Professional Documents
Culture Documents
Abstract Buckling and post buckling analysis of composite to confirm the results produced by the other two methods;
T
laminated structures having delaminations were studied therefore, in here we will focus on the analytical and numerical
numerically. The analysis were performed using the generalized works and will also mention their experimental validations.
differential quadrature method. Several one- and two- Later we will further subdivide the works based on different
dimensional models were developed and showed to be capable of sub categories, such as: one-, two- or three-dimension
predicting the buckling and post buckling responses of composite delamination modeling, single or multiple delaminations, etc.
beams and plates. Prediction of the buckling strength of the One of the earliest works in delamination buckling and growth
delaminated composite beams and the examination of the analysis of beams was came out by Chai et al [1]. They studied
ES
influencing factors were came out by employing several models
.The effects of the shear deformation and the bending-stretching
coupling were added by incorporating a shear deformation beam
theory. The accuracy and efficiency of the proposed method were
evaluated through several case studies. In addition to the effect of
the behavior of an isotropic homogeneous beam-column under
axial compression from a thin film model to the general case
(when the supporting base laminate buckles globally, so that
the zero-slope boundary condition for the thin sub laminate
shear deformation, bending- stretching coupling, the influence of becomes invalid). Simitses et al [2] also employed a similar
material properties, lamination sequence and fiber- orientation, model to study delamination buckling. They studied the effect
through-the-thickness and through-the-length locations of of the location, length and thickness of delamination on the
delaminations on buckling and post buckling responses were buckling load of a beam with clamped and simply supported
investigated. For the post buckling analysis, the GDQM ends, having a single across-the-width delamination. The
combined with an arc-length strategy was used to mode1 the post perturbation method was used to solve the buckling equation.
buckling analysis of composite beams having single or multiple Kardomateas and Shmueser [3] used perturbation technique to
delaminations. The nonlinear Von Karman strain-displacements analyse the compressive stability of a one-dimensional across-
A
and the exact curvatures were used to formulate the problem. the-width delaminated orthotropic homogeneous elastic beam.
The results show that the GDQ technique can be used as a They also considered the transverse shear effect on the
powerful, reliable, accurate and efficient numerical tool in buckling load and post-buckling behavior of the beam. Using a
assessing the buckling and post buckling responses of variation energy approach and a shear-deformation theory,
delaminated composite structures. Beside the excellent quality of Chen [4] formulated the same problem. According to his
results that obtained through GDQM, the method is relatively results, inclusion of the shear deformation causes reduction in
simple to formulate, and it requires less effort to implement.
the buckling and ultimate strength of delaminated composite
IJ
asymptotic solution of the post buckling behavior of the beam. Shu and Du [23] used DQM to address the analysis of
In both of the above works the contact between the free vibration of laminated composite cylindrical shells.
delaminated layers and delamination growth was neglected. Application of DQM in nonlinear analysis of structures has
Sheinman et a1 [8] solved the differential equations of a been also reported in several studies. Bert et al [24] analyzed
delaminated composite beam under arbitrary loading and the large deflection problem of a thin orthotropic rectangular
boundary conditions with a finite difference method. plate in bending. The three nonlinear differential equations of
equilibrium of the plate were transformed into differential
Bending-stretching coupling was taken into account quadrature form and solved numerically using the method of
which was show to significantly influence the buckling loads. Newton-Raphson. Lin et al [25] used the same procedures to
Sheinman and Soffer [9] used the finite difference method to solve the problem of large deflection of isotropic plates under
solve the post buckling problem of an imperfect composite thermal loading. In their analysis they used the generalized
laminate having a through-the-thickness delamination. They differential quadrature of Shu and Richards [26]. Moradi and
employed the commonly used one-dimensional beam model Taheri used equally and unequally spaced sampling points in
and formulated the response of the beam by dividing the the delamination buckling analysis of beams. Bert and Malik
delaminated beam into four regions. Using the Von Kaman [27] also examined the equally spaced grids as well as
kinematic approach, the resulting non-linear differential unequally spaced in several structural mechanics applications.
equations were solved by the method of Newton-Raphson. The domain is divided into N discrete points and cij are the
Davidson [10] used the Rayleigh-Ritz method to compute the weighting coefficients of the derivative. i values are important
buckling strains of a composite laminate containing an factors control the quality of the approximation, resulting from
elliptical delamination. The influence of the bending stretching the application of GDQM.
coupling behavior of the delaminated region and the Poisson's
T
ratio mismatch between the delaminated and base regions were N
plates. Bert et al [16] , [17] used DQM for static and free Cij( m1)
Ciim m Ciim1Cij i j m 2 , 3 , ...., N 1 (6)
vibration analysis of anisotropic plates, while Laura and xi x j
Gutierrez [18] used the method in vibration analysis of
rectangular plates with non-uniform boundary conditions. i, j 1, 2 , .... , N
Sherboume and Padney [19a] and Padney and Sherboume N
[19b] used DQM to analyze the buckling of composite beams Ciim Cijm i 1, 2 , ..., N
and plates. They used different number of grid spacing in their j 1, j i
(7)
analysis. The same problem was addressed by Wang [20], who
also used different grid spacing. He found that employing Also use of zeros of the shifted Legendre polynomials good
uniform grid spacing could result in an inaccurate solution; results, while some authors have chosen the grid points based
therefore, caution should be exercised when using such on trial (Sherbourne and Pandey [19]).
spacing. Liew et al [21] used the method for the analysis of
thick symmetric cross-ply laminates with first order shear
deformation while Kang et al [22] used it to address the
vibration and buckling analysis of circular arches.
B. Treating Boundary Conditions and the compatibility conditions at the subdomain interfaces
Essential and natural boundary conditions can be should be taken into consideration and satisfied.
approximated by DQM; they are treated the same way as the
differential equations are. In the resulting system of algebraic D. Differential Quadrature Formulation Of Delamination
equations from GDQM, each boundary condition replaces the Buckling
corresponding field equation. Note that at each boundary point nx nx ny
only one boundary condition be satisfied. However, in the case D11 Aik4 Wkj 4 D16 Aik(3) B (jm1)Wkm
of fourth order differential equations (such as those for beams k 1 k 1 m 1
and plates) or the higher order, one must satisfy more than one nx ny nx ny
boundary condition at each boundary. Wang and Bert [28] 2D12 2 D66 A B W
ik
2 3
jm km 4 D26 Aik(1) B jm3Wkm
k 1 m 1 k 1 m 1
proposed a method in which the weighting coefficient matrices ny nx
(12)
for each order derivative can be developed by incorporating the D22 B jm Wim N x A Wkj
4 ( 2)
dealing with the boundary conditions other than simply N y B jm2 Wim 2 N xy Aik(1) B (jm1)Wkm 0
supported or clamped. Malik and Bert [29] also explored the m 1 k 1 m 1
benefits and the limitations of this method for various types of i 1 , ... , n x j 1 , ... n y
boundary conditions. Shu and Du [23] proposed another
approach in which the derivative representing the two opposite ʎ is the normalized buckling load which is the ratio of the
edges are coupled to provide two solutions at two neighboring compressive load to the Euler buckling load of a beam with
T
points to the edges. a and b are dimensions and nx , ny are clamped ends. Here N represent force vector. where A , B and
number of gid points (test points) in direction of X and Y D are the stiffness terms.
respectively where A , B and D are the stiffness terms and W
represent Transverse deflection. S bb S bd wb 0 0 wb
S S w B B w
For simply support db dd d db dd d
(13)
x 0, a
nx
Wij 0
ny
D11 Aik( 2)Wkj D12 B (jm2)Wim 2 D16
k 1 m 1
nx
k 1
ny
m 1
ES Aik(1) B (jm1)Wkm 0
(8)
Pcr .a 2
E22.t 3
(14)
i 1 , ... , nx , j 1, ny
(11) for a plate can be written as (Whitney (1987)):
M x , xx M xy , xy M y , yy N x .W , xx
C. Domain Decomposition
N y .W , yy N xy .W , xy 0 (15)
For problems having complicated domains such as those
in delaminated plates or beams or plates with cutouts, the
N x , x N xy , y 0
concept of domain decomposition may be used for solving the
problems. With this concept, first the domain is divided into N y , y N xy , x 0 (16)
several subdomains. A local mesh can be generated for each
subdomain with more density near the boundaries. Then, the where w represents the out of plane deflection. Substituting
generalized differential quadrature representation of the stress-strain and strain-displacement relations yields the
governing differential equations for each domain can be following equilibrium equations in terms of in-plane
formulated. In this approach, each region may have different displacements u and v and out of plane deflection w:
number of sampling points. Finally, the boundary conditions
A11U , xx 2 A16U , xy A66U , yy A16V , xx the post buckling response of laminated plates. Then the
(17)
( A1 2 A66 )V , xy A26V , yy B11W , xxx generalized differential quadrature formulation of imperfect
3B16W , xxy ( B12 2 B66 )W , xyy B26W , yyy 0 laminated composite beam having a single or multiple
delaminations will be presented.
A16U , xx 2 A26V , xy A66V , xx A22V , yy
(18)
( A1 2 A66 )U , xy A26U , yy B16W , xxx 3B26W , xxy A. Arc-Length Method
( B12 2 B66 )W , xxy B22W , yyy 0 The conventional Newton-Raphson iterative algorithms,
which have been used extensively to solve nonlinear
D11W , xxxx 4 D16W , xxxy 2( D1 2 2 D66 )W , xxyy problems, develop difficulties when tracing the points near
4 D 26W , xyyy B11U , xxxx 3B16U , xxy ( B12 2 B66 )U , xyy (19) buckling or limit points of the equilibrium path. In these
B 26U , yyy B16V , xxx ( B11 2 B66 )V , xxy algorithms load controls the path; that is, the load is
3B 26V , xyy B 22V , yyy N xW , xx 2 N xy W , xy N y W , yy 0 incremented by constant values and the iteration is performed
on the nodal displacements, only. In these algorithms, the
solution often halts near the limit points. To resolve this
where A , B and D are the stiffness terms. The first two problem, several strategies have been suggested by researchers
equations are concerned with the in-plane equilibrium while (see for example Clarke and Hancock [31]). In some of the
the last one represents the buckling equation. A close look at suggested methods, the iterative procedure is controlled by
equation (19), indicates that the coefficients of in-plane
displacement, where a constant displacement increment or a
T
displacements u and v are those which are only associated with
the bending-stretching coupling terms. So, if one cm eliminate combination of, controls the equilibrium path (see Haisler and
these terms from (17,18,19), then, one can determine the Stricklin [32], Batoz [33] and Powell and Simons [34]). In
buckling strains by only one equation (Le., 19). This will addition to the above methods, there is another strategy that
dramatically reduce the computational efforts. For symmetric uses a constraint to Limit the load increment. The method,
delaminated plates, Bij=O and thus, the extension and bending which was introduced by Wempner and by Riks
are uncoupled. One therefore, only needs to solve equation independently, uses the arc length of the load-displacement
(17,18,19). But in general, the delamminated regions do not
possess mid-plane symmetry (even if the laminate is
ES
symmetric), and therefore, bending-stretching coupling exists.
In this case, some researchers have recommended the reduced
path as the constraint equation, and therefore is referred to as
the arc-length method. The method was modified by Crisfield
(1981, 1983), thus being referred to as the "modified Riks"
bending stiffness approximation. In this approach, the effect of method. There also exist other types of constraint equations
coupling terms Bij is implicitly accounted for by reducing the such as the "constant external work". Clarke and Hancock [31]
flexural stiffness matrix D such as:
performed a detailed study on the available incremental
iterative strategies for the nonlinear analysis.
D D BAB 1
(20)
IV. CONCLUSION
The accuracy of the reduced bending stiffness In the "fig 1", the ordinate represents the applied load
A
approximation in analyzing the delamination buckling of normalized with respect to the buckling load. For GDQM,
composite laminates was examined by Dost et al [30]. They 13x13 grid points were used to discretized the plate. The results
performed finite element analysis on delaminated composite are shown in "Fig 1" again the GDQM shows excellent
plates and found that the results with reduced bending stiffness agreement with Lanzo et aI [34] finite elements method. Also,
agree well with those with the usual bending and bending- it can be seen that increasing the amplitude of imperfection
stretching stiffness matrices. dramatically influences the response of the rectangular plate.
IJ
T
[36]. It is observed that decrease normalized buckling load buckling mode can occur at specific a/L ratios Lee [37];
with growth fiber orientation angle in three methods. ( a = therefore, by imposing the symmetry condition, accurate
delamination length , h = Thickness of the upper delaminated simulation of the buckling modes for all a/L ratios is not
region , t = thickness of beam and plate ). achievable. The result reported by Lee et all are provided for
The normalized buckling loads are also shown graphically the two possible conditions. Also, note that the "non
in "Fig 3". The results are derived using different rotations at symmetric" terminology used to identify some of the GDQM
the delamination fronts. In all of the delamination length ratios
decrease normalized buckling load. For GDQM, 13x13 grid
ES
points were used to discretized the beam. Beyond that rang
one can show a rapid decrement in the value of normalized
results conforms to the terminology used by Simitses et al. As
it can be seen, there is an excellent agreement between the
results obtained by GDQM and those of Simitses et al, Chen,
and Lee et al. Indeed, the GDQM results are closer to the
buckling load for delamination length ratio less than 0.2.
Table 1 tabulates the buckling loads obtained by GDQM results obtained based on the analytical solutions (Simitses et
for a specially-orthotropic composite laminate containing a al and Chen), than those calculated based on the finite element
delamination at its mid-plane (Le., h/T=0.5). This problem is analysis of Lee et al. Also note that, while Lee et al used two
analyzed for various delamination length ratios (a/L). The different models to obtain the symmetric and anti symmetric
buckling loads are normalized with respect to the Euler buckling loads, the same results could be achieved by GDQM
buckling load of a perfect column. The "symmetric" and "anti- by only using different rotations at crack tips.
symmetric" notations in Table1 conform to the terminology
A
used in past sections. Smitses [2] and Chen [4] results present TABLE 2. COMPARISON OF THE FIRST THREE NORMALIZED
BUCKLING LOAD FOR A BEAM WITH SINGLE DELAMINATION,
for symmetric beam. Note that the results reported by Chen
H/T=0.5
were obtained by imposing the symmetry condition in the
axial direction, while Simitses et al did not apply such an
present study (GDQM) Lee et al. (1993)
imposition.
buckling load global local anti
IJ
T
these buckling loads with those of Lee et al shows that the first GDQM, however, produces acceptable results for all the
buckling loads in Table 2 is essentially the minimum value of buckling modes. Based on the results of the author's
the three possible mode shapes. investigations presented in the previous, the following
The influence of the longitudinal delamination position concluding remarks are made:
on the buckling response of the beams is shown in"Fig 4" for
beam with h/T = 0.5. For the delamination length ratios 0.2 < 1) The paper reports the first attempt in applying the
ES
a/L < 0.6, the buckling response varies depending on the
location of the delamination along the span of the beams. As
expected, the highest buckling strength is obtained when the
delamination is at the center of the span of the beam. The
generalized differential quadrature analog to the
delamination buckling of composite laminated
structures.
2) GDQM was show to be capable of determining ail
behavior is consistent. throughout, including in beams with possible buckling loads and their corresponding
h/T ratios other than 0.5, as illustrated in "Fig 4". modes.
" Figure 5" illustrates the effects for three different types of 3) For a forth order differential equation (i.e.,
composites, namely: graphite/epoxy, glass/epoxy and Kevlar differential equilibrium equation of the Euler beam),
epoxy, having various delamination length ratios (a/L). Al1 the use of the technique was found to enhance the
composites have ply sequence of [(90/0)4 /(90/0)12] with the accuracy of the solution. For such a problem, the use
delamination thickness ratio of h/T=0.25. Normalization of the of the non-equally spaced grid schemes such as the
A
buckling loads in "Fig 5" was performed with respect to the Chebyshev or the Legendre, which allocate denser
modulus of glass epoxy. The figure illustrates that the grid spacing near the boundaries, is recommended.
composite having the largest modulus (Le. the graphite/epoxy) 4) While using a grid spacing with an odd number of
exhibits the highest buckling capacity regardless of nodes is highly recommended, the author's
delamination length, as it was expected. experience indicates that a relatively low number of
grid points can produce results with acceptable
accuracy in the buckling analysis of delaminated
IJ
T
the other numerical approaches Laminated Plates, Computers & Structures, vol. 57, pp. 715-719.
8) In order to solve for the buckling strains of an
elliptical delamination, the author used the reduced [7] Kardomateas, G. A., (1993) The Initial Post-Buckling and Growth
Behavior of Internal Delaminations in Composite Plates, Journal of Applied
bending stiffness approach. This resulted in a Mechanics, vol. 60, pp. 903-9 10.
considerable reduction in the computational effort. in
fact, instead of using the in-plane deformations (i.e. u [8] Sheinman, C., (1991) Generalized Differential-Internal Quadrature and
ES
and v) along with the out of plane deflection was
unknowns, it was shown that by using only one
parameter, w, DQM could provide accurate results
with minimal computational effort. The numerical
Application to the Simulation of incompressible Viscous Flows Including
Parallel Computation, Ph.D. Thesis, University of Glasgow
2341-2356.
method enjoys a mathematical foundation that is simpler to
digest compared to those associated with the other numerical [14] Farsa, J. E. and Roth, R. S., (1986) in: Methods ideal Publishing,
methods. Moreover, the implementation of the method into a Dordrecht, Netherlands.
computer code is relatively easier. The efforts needed for
setting up a problem, and for solving it with GDQM is also [15] Farsa, J. E. and Fei, Z., (1988) Delamination Buckling and Growth in a
relatively less in comparison to the other numerical methods. Clamped Circular Plate, AIAA Journal, vol. 26, pp. 438-445.
The only disadvantage of the proposed method at this stage of
its evolution is that it is not as robust as FEM in [16] Bert, C. W., Jang, S. K. and Striz, A. G., (1988) Two New Approximate
Methods For Analyzing Free Vibration of Structural Components, AIAA
accommodating problems with complicated geometries. This
Journal, vol. 26, pp. 612-618.
issue is however resolvable, as one can use the same type of
remedy that is available for treating the same anomaly that
[17] Bert, C. W., Wang, X. and Striz, A. G., (1994) Static and Free Vibration
exists with The FDM. Analysis of Beams and Plates by Differential Quadrature Method, Act
Mechanical, vol. 102, pp. 1 1-24.
T
in Nonlinear Structural Analysis by the Self connecting Method, International
Journal for Numerical Methods in Engineering, vol. 1 1, pp. 3-10.
[22] Kang, S. K., Bert, C. W., and Striz, A. G., (1989) Application of
Differential Quadrature to Static Analysis of Structural Components,
[33] Batoz, J. -L. and Dhatt, G., (1979) Incremental Displacement Algorithms
International Journal for Numerical Methods in Engineering, vol. 28, pp. 561-
for Nonlinear Problems, International Journal for Numerical Methods in
577.
Engineering, vol. 14, pp. 1262- 1267.
ES
[23] Shu, C. and Du, H., (1997b) Implementation of Clamped and Simply
Supported Boundary Conditions in the GDQ Free Vibration Analysis of
Beams and Plates, International Journal of Solids and Structures, vol. 34, pp. 8
19-835.
[34] Lanzo, W. Y., Cheung, Y. K., and Tham, L. G., (1986) Spline Finite Strip
Analysis of General Plates, ASCE Journal of Engineering Mechanics, vol.
112, pp. 43-54.