½B
fN g ¼ feg ¼ ½Kfeg ð2Þ extensional deformations, but also twisting and/or bend-
½B ½D
ing of the laminate [4,6]. Moreover, such a laminate
where cannot be subjected to moment without suffering simul-
1X N
 Þ ðzk  zk1 Þ taneously from extension of the middle surface.
Aij ¼ ðQ ð3Þ
h k¼1 ij k The form of constitutive Eq. (2) assumed here is very
general, allowing for anisotropic laminate properties
1 X N with regard to both in-plane and out-of-plane behav-
Bij ¼  ij Þ ðz2  z2 Þ
ðQ ð4Þ iours, and for full coupling between in-plane and out-
k k k1
2h2 k¼1
of-plane behaviours, as well as for through-thickness
shearing actions.
1 XN
 ij Þ ðz3  z3 Þ The complexity of Eq. (2) is reduced somewhat for
Dij ¼ 3
ðQ k k k1 ð5Þ
3h k¼1 laminated plates of particular specialised configurations.
  For example, suppose that the reference surface is se-
½A ½B lected at the middle surface of the plate and the laminate
½K ¼
½B ½D is symmetrical with respect to this surface.
2 3 For laminates that are symmetric in both geometry
A11 A12 A16 B11 B12 B16
6A and material properties about the middle surface, the
6 21 A22 A26 B21 B22 B26 7
7
6 7 general stiffness equations, Eq. (6), simplify consider-
6 A61 A62 A66 B61 B62 B66 7 ably. In particular, the coupling stiffness, that is Bij,
¼6
6B
7
6 11 B12 B16 D11 D12 D16 7
7 can be shown to be zero. The elimination of coupling be-
6 7 tween bending and tension has important practical ram-
4 B21 B22 B26 D21 D22 D26 5
ð6Þ ifications. Such laminates are usually much easier to
B61 B62 B66 D61 D62 D66
2 3 analyse than laminates with coupling and they do not
K 11 K 12 K 13 K 14 K 15 K 16 have a tendency to twist from the inevitable thermally
6K K 23 K 24 K 25 K 26 7
6 21 K 22 7 induced contractions that occur during cooling that fol-
6 7 lows the curing process [6].
6 K 31 K 32 K 33 K 34 K 35 K 36 7
¼6
6K
7 As another example of a specialised configuration, let
6 41 K 42 K 43 K 44 K 45 K 46 7
7
6 7 us consider a regular symmetric cross-ply laminate that is
4 K 51 K 52 K 53 K 54 K 55 K 56 5 symmetrical with respect to the middle surface and the
K 61 K 62 K 63 K 64 K 65 K 66 laminae are all of the same thickness and material prop-
 ij is
in which Aij = Aji, Bij = Bji, Dij = Dji, Kij = Kji and Q erties, for example [0/90/90/0]T. For this laminate, the
the transformed reduced stiffness matrix [4,6]. stiffness Bij = A16 = A26 = D16 = D26 = 0. This laminate
In the above equations, N1, N2, N3 are the dimension- is called a quasi-orthotropic material.
less sectional forces; N4, N5, N6—the dimensionless sec- Regular symmetric angle-ply laminates have orthotro-
tional moments: pic laminae of equal thickness. The stiffness Bij is zero
Nx Ny N xy because of symmetry. The adjacent layers have opposite
N1 ¼ ; N2 ¼ N3 ¼ ; signs of the angle of orientation of the principal material
E0 h E0 h E0 h
ð7Þ properties with respect to the laminate axes, for exam-
Mx My M xy
N4 ¼ ; N 5 ¼ N 6 ¼ ple, [+h/h/h/+h]T. Thus, for symmetry, there must
E 0 h2 E 0 h2 E 0 h2 be an odd number of layers. For many layered symmet-
where E0 is the elastic modulus of reference. ric angle-ply laminates, the values of A16, A26, D16 and
The reverse relation with respect to (2) can be written D26 can be quite small when compared to other Aij
as and Dij, respectively. Only in the situation where the
1 
feg ¼ ½K fN g ¼ ½KfN g ð8Þ stiffness A16, A26, D16 and D26 is exactly zero, it can be
ignored without analysis [6].
In the constitutive matrix of Eq. (2), the submatrix [A], In many physical applications of laminated compos-
detailed in Eq. (3) and related to the in-plane response of ites, antisymmetric laminates are needed to fulfil design
the laminate, is called extensional stiffness. The subma- requirements. If the shear stiffness of a laminate made of
trix [D], described by Eq. (5), is associated with the out- laminae with unidirectional fibres must be increased,
of-plane bending response of the laminate and is called one way to satisfy this requirement is to position layers
bending stiffness, whereas the submatrix [B], illustrated at some angle with respect to the laminate axes. An even
by Eq. (4), is a measure of an interaction (coupling) be- number of such layers may be necessary at orientations
tween the membrane and the bending action. Thus, it is that alternate from layer to layer, e.g. [+h/h/+h/h]T.
306 Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313
Z
A regular antisymmetric cross-ply laminate consists of  ij N j ÞdN i dS ¼ 0
½ei  K
an even number of orthotropic laminae laid on one an- S
other with principal material directions alternating at 0 where : i; j ¼ 1; 2; . . . ; 6 and no summation to
and 90 to the laminate axes and has all laminae of the
‘i’ occurs ð12Þ
same thickness. These laminates do not have A16, A26,
D16, D26, B12, B16, B26 and B66, but have coupling be- for x1 = const
tween bending and extension. The coupling is such that Z b
B11 = B22 5 0. ½N 1 þ N 1 u1;1 þ N 3 u1;2  hp0 ðx2 Þdu1 dx2 jx1 ¼ 0
A regular antisymmetric angle-ply laminate has all 0
Z b
laminae of the same thickness oriented at +h degrees
to the laminate coordinate axes on one side of the mid- ½N 3 þ N 1 u2;1 þ N 3 u2;2 du2 dx2 jx1 ¼ 0
0
dle surface and corresponding, equally thick laminae Z b
oriented at h degrees on the other side. In this laminate N 4 du3;1 dx2 jx1 ¼ 0
A16 = A26 = B11 = B12 = B22 = B66 = D16 = D26 = 0 but 0
B16 5 0 and B26 5 0. Z b
The equations of stability of thin-walled structures ðhN 4;1 þ 2hN 6;2 þ N 1 u3;1 þ N 3 u3;2 Þdu3 dx2 jx1 ¼ 0
0
have been derived using a variational method [4,8,9].
The total potential energy variation of a multi-layer ð13Þ
sector of the panel P can be written as for x2 = const
Z
Z ‘
dP ¼ dU  dW ¼ ðrx dex þ ry dey þ sxy dcxy Þ dX
X ½N 2 þ N 2 u2;2 þ N 3 u2;1 du2 dx1 jx2 ¼ 0
Z b  0
0 x¼‘ Z
 hp ðyÞdu dyjx¼0 ð9Þ ‘
0 ½N 3 þ N 2 u1;2 þ N 3 u1;1 du1 dx1 jx2 ¼ 0
0
where U—internal elastic strain energy, W—work of Z ‘
external forces, p0(y)  p0(x2)—pre-buckling external
N 5 du3;2 dx1 jx2 ¼ 0
load in the panel middle surface. The expression 0
X = ‘ · b · h = S · h has been employed in the above Z ‘
relation. ðhN 5;2 þ 2hN 6;1 þ N 2 u3;2 þ N 3 u3;1 Þdu3 dx1 jx2 ¼ 0
0
The above equation says that in the equilibrium state
the potential energy P of the panel has a non-varying ð14Þ
value in the class of admissible displacement variations. for the plate corner, that is to say, for x1 = const and
Hence, the equation has to be satisfied for all admissible x2 = const
virtual displacements that comply with the constraints
imposed. N 6 jx1 jx2 du3 ¼ 0 ð15Þ
In order to determine the variation of the potential Relation (11) is a system of equilibrium equations.
energy of a single multi-layer panel, the following Systems (12) are the already employed relations between
identity strains and external forces, whereas relations (13)–(15)
X dY ¼ dðXY Þ  Y dX ð10Þ correspond to the boundary conditions at the edge
x1 = const, x2 = const, respectively, and in the plate
has been employed. corner.
After grouping the components at respective varia- In the case of a plate, the last term kN2 should be ne-
tions, the following system of equations has been glected in Eq. (11) because k = 1/R = 0.
obtained: An introduction of the full functional into the solu-
Z tion allows for satisfying the subsequent and consistent
f½N 1 ð1 þ u1;1 Þ þ N 3 u1;2 ;1 systems of simplified mathematical models correspond-
S
ing to the characteristic types of composite panels.
þ ½N 2 u1;2 þ N 3 ð1 þ u1;1 Þ;2 gdu1 dS ¼ 0
Z
f½N 1 u2;1 þ N 3 ð1 þ u2;2 Þ;1
S 3. Solution to the problem
þ ½N 2 ð1 þ u2;2 Þ þ N 3 u2;1 ;2 gdu2 dS ¼ 0
Z
The problem has been solved by the variational meth-
½ðhN 4;1 þ N 1 u3;1 þ N 3 u3;2 Þ;1 od using KoiterÕs asymptotic theory of conservative sys-
S
þ ðhN 5;2 þ 2hN 6;1 þ N 2 u3;2 þ N 3 u3;1 Þ;2 þ kN 2 du3 dS ¼ 0 tems [7]. In the solution of the problem and in the
computer program developed, the following have been
ð11Þ employed: Byskov–HutchinsonÕs asymptotic expansion
Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313 307

[1], the numerical transition matrix method using Godu- The corresponding displacements are treated as infin-
ð0Þ ð0Þ
novÕs orthogonalization method [8–11]. itesimally small, so that for any uj ; L11n ðuj ; vj Þ ¼ 0 for
The girders under consideration can be loaded with all i and j, any vj, and L11 is a bilinear operator, follow-
axial compression and bending in the cross-section sym- ing the notation of [1].
metry plane, that is to say, with the so-called flat bending. Numerical aspects of the problem being solved for
The aim of this study is to generate the stability equations the first-order fields (like in papers by Kolakowski and
and to solve the stability problem (to find the values of coworkers [8,10,11]) have resulted in an introduction
the critical load of local and global buckling), as well as of the following new orthogonal functions for kth com-
to determine the initial equilibrium paths in the elastic posite panel in the sense of the boundary conditions for
post-buckling for the structures described above. two longitudinal edges (14):
As has been mentioned above, after expanding the ðnÞ ðnÞ ðoÞ ðnÞ
fields of displacements U  and the fields of sectional ak ¼ N 2k ð1 þ kDk K 21k =K 22k Þ þ kN 3k dk;n =bk

forces N into power series with respect to the buckling bðnÞ ¼ N ðnÞ ð1  kDk Þ þ kN ð0ÞlðnÞ =bk
k 3k 3k k;n
mode amplitudes fn (the amplitude of the nth buckling ðnÞ ðnÞ
mode divided by the thickness h1 of the wall assumed ck ¼ u1k
to be the first one), KoiterÕs asymptotic theory has been ðnÞ ðnÞ
dk ¼ u21k
employed: ðnÞ ðnÞ
ð18Þ
ek ¼ u3k
U  ð0Þ þ fn U
 ¼ kU  ðnÞ þ    ðnÞ ðnÞ ðnÞ
k k
ð16Þ f k ¼ u3k;g =bk ¼ ek;g =bk
N  ð0Þ þ fn N
 ¼ kN  ðnÞ þ    ðnÞ ðnÞ
k k
gk ¼ N 5k
where U  ð0Þ ; N
 ð0Þ are the pre-buckling state fields, and
k k hðnÞ ¼ hk gðnÞ =bk þ 2hk N ðnÞ =bk þ kN ðo ÞeðnÞ =bk
 ðnÞ ; N
U  ðnÞ —the nth buckling mode fields for the compos- k k;g 6k;n 3k k;n
k k
ite kth panel. The range of indices is [1, N], where N is where nk = x1k/bk and gk = x2k/bk.
the number of interacting modes. In the case of arbitrary composite structures, the
After substitution of expansions (16) into equilibrium solution to the problem within the first-order approxi-
Eq. (11), continuity conditions (14), boundary condi- mation can be sought in the form of trigonometric series
tions (13) (corresponding to the free support at the seg- in the light of inconsistency of these functions in the
ment ends) and relations (12), the boundary problem of equilibrium equations along the longitudinal direction
the zero and first-order for the case of uniform compres- or in an approximate way.
sion along the generating lines of the panel has been In the majority of studies, only conditions of orthog-
obtained. onality (11) of the assumed class of functions of the
The panels with the linearly varying pre-buckling problem solution with respect to displacements u = u1,
state along their widths are divided into several strips v = u2, w = u3 are fulfilled. In the solution method pre-
under uniformly distributed compressive (tensile) stres- sented here, additional orthogonality conditions (i.e.
ses. Instead of the finite strip method, the exact transi- (12) and (14)) are satisfied.
tion matrix method is used in this case [10,11]. A separation of variables in the linear boundary
The pre-buckling solution of the kth composite panel problem has been solved by means of KantorowichÕs
consisting of homogeneous fields is assumed as method. The first-order solutions may be formulated
ð0Þ as follows:
u1k ¼ ð‘=2  x1k ÞDk
ðnÞ ðnÞ mpbk n ðnÞ
 ðnÞ ðgk Þ cos mpbk n
ð0Þ
u2k ¼ x2k Dk K 12k =K 22k ð17Þ  ðgk Þ sin
ak ¼ A k ; bk ¼ B k
‘ ‘
ð0Þ
u3k ¼ 0 ðnÞ ðnÞ mpb k n ðnÞ ðnÞ mpb kn
ck ¼ C ðgk Þ cos
k ; dk ¼ D  ðgk Þ sin
k
‘ ‘
where Dk is the actual loading. This loading is specified
as the product of a unit loading system and a scalar load
ðnÞ
ek ¼ E ðnÞ ðgk Þ sin mpbk n ; fkðnÞ ¼ F ðnÞ ðgk Þ sin mpbk n
k k
‘ ‘
factor Dk.
ðnÞ  ðnÞ ðgk Þ sin mpb k n ðnÞ
 ðnÞ ðgk Þ sin mpb kn
As can be easily seen, thus assumed field of displace- gk ¼ G k ; hk ¼ H k
ments and the field of inner forces, corresponding to it ‘ ‘
for the pre-buckling state (see Appendix B), fulfils equi- ð19Þ
librium Eq. (11) for the zero state as an identity.  ðnÞ ;
A  ðnÞ ;
B  ðnÞ ;
C  ðnÞ ;
D  ðnÞ ;
E
ðnÞ
F k ;
(with the mth  ðnÞ ;
G  ðnÞ
H
k k k k k k k
The zero approximation describes the pre-buckling harmonic) are initially unknown functions that will be
state, whereas the first-order approximation, being the determined by the numerical method of transition matri-
linear problem of stability, allows for determination of ces. The solution assumed in this way (19) allows one to
values of critical loads, buckling modes, and initial determine dimensionless sectional forces for the first-
post-buckling equilibrium paths. order approximation in the form:
308 Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313

ðnÞ ðnÞ
 ðgk Þ sin mpbk n ðnÞ  ðnÞ ðgk Þ sin mpbk n Expressions for a0, ar, ajkr are calculated by known
N 1k ¼ N 1k ; N 2k ¼ N 21k
‘ ‘ formulae [1]. The formulae for the post-buckling coeffi-
cients ajkr depend only on the buckling modes.
ðnÞ
N 3k  ðnÞ ðgk Þ cos mpbk n ; N 4k
¼N
ðnÞ  ðnÞ ðgk Þ sin mpbk n
¼N The consideration of only linear initial imperfections
3k 41k
‘ ‘
(determined by the shape of N coupled buckling modes)
ðnÞ
N 5k  ðnÞ ðgk Þ sin mpbk n ; N 6k
¼N
ðnÞ  ðnÞ ðgk Þ cos mpbk n
¼N and components of displacement and force fields for the
5k 6k
‘ ‘ first-order allows us, to some extent, to account for
ð20Þ residual stresses. However, the residual stresses are not
The presented way of solution allows for carrying out assumed in advance and this approach can only be trea-
a modal analysis of buckling of complex composite thin- ted as an attempt to consider their most unfavourable
walled structures. distribution. For a rather extensive discussion, see pa-
The obtained system of homogeneous ordinary differ- pers by Pignataro and Luongo [19,20].
ential equations, with the corresponding conditions of
the co-operation of walls, has been solved by the transi-
tion matrix method, having integrated numerically the 4. Analysis of the calculation results
equilibrium equations along the circumferential direc-
tion in order to obtain the relationships between the The developed calculation program has been tested in
state vectors on two longitudinal edges. During the inte- many aspects, among others in order to determine the
gration of the equations, GodunovÕs orthogonalization range of applicability of the cylindrical shell theory as-
method is employed. The column global buckling occurs sumed in the considerations.
at one sinusoid half-wave on the column length, whereas A comparison of the results obtained by means of the
the local buckling takes place at the number of half- method discussed above and by the finite element meth-
waves m > 1 with bk  ‘ [8–11]. od—ANSYS 5.4 package—is presented below. A multi-
In systems of Eqs. (11), (12) and (14) for the first- layer channel with the following geometrical dimensions
order approximation, there are two components of the (Fig. 1):
pre-critical loading N  ð0Þ and N  ð0Þ (as can be easily seen b1 ¼ 25 ½mm; b2 ¼ 50 ½mm; ‘ ¼ 300 ½mm;
1k 3k
ð0Þ
in (18)). The component N  has an insignificant effect
3k
 ð0Þ . h ¼ 1 ½mm
on the value of critical load in comparison with N 1k
The developed computer program allows for a divi- has been considered.
sion of each plate or/and shell element into several or Each wall is made of four layers of two hypothetical
even more than 50 strips made of different materials materials characterised by the following parameters:
and with various wall thickness. The presented solution
method enables a multi-modal analysis of buckling. • h1 = 0.2 [mm], h = 0 [mm], E1 = 20 [GPa], E2 = 10
Taking into account the components of membrane [GPa], G = 5 [GPa], m12 = 0.3;
forces and displacements within the first-order approxi- • h1 = 0.3 [mm], h = 35 [mm], E1 = 10 [GPa], E2 = 30
mation allows one to consider a shear–lag phenomenon [GPa], G = 10 [GPa], m12 = 0.3;
and distortions of cross-sections. • h1 = 0.3 [mm], h = 52 [mm], E1 = 10 [GPa], E2 = 30
At the point where the load parameter k reaches its [GPa], G = 10 [GPa], m12 = 0.3;
maximum value ks (secondary bifurcation or limit • h1 = 0.2 [mm], h = 12 [mm], E1 = 20 [GPa], E2 = 10
points) for the imperfect structure with regard to the [GPa], G = 5 [GPa], m12 = 0.3
imperfection of the buckling mode with the amplitude
fr , the Jacobian of the non-linear system of equations where E1m21 = E2m12 and h = h1 + h2 + h3 + h4.
[1]:
 
k k
ar 1  fr þ ajkr fj fk þ    ¼ ar fr at r ¼ 1; 2; . . . ; N
kr kr
b2
ð21Þ
is equal to zero.
h
The corresponding expression for the total elastic po-
tential energy of the structures has the following form: b1
P ¼ ao k2 =2 þ ar ð1  k=kr Þf2r =2 þ ajkr fj fk fr =3
 ar fr fr k=kr þ    ð22Þ
h
where k—load parameter, kr—critical value of k,
Po = aok2/2—energy of the pre–buckling state. Fig. 1. Cross-section of the column.
Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313 309

Table 1 (17). For a multi-layer symmetric laminate, the coeffi-


Comparison of buckling loads for channels cients K13 = A16  0, K32 = A26  0, thus, as has been al-
Pcr [kN] ready mentioned, it follows that N  ð0Þ  0.
3k
Presented method ANSYS 5.4 j(1)–(2)j/(2) The value of the coefficient a0 in Eq. (22) depends
(1) (2) (3) only on N ð0Þ , as N 1ð0Þ  e1ð0Þ 6¼ 0 and N 2ð0Þ  e2ð0Þ þ N 3ð0Þ  e3ð0Þ þ
1k
ð0Þ ð0Þ ð0Þ ð0Þ ð0Þ ð0Þ
1.589 (4) 1.547 2.7% N 4  e4 þ N 5  e5 þ N 6  e6 ¼ 0, following the nota-
1.649 (5) 1.591 3.6%
1.706 (3) 1.681 1.5% tion of [1]. The values of ar (Eqs. (21) and (22)) do not
ð0Þ
1.802 (6) 1.743 3.4% depend on N i for i = 2, 3, . . . , 6, either.
2.022 (7) 1.929 4.8% Below, the results of calculations for thin-walled
beam-columns with the cross-sections under compres-
sion and bending are presented in Fig. 2.
Completely arbitrary arrangements of layers have An influence of the curvature radius R of the flange
been assumed, as in this case the stiffness matrix K (or of the cylindrical shell segment rise) on the global
(Eq. (6)) is filled up as a whole, i.e. Kij 5 0 for i, and local critical load Ncr [N/m] of the column with
j = 1, 2, . . . , 6. the length ‘ = 2000 mm, made of a three-layer composite
Table 1 shows a comparison of the results for the first of the following structure of the layers:
five lowest values of critical loads (eigenvalues) of the
channel under compression. For the present modal • h1 = 0.2 mm, E1 = 200 GPa, m12 = 0.3, E2 = E1, G =
method, a number of half-waves that generate along E1/(2(1+m12)) steel;
the length of the channel is given in brackets. The buck- • h2 = 0.6 mm, E = 139.3 GPa, m12 = 0.3, E2 = 11.1
ling modes (eigenmodes) obtained by both the methods GPa, G = 6 GPa composite [22,23];
are identical in practice. • h3 = 0.2 mm, E1 = 200 GPa, m12 = 0.3, E2 = E1, G =
Next, the present method has been compared with the E1/(2(1+m12)) steel;
results obtained by other authors. An example has been
taken from Shen and Williams [22] for an unstiffened (where h = h1 + h2 + h3) has been analysed. Structures
six-layer graphite/epoxy laminated square box column made of a composite with such an arrangement are
with the length ‘ = 305 [mm] and the width b = referred to as metal-composite-metal structures.
25.4 [mm]. The layer material properties were: E1 = The curves in Fig. 3 denote, respectively: 1—global
139.3 [GPa], E2 = 11.1 [GPa], G = 6.0 [GPa], m12 = 0.3 flexural-torsional buckling, 2—local antisymmetric
and the thickness hply = 0.134 [mm]. The buckling loads buckling, 3—global flexural buckling, 4—local symmet-
calculated by the present method are listed in Table 2 ric buckling.
and compared with the numerical results given by Shen The negative dimensionless curvature b1/R < 0
and Williams [22] and with the experimental and numer- (b1 = 100 mm) corresponds to the flange deflected to-
ical results given by Wang et al. [25]. wards the inside of the column contour, whereas b1/
The present method has been also compared with the R > 0—to the flange deflected towards the outside of
results obtained by Loughlan [15] for multi-layer sym- the column contour.
metric and antisymmetric plates under compression. A The value of the critical load corresponding to the
very good agreement between the results has been ob- global flexural-torsional buckling decreases with an in-
tained as well. crease in the curvature b1/R by nearly 1.5 times. The
Further calculations have been limited to thin-walled value of the global flexural critical load undergoes slight
multi-layer symmetric and antisymmetric structures. In changes. A similar situation occurs for the local anti-
the case of an antisymmetric laminate, the zero state symmetric buckling. The shell element exerts the stron-
component N  ð0Þ ¼ 0 and the determined values of the gest influence on the local symmetric buckling. Even a
3k
critical load are precise within the asymptotic method slight flange curvature results in a rapid increase in the
applied and the pre-critical distribution field assumed critical load.

Table 2
Comparisons of buckling loads of graphite/epoxy laminated square box columns
Pcr [kN]
Lay-up Presented method Shen and Williams (1993a) Wang et al. [25] calculated Wang et al. [25] experiments
(±15/0)S 9.09 8.93 8.59 8.17
(±15/0)2T 8.71 8.48 7.74 8.01
(±152/02)T 7.10 8.17 6.87 7.46
(±45/(±15)2)T 11.02 9.72 10.91 8.81
310 Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313

theoretical load-carrying capacity, determined within


the first-order non-linear approximation, to the mini-

b1/R<0
mum critical load value, that is to say, Ns/Nm, where
Nm = min (Ng, Nl), are presented. In each case the sign
of the imperfection has been chosen in the most unfa-
vourable way, i.e. so that Ns would have its minimum
1
b1/R>0
value (for more a detailed analysis, see [8,10,13]).
In Table 3 the ratios of the global critical load to the
local one, for the same conditions along the symmetry

100
axis of the cross-section, i.e. the ratios of the critical
loads: global flexural-torsional (antisymmetric) to local
antisymmetric; global flexural (symmetric) to local sym-
1

metric, are included as well.


12. 5
100
The interaction of the global flexural buckling mode
with the local symmetric one has turned out to be the
Fig. 2. Cross-section of the column with the length ‘ = 2000 mm. most dangerous for 0.5 6 b1/R 6 0.25.
The further analysis has been limited to thin-walled
structures made of an eight-layer composite of a sym-
Ncr [N/m] metric and antisymmetric structure.
0.8
Each layer is made of the same material characterised
0.7
by the following data [26]:
0.6
1
0.5 2 E1 ¼ 129:1 ½GPa; E2 ¼ 9:43 ½GPa; G ¼ 5:39 ½GPa;
3
0.4
4
m12 ¼ 0:288; hply ¼ 0:125 ½mm
0.3
0.2 Individual layers were turned by a directed angle H
0.1 with respect to the assumed local system of co-ordinate
0
axes, which overlapped the contour of each plate-panel
-0.5 -0.3 -0.1 0.1 0.3 0.5 ([6,14–16], Wang et al., 1994).
b 1/R Again, the previously discussed channel of the follow-
Fig. 3. Critical load Ncr versus b1/R. The curves denote, respectively:
ing geometrical dimensions (Fig. 1):
1—global flexural-torsional buckling, 2—local antisymmetric buck- b1 ¼ 25 ½mm; b2 ¼ 50 ½mm; h ¼ 1 ½mm; ‘ ¼ 650 ½mm;
ling, 3—global flexural buckling, 4—local symmetric buckling.
made of a composite with a symmetric arrangement of
The lowest value of the critical load of the beam-col- layers [H/H/H/H/H/H/H/H]T and an antisym-
umn under analysis occurs for: local symmetric buckling metric arrangement of layers [H/H/H/H/H/H/H/
mode for 0.5 6 b1/R 6 0.25; local antisymmetric buck- H]T, has been analysed.
ling mode for b1/R = 0.5. In Fig. 4 a variability of critical loads for three vari-
The further analysis has concerned the interactive ous buckling modes, i.e.: global flexural, global flexural-
buckling and has been aimed at determination of the torsional and the lowest local, denoted by the indices f,
sensitivity of the columns under discussion to global ft and l, respectively, as a function of the angle H for
and local geometrical imperfections. It has been as- both the ways of arrangement of channel layers, is
sumed that the beam-column has the following initial presented.
deflections As can be easily seen, the values of global loads are
jfg j ¼ 1:0, jfl j ¼ 0:2, where the index denotes: g—glo- the same practically for the case A and S for the same
bal buckling mode, l—local one. In Table 3 the theoret- angle H. Maximum values of global critical loads occur
ical load-carrying capacity Ns [N/m], the ratios of the for the angle 0 6 H 6 15 [deg], whereas maximum

Table 3
Load carrying capacity within the first-order approximation and the imperfections jfg j ¼ 1:0; jfl j ¼ 0:2
b1/R Antisymmetric modes Symmetric modes
Ns [N/m] Ns/Nm Ng/Nl Ns [N/m] Ns/Nm Ng/Nl
0.5 0.139 0.784 1.949 0.138 0.858 4.077
0.0 0.115 0.766 1.626 0.0621 0.797 9.085
0.5 0.115 0.762 1.344 0.133 0.858 4.423
Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313 311

Ncr /h [N/mm2] Ncr/h [N/mm2]


140
160

120
140
100
1A 120
2A
80 1S
3A
100
1S 2S
60
2S 3S
3S 80
40 1SS
60 2SS
20 3SS
40
0 Θ [deg]
0 10 20 30 40 50 60 70 80 90
20
Fig. 4. Influence of ply-angle H on critical compressive loads for
‘ = 650 mm. The curves: 1—flexural buckling, 2—local buckling, 3— 0 Θ [deg]
flexural-torsional buckling. The notes: A—antisymmetric angle-ply 0 10 20 30 40 50 60 70 80 90
laminates H/H/H/H/H/H/H/H; S—symmetric angle-ply lami-
nates H/H/H/H/H/H/H/H. Fig. 5. Influence of ply-angle H on critical compressive loads for
‘ = 650 mm. The curves: 1—flexural buckling, 2—local buckling, 3—
flexural-torsional buckling. The notes: S—symmetric angle-ply lami-
values of local critical loads for H  45 [deg]. Maximum nates H/H/H/H/H/H/H/H; SS—symmetric angle-ply laminates
differences between the values of critical local loads for H/H/0/0/0/0/H/H.
both the types of layer arrangement occur also for
H = 45 [deg] and are equal to approx. 8%. local buckling mode, 3—flexural-torsional buckling
Then sensitivity to imperfections of these arrange- mode.
ments of layers for H = 45 [deg] has been analysed, i.e. As one can see, a change in the way of layer arrange-
theoretical values of load carrying capacity have been ment from S to SS, according to the notations assumed
determined within the first-order approximation for in Fig. 5, can have a significant influence on the values of
the assumed initial imperfections jff j ¼ 1:0; jfft j ¼ global loads. For the channel under consideration and
1:0; jfl j ¼ 0:2 for a smaller length of the channel the angle H = 45 [deg], an increase in the values of crit-
‘ = 390 mm. A decrease in the length of the analysed ical load of the flexural buckling mode is more than 3.5
channel has of course caused an increase in the values times, whereas for the flexural-torsional buckling mode,
of global loads (flexural and flexural-torsional buckling it is more than 2.5 times. An increase in global loads re-
mode), whereas the values of local critical loads are sults from a significant increase in the global flexural
practically the same. The obtained results are presented stiffness of the cross-section. The local flexural stiffness
in Table 4. As can be seen, sensitivity for both the ways of individual walls remains practically unaltered in com-
of layer arrangement is the same in practice. parison with regular symmetric angle-ply laminates.
Next, another way of a symmetric arrangement of Together with a change in the angle H, the number of
layers [H/H/0/0/0/0/H/H]T—denoted as SS—has halfwaves m that corresponds to the lowest value of the
been analysed. In Fig. 5 a change in critical loads as a local buckling changes as well. For the plies arranged at
function of the angle H for a channel with the length the angle H = 45, the number of halfwaves that form
‘ = 650 mm is shown. For comparison, the results for along the column length is m = 10, whereas for the angle
the above mentioned symmetric arrangement of layers, H = 0—m = 5, and for H = 90—m = 19, respectively.
i.e. [H/H/H/H/H/H/H/H]T—denoted as S—have This result agrees with the results obtained by the
been presented as well. The curves have been denoted, authors (see papers by [8,13]). As can bee seen from this
respectively: 1—flexural buckling mode, 2—the lowest comparison, the fact that a change in the arrangement of

Table 4
Load-carrying capacity within the first-order approximation for the channel length ‘ = 390 mm and the imperfections j fft j¼ 1:0; j ff j¼ 1:0; j fl j¼ 0:2
H = 45 [deg] Flexural-torsional global Flexural global Local buckling Load-carrying
buckling mode buckling mode mode capacity
Nft/h [N/mm2] Nf/h [N/mm2] Nl/h [N/mm2] Ns/h [N/mm2]
Antisymmetric angle-ply laminates 59.55 76.60 68.72 34.78
Symmetric angle-ply laminates 58.72 76.69 63.87 32.40
312 Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313

layers can result in a change in the number of halfwaves Ncr /h [N/mm2]


that form along the longitudinal direction should be ta- 50
ken into consideration in the analysis. This can explain 45
the practically symmetric distribution of the lowest val- 40
ues of critical loads of the local buckling with respect to 35
the value H = 45 in Figs. 4 and 5. 30
1
The above observation allows for a conclusion that a 25
rational way of designing the way of laminate ply 2
20
arrangement is purposeful (Fig. 6). 15
In the next figure (Fig. 7) a variability of global flex- 10
ural and lowest local critical loads for the cylindrical 5
Θ [deg]
shell under compression and characterised by the fol- 0
0 10 20 30 40 50 60 70 80 90
lowing geometrical dimensions:
Fig. 7. Global flexural and lowest local critical loads for cylindrical
R ¼ 430 ½mm; ‘ ¼ 4130 ½mm; h ¼ 1 ½mm
shell.
made of an 8-layer symmetric laminate [H/H/H/H/
H/H/H/H]T with the following properties [26]:
The present solution method allows one to analyse
E1 ¼ 129:1 ½GPa; E2 ¼ 9:43 ½GPa; G ¼ 5:39 ½GPa;
the modal buckling of plate-shell structures made of
m12 ¼ 0:288; hply ¼ 0:125 ½mm multi-layer composites.
as a function of the angle H is presented. The identical
dimensional ratios have been assumed as in the cylindri-
cal vessel experimentally investigated by Magnucki et al. 5. Conclusions
[17]. In this figure the lowest critical values of local
buckling of the shell are shown. The curves are denoted, A modified method of solution to the stability prob-
respectively: 1—flexural buckling mode, 2—local one. lem and determination of initial post-buckling equilib-
As can be seen, the critical values of local buckling rium paths for plate-shell multi-layer structures has
reach the maximum for H = 15 [deg] and for H = 75 been postulated. The present method can be easily
[deg], whereas the minimum for H = 0 and 90 [deg]. adopted in rational design of layered thin-walled struc-
The global critical load reaches the minimum for tures within the linear and non-linear theory of buckling
H = 50 [deg], whereas the maximum for 0 6 H 6 20 of multi-layer structures.
[deg] and H  90 [deg]. The most advantageous case
with respect to maximum global and local critical loads
is the arrangement of layers at the angle H = 15 [deg] for Appendix A. Kinematic and static conditions of
the given geometrical ratios of the cylindrical shell and interaction between adjacent plates
for the way of its arrangement under consideration. In
the case of the shell, sensitivity to imperfections cannot According to (14), kinematics and static conditions of
be analysed within the first-order approximation. the interaction on the longitudinal edges of adjacent
plates (for ‘‘i’’ and ‘‘i + 1’’) can be written as (Fig. A.1):

20
m
15

10

5
ϕ
Θ [deg]
0
0 15 30 45 60 75 90

Fig. 6. The number of halfwaves m for the lowest buckling modes


versus the angle H. Fig. A.1. Prismatic plate structure and the local co-ordinate system.
Z. Kolakowski, M. Krolak / Composite Structures 76 (2006) 303–313 313

0 þ [7] Koiter WT. General theory of mode interaction in stiffened plate


u1ðiþ1Þ j ¼ u1i j
0 þ and shell structures. WTHD, Report 590, Delft, 1976.
u2ðiþ1Þ j ¼ u3i j þ sin u þ u2i j cos u [8] Kolakowski Z, Krolak M. Interactive elastic buckling of thin-
0 þ þ walled closed orthotropic beam-columns. Eng Trans 1995;
u3ðiþ1Þ j ¼ u3i j cos u  u2i j sin u
0 þ 43(4):571–90.
u3ðiþ1Þ;y j ¼ u3i;y j [9] Kolakowski Z, Kowal-Michalska K. Selected problems of insta-
0 þ
N 5ðiþ1Þ E0 h2iþ1 j  N 5i E0 h2i j ¼ 0 bilities in composite structures. Technical University of Lodz, A
0 þ þ series of monographs, 1999:222p.
N 2ðiþ1Þ E0 hðiþ1Þ j  N 2i E0 hi j cos u  Q2i j sin u ¼ 0 [10] Kolakowski Z, Teter A. Interactive buckling of thin-walled closed
0 þ þ elastic beam-columns with intermediate stiffeners. Int J Solids
Q2ðiþ1Þ j þ N 2i E0 hi j sin u  Q2i j cos u ¼ 0
Struct 1995;32(11):1501–15.
0 þ
N 3ðiþ1Þ E0 hiþ1 j  N 3i E0 hi j ¼ 0 [11] Kolakowski Z, Teter A. Interactive buckling of thin-walled beam-
columns with intermediate stiffeners or/and variable thickness. Int
ðA:1:1Þ J Solids Struct 2000;37:3323–44.
where [12] Kollar LP. Flexural-torsional buckling of open section composite
columns with shear deformation. Int J Solids Struct 2001;
N y ¼ N y þ N y v;y þ N xy v;x N xy ¼ N xy þ N y u;y þ N xy u;x 38:7525–41.
[13] Krolak M, Kolakowski Z. Interactive elastic buckling of thin-
Qy ¼ M y;y þ 2M xy;x þ N y w;y þ N xy w;x u  ui;iþ1 walled open orthotropic beam-columns. Eng Trans 1995;43(4):
591–602.
ðA:1:2Þ
[14] Loughlan J. The buckling performance of composite stiffened
panel structures subjected to combined in-plane compression and
shear loading. Compos Struct 1994;29:197–212.
[15] Loughlan J. The effect of membrane–flexural coupling on the
Appendix B compressive stability of antisymmetric angle-ply laminated plate.
In: Proceedings of the Third International Conference on Thin-
Inner sectional forces of the pre-critical state for the Walled Structures. Elsevier; 2001. p. 507–14.
assumed homogeneous field of displacements (11) are [16] Loughlan J. The shear buckling behaviour of thin composites
plates with particular reference to the effects of bend-twist
expressed by the following relationships:
coupling. Int J Mech Sci 2001;43:771–92.
 ð0Þ ¼ ðK 11  K 2 =K 22 ÞD
N [17] Magnucki K. Wytrzymałość i optymalizacja zbiorników
1k 12
cienkościennych. PWN, Warsaw-Poznań, 16p. /in Polish/, 1998.
 ð0Þ ¼ 0
N [18] Matsunaga H. Buckling analysis of multilayered angle-ply com-
2k
posite plates. In: Proceedings of the Third International Confer-
 ð0Þ ¼ ðK 31  K 32 K 21 =K 22 ÞD
N ence on Thin-Walled Structures. Elsevier; 2001. p. 507–14.
3k
ðA:2:1Þ [19] Pignataro M, Luongo A. Asymmetric interactive buckling of thin-
 ð0Þ ¼ ðK 41  K 42 K 21 =K 22 ÞD
N walled columns with initial imperfection. Thin-Walled Struct
4k
1987;3:365–86.
 ð0Þ ¼ ðK 51  K 52 K 21 =K 22 ÞD
N 5k [20] Pignataro M, Luongo A. Multiple interactive buckling of thin-
 ð0Þ ¼ ðK 61  K 62 K 21 =K 22 ÞD walled members in compression. In: Proceedings of the Interna-
N 6k tional Colloquiim on Stability of Plate and Shell Structures,
Ghent, University Ghent, 1987b, p. 235–40.
[21] Rand O. Nonlinear in-plane warping deformation in elastically
coupled open thin-walled beams. Comput Struct 2001;79:281–91.
References [22] Shen HS, Williams FW. Postbuckling analysis of stiffened
laminated box columns. J Eng Mech 1993;119(1):39–57.
[1] Byskov E, Hutchinson JW. Mode interaction in axially stiffened [23] Shen HS, Williams SF. Postbuckling analysis of stiffened lami-
cylindrical shells. AIAA J 1977;15(7):941–8. nated panels loaded in compression. Int J Solids Struct 1993;
[2] Calcote LR. The analysis of laminated composite structures. New 30(12):1589–601.
York: Van Nostrand Reihold Company; 1969. [24] Walker M, Adali S, Verijenko VE. Optimal design of symmetric
[3] Chandra R, Raju BB. Postbuckling analysis of rectangular angle-ply laminates subject to nonuniform buckling loads and in-
ortothropic plates. Int J Mech Sci 1973;16:81–9. plane restraints. Thin-Walled Struct 1996;26(1):45–60.
[4] Dawe DJ, Wang S. Buckling of composite plates and plate [25] Wang C, Pian THH, Dugundji J, Lagace PA. Analitycal and
structures using the spline finite strip method. Compos Eng experimental studies on the buckling of laminated thin-walled
1994;4(11):1099–117. structures. In: Proc. AIAA/ASME/ASCE/AHS 28th Structures,
[5] Flügge W. Stresses in shells. New York: Springer-Verlag; 1967. Struct. Dyn. and Materials Conf., Part 1, 1987, p. 135–40.
[6] Jones RM. Mechanics of composite materials. International [26] Wilczynski AP. Polimerowe kompozyty włókniste—własności,
Student Edition. Tokyo: McGraw-Hill Kogakusha, Ltd.; 1975. struktura, projektowanie. Scientific and Technical Publishers,
365p. Warsaw, 297p. (in Polish), 1996.

You might also like