www.elsevier.com/locate/compositesb
Abstract
The weakest part of fiber reinforced polymer composites (FRP) sandwich structures is commonly the interfacial zone between layers.
Under service loading, delamination failure will be most likely to initiate from these interfaces. Typical failure analysis for such sandwich
structures usually depends on the modeling details of the interface. In this paper, the sandwich panel was modeled as separated layers with
appropriate constrains imposed between them. These constrains guaranteed a smooth stress transfer from the skins to the core. The proposed
FEM model was applied to simulate the failure behavior of a FRP sandwich panel that is used in bus body. The simulation results were
compared with other numerical predictions and the experiment. It was concluded that this model is very efficient computationally for
analyzing the failure issues of FRP sandwich structures.
q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: A. Fiber reinforced polymer; B. Finite element method; C. Buckling; D. Laminates
2. Proposed layered FEM model to be able to transfer shear, tension, compression and
moment.
Since Arendt’s and Sun’s research [22] focused on the Shear stress in the upper interfacial layer (Fig. 2)
patched structural repair the interfacial adhesive layer was
›u ›w
not supposed to undertake large scale bending deformation. txz ¼ Ggxz ¼ G þ ¼ tð1Þ ð2Þ
xz þ txz
›z ›x
However, in our project, the bus panels are designed for ð2Þ
›v ›w
large flexural deformations under compression, hence the tyz ¼ Ggyz ¼ G þ ¼ tð1Þ
yz þ t ð2Þ
yz
›z ›y
interface layer should be able to transfer all of the possible
forces and moments. Indeed, our study to be discussed later where
indicates that moment transferability of the interface is very
›u u 2 u3
important. In the proposed FEM model herein, each lamina txz ð1Þ ¼ G ¼G 2
(skin or core) is modeled as a Mindlin shell structure [23]. ›z te
The interfaces between lamina are assumed to be able to ›w
txz ð2Þ ¼G ¼ 2Gðu3y 2 u2y Þ
transfer shear, tension, compression and moment. ABAQUS ›x
›v v 2 v3 ð2aÞ
is employed to implement this FEM model. The results are tyz ð1Þ ¼G ¼G 2
compared with those predicted by the integral method and ›z te
experimental data. ›w
tyz ð2Þ ¼G ¼ 2Gðu3x 2 u2x Þ
It is assumed in our model: ›y
† Material is homogeneous and linearly elastic in each te is the thickness of the epoxy interface and u is the rotation
layer; degree of freedom of the shell node.
† Displacement in z-direction is small and interfacial layer It can be seen from Eqs. (2) and (2a) that the first part of
is under anti-plane shear and shear stress throughout the the shear stress relates to relative displacement of the
plate thickness is uniform. adherends, and the second part relates to relative rotation of
the adherends. From Fig. 2, the displacement relationship
For the FRP sandwich bus panel analyzed here, high between the top skin and the corresponding interfacial layer
transverse shear deformation can be expected, therefore the can be expressed as:
first-order shear deformation shell theory is adopted in ts tp
modeling the FRP skins and paper core [23]. The u2 ¼ u1 2 u ; u3 ¼ u4 þ u
2 1y 2 4y ð2bÞ
displacement field is given by t tp
v2 ¼ v1 þ s u1x ; v3 ¼ v4 2 u4x
uðx; y; zÞ ¼ u0 ðx; yÞ þ zcx ðx; yÞ ð1Þ 2 2
vðx; y; zÞ ¼ v0 ðx; yÞ 2 zcy ðx; yÞ Suppose u1x < u2x ; u3x < u4x and u1y < u2y ; u3y < u4y ; then
the following expressions can be obtained:
wðx; y; zÞ ¼ wðx; yÞ
G t tp
where u, v, w are displacement components in the x, y, z txz ¼ u1 2 s 2 te u1y 2 u4 þ þ te u4y
te 2 2
directions, respectively; cx and cy are rotations of the cross-
G ts tp
section about the x and y axis; and u0 and v0 are tyz ¼ v1 þ 2 te u1x 2 v4 2 þ te u4x
displacement components at the midplane of the plate. te 2 2
The interfaces between the skins and the core are assumed ð3Þ
54 H.-C. Wu et al. / Composites: Part B 34 (2003) 51–58
Fig. 5. Compression of the FRP bus panel showing interface delamination between the skin and the core.
From Eqs. (4), (6), (9) and (10), it can be seen that the by a MTS 810 system under a displacement control. The
temporary interfacial nodes, 2, 3, 5 and 6 are eliminated coupon specimens were 75 £ 75 mm2 (3 £ 3 in.2). A typical
(Fig. 2). The shear forces, tensile or compressive forces and failure mode from the compression tests involved delamina-
moments in the interface can be expressed in terms of the tion between the FRP skins and paper core as shown in
degrees of freedom of the shell nodes in the skins and core. Fig. 5. In the test, the thicker skin was first debonded
This significantly increases the computation efficiency of outwards at a location about 1/4 to 1/3 distance from the
the current model. edge of the specimen when the compressive load reached its
maximum. The delamination propagated rapidly and
through the interface between the thicker skin and the
3. Simulation results and discussion paper core. Then the specimen was crashed. Multiple failure
modes, such as buckling and rupture of the samples might
Before launching a full-scale numerical simulation, the exist. The purpose of the FEM simulation is to find a
proposed model was used to simulate three point bending particular failure mode corresponding to the lowest external
tests with two extreme conditions: perfect rigid bonding load, i.e. true testing failure load. Herein both the integral
between layers and totally debonding between layers. The model based on the integral method (first-order theory)
three point bending test results [1] were used to verify [13,14] and the proposed layered model were employed first
the FEM predictions (Fig. 4). In this three bending test, the to simulate the overall compressive behavior. Then the
size of the samples was 73.66 £ 304.8 mm2 (2.9 £ 12 in.2). layered model was used to investigate the possible failure
Instron MTS 1.09 was used at a ramp rate of 5.08 mm/min locations at the interface. The numerical predictions of the
(0.2 in./min). The maximum external load was conveniently sample’s overall compressive behavior are shown in Fig. 6.
obtained by a linear static calculation process in the ABAQUS The global critical buckling status was determined by the
(point A in Fig. 4) by assuming a perfect bonding condition energy criterion. The buckling load is around 25 MPa
between the skins and the core. In the linear static (3.6 ksi). It can be seen that both models predict the overall
calculation, ABAQUS did not consider any geometrical and
compressive behavior of this FRP sandwich panel well.
material nonlinearities. After the maximum external load, a
sudden load drop was observed from the experiment. The
specimen still had some residual strength. Some local
delamination between the FRP skins and the paper core was
observed in the specimens at this point. However, no
detailed description of this observed delamination was
given in the report. Here we considered an extreme case in
the FEM simulation, i.e. assuming a complete debond
between the FRP skins and the paper core. The FEM
prediction was marked point B in Fig. 4. It can be found that
the experimental data approach point B at large mid-span
displacements, i.e. severe delamination that is close to a
complete delamination assumption used in the FEM
simulation.
The model then was employed to predict the compressive
behavior of the same sandwich sample and to locate
the initial failure position. The compression test was done Fig. 6. FEM simulations for compression of FRP bus panels.
56 H.-C. Wu et al. / Composites: Part B 34 (2003) 51–58
neglected the effect of not only Eq. (10) but also the second
part of the shear stress in Eqs. (2) and (2a). (The second part
relates to relative rotation of the adherends.) It can be found
that there is a large difference between the predictions by the
layered model and the layered model-1, while the predic-
tions by the layered model-1 and the layered model-2 are
very similar (within a 3% difference). This suggests that the
moment transfer in the interface plays a very important role
in the mechanical behavior of the FRP sandwich structures.
Though this epoxy interface is very thin (around 0.01 in.),
neglecting its moment transferability would lead to a large
discrepancy in the strength prediction. However, the
influence of the second part in the shear stress term
(Eq. (2)) is quite limited. Considering that it is very easy to
implement this second part of Eq. (2) in ABAQUS , even the
final improvement of the prediction is not significant, it will
Fig. 7. Compressive buckling loads given by experiment and FEM.
be beneficial to include this part. Since both this moment
transfer equation and the second part of the shear stress term
Fig.REL-P 7 compares the ultimate loads determined by the are related to the rotation of interface, it can be concluded
experiment and by the FEM models. In Fig. 7, the layered that relative rotation of the interfacial layer is important, at
model-1 neglected the effect of moment transfer in the least in the case investigated here. The ultimate buckling
interface, i.e. neglecting Eq. (10) and layered model-2 loads predicted by the integral model and the layered model
considering all normal, shear, and moment at the interface Jonson is the program manager; his valuable insights and
agree quite well with the experimental results. All predicted assistance are greatly appreciated. We are also indebted
compressive failures are dominated by buckling, not by to Mr John Resnik and Mr Ken Huser of ElDorado
static strength. National for their very useful discussions and for
Buckling modes simulated by the FEM models are providing the composite panels used in this study. The
shown in Fig. 8. All the three simulated modes by the authors want to thank you the reviewers for their useful
layered model, layered model-1 and integral model are very suggestions.
similar. However, only the layered model is able to predict
precisely the stress state at the interface. In the interface, the
maximum tensile stress occurs at the edge of the specimen
and the maximum shear stress occurs at about 1/4 of the References
specimen from the top edge, both on the thicker skin side.
Under compression, delamination between the skin and the [1] ElDorado National. Mechanical test composite bus panels. Exper-
imental Report; 1992.
paper core due to global buckling would initiate when the [2] Wu HC, Mu B, Warnemuende K. Fiberglass reinforced plastic
maximum shear stress exceeded the corresponding shear composite bus body structural integrity analysis. Final report to the
strength. Subsequently tensile failure in the interface would Michigan Department of Transportation, the Urban and Public
occur at the top boundary of the compressed sample (Fig. 5) Transportation Passenger Transportation Division and the Michigan
due to the maximum tensile stress exceeding the Transit Center for Excellence; 2001.
[3] Lekhnitskii SG. Theory of elasticity of an anisotropic elastic body.
corresponding tensile strength. This simulation agrees well
San Francisco: Holden-Day; 1963.
with the experimental observation (Fig. 5). [4] Reissner E. The effect of transverse shear deformation on the bending
of elastic plates. ASME J Appl Mech 1945;12:A69–A77.
[5] Mindlin RD. Influence of rotary inertia and shear deformation on
4. Conclusion flexural motions of isotropic elastic plates. ASME J Appl Mech 1951;
18:31–8.
[6] Librescu L, Hause T. Recent developments in the modeling and
It is difficult to analytically predict the performance of
behavior of advanced sandwich constructions: a survey. Compos
the FRP sandwich composites especially when local Struct 2000;48:1–17.
damage and failure develop. To lower analysis cost and to [7] Mallikarjuna, Kant T. A critical review and some results of recently
have an accurate prediction capability of FRP structural developed refined theories of fiber-reinforced laminated composites
performance, it is highly desired to use computer simu- and sandwiches. Compos Struct 1993;23:293– 312.
lations instead of conducting full-scale tests. For accurate [8] Mallikarjuna, Kant T. Free vibration of symmetrically laminated
plates using a high order theory with finite element technique. Int J
numerical simulation of composite panels, the FEM model Numer Mech Engng 1989;28:1875– 89.
should describe a true stress state in the structure. [9] Correia VMF, Soares CMM, Soares CAM. Design sensitivity analysis
The proposed layered FEM model assumes the skins and and optimal design of composite structures using higher order discrete
the middle core as Mindlin plates, and the interfacial epoxy models. Engng Optim 1997;29:85–111.
layer can transfer normal stresses, shear stresses and [10] Botello S, Oñate E, Canet JM. A layer-wise triangle for analysis of
laminated composite plates and shell. Comput Struct 1999;70:
moments. It has been demonstrated that the concept of
635–46.
this model is clear and the mathematical computation is [11] Reddy JN. Mechanics of laminated composite plates—theory and
simple. The variables in the interfacial layer can be analysis. Boca Raton: CRS Press; 1997.
completely eliminated. As a result, the computation [12] Reddy JN. A generalization of two-dimensional theories of
efficiency is greatly improved. laminated composite plate. Commun Appl Numer Meth 1987;3:
From the simulations, it can be concluded that the 173 –80.
[13] Jones RM. Mechanics of composite materials. New York: McGraw-
proposed layered model is able to not only predict the Hill; 1975.
overall mechanical behavior of the FRP sandwich struc- [14] Whitney J. Structure analysis of laminated anisotropic plates. Oxford:
tures, but also to analyze local failure issues, such as Pergamon Press; 1987.
location of failure initiation point and failure modes. The [15] Riccio A, Scaramuzzino F, Rerugini P. Embedded delamination
overall mechanical behavior predicted by the layered model growth in composite panels under compressive load. Composites, Part
B 2001;21:209–18.
is similar to that by the first-order integral model. The
[16] McCarthy MA, Wiggenraad JFM. Numerical investigation of a crash
moment transferability of the interface is found to be very test of a composite helicopter subfloor structure. Compos Struct 2001;
important in the numerical simulation process. 51:345–59.
[17] Choi Y, Chang FK. A model for predicting damage in graphite/epoxy
laminate composites resulting from low velocity impact. J Compos
Acknowledgements Mater 1992;26:2134 –67.
[18] Wang JT, Raju IS, Davila CG, Sleight DW. Computation of strain
energy release rate for skin-stiffener debonds modeled with plate
This project was funded by the Urban and Public elements. 34th AIAA/ASME/ASCE/AHS/ASC Structures, Structural
Transportation Passenger Transportation Division of Dynamics and Materials Conference, La Jolla, CA, USA; 1992.
the Michigan Department of Transportation. Mr Jerome p. 1680–92.
58 H.-C. Wu et al. / Composites: Part B 34 (2003) 51–58
[19] Sankar BV, Rao VS. A plate finite element for modeling delamina- [22] Arendt C, Sun CT. Bending effects of unsymmetric adhesively
tion. J Reinforced Plastics Compos 1993;12:227–36. bonded composite repairs on cracked aluminum panels. FAA/
[20] Rose LRF. A cracked plate repaired by bonded reinforcements. Int J NASA International Symposium, Hampton, VA, Part I; 1994.
Fract 1982;18:135–44. p. 33–48.
[21] Jones R, Callinan RJ. Finite element analysis of patched cracks. [23] Zienkiewicz OC. Finite element method. New York: McGraw-Hill;
J Struct Mech 1979;7:107–30. 1989.