Applied Computer Science, vol. 18, no. 2, pp.

43–55
doi: 10.35784/acs-2022-12

Submitted: 2022-03-28 | Revised: 2022-05-16 | Accepted: 2022-05-21

Keywords: finite element method (FEM), post-buckling, progressive failure analysis (PFA),
delamination, cohesive zone model (CZM)

Blazej CZAJKA [0000-0002-0870-5334]*, Patryk ROZYLO [0000-0003-1997-3235]*,

Hubert DEBSKI [0000-0002-1916-8896]*

STABILITY AND FAILURE OF THIN-WALLED

COMPOSITE STRUCTURES WITH A SQUARE
CROSS-SECTION

Abstract
This paper is devoted to the analysis of the stability and load-carrying capacity of thin-
walled composite profiles in compression. The specimens reflect elements made of
carbon fibre reinforced laminate (CFRP). Thin-walled columns with a square cross-
section were made from 4 layers of composite in 3 different combinations of layer
arrangements. Advanced numerical analyses have been carried out. In the first stage
of the study, a buckling analysis of the structure was performed. In further numerical
simulations, two advanced models were used simultaneously: the Progressive Failure
Analysis (PFA) and the Cohesive Zone Model (CZM). The results showed significant
differences between the critical load values for each layer configuration. The forms of
buckling and the areas of damage initiation and evolution were also dependent on the
applied layup.

1. INTRODUCTION

Thin-walled structures have been a very common type of load-bearing structure in many
sectors of industry for many years. They have a major role to play in the construction
of aircraft, vehicles and modern buildings. Thin-walled structural elements with different
cross-sections, both open and closed (stringers, frames, profiles), are used to transfer loads.
Due to the favourable weight/strength ratio, especially for aeronautical structures, more and
more components are being manufactured from composite materials in favour of traditional
engineering materials. One of the widely used composites is the continuous carbon fibre
reinforced polymer (CFRP) laminate. It is also characterised by chemical and corrosion
resistance and high fatigue strength (Chung, 1994). It is used in aviation to manufacture
many responsible parts such as fuselage and landing gear components, airframes, and
helicopter blades (Freeman, 1993). The most commonly used method of manufacturing
critical composite parts is the autoclave technique (Campbell, 2004, 2006), which ensures
high strength of these parts, repeatability of the manufacturing process, as well as low

*Lublin University of Technology, Faculty of Mechanical Engineering, Department of Machine Design and
Mechatronics, Lublin, Poland, [email protected], [email protected], [email protected]

43
internal porosity of the composite material. Thin-walled load-bearing elements during
exploitation should be used in the stable range, which is the subject of many scientific papers
(Berardi, Perrella, Feo & Cricri, 2017; Fascetti, Feo, Nistici & Penna, 2016; Kubiak,
Kolakowski, Swinarski, Urbaniak & Gliszczynski, 2016).
Buckling of a structural member may occur due to, for example, compressive loading
(Debski, Teter, Kubiak & Samborski, 2016). This results in operation of the structure in the
coverage range and accelerated failure. During stable working, thin-walled composite
structures show the possibility of continuing to carry the axial compressive load, even after
the buckling phenomenon has occurred (Koiter, 1963; Kubiak, Kolakowski, Swinarski,
Urbaniak & Gliszczynski, 2016; Singer, Arbocz & Weller, 2000). As shown by previous
studies (Debski, Rozylo, Gliszczynski & Kubiak, 2019; Falkowicz, Mazurek, Rozylo,
Wysmulski & Smagowski, 2016; Wysmulski, Debski, Rozylo & Falkowicz, 2016), they
have a large reserve of load-carrying capacity as long as the buckling is elastic in nature and
the post-buckling equilibrium path remains stable. Therefore, the study of stability and the
load-carrying capacity of thin-walled composite structures, requires analysis both in the state
before and after the occurrence of loss of stability (Paszkiewicz & Kubiak, 2015; Rozylo,
Debski, Wysmulski & Falkowicz, 2018). The analysis of composite structures determines
the study of the full load range up to failure (Abrate, 1998) and the description of the
phenomena of initiation and evolution of failure (Liu, Gu, Peng & Zheng, 2015). Many
papers also demonstrate the ability of the described structures to carry loads after failure of
the first laminate layer, as well as significant differences in limit load values, depending on
the configuration of the composite layers (Debski, Teter, Kubiak & Samborski, 2016).
Composite elements are characterised by a more complex failure mechanism than traditional
materials (e.g. metals). They may fail as a result of tension or compression of the fibres,
tension or compression of the matrix, and shear between layers (Camanho & Matthews,
1999; Lapczyk & Hurtado, 2007). This requires that experiments be conducted using several
measurement methods simultaneously and that advanced failure models be applied during
numerical simulations. Numerical analyses allowing observation of failure initiation and
evolution phenomena are usually carried out using the progressive failure analysis (PFA)
model (Camanho & Matthews, 1999; Lapczyk & Hurtado, 2007), whereas the cohesion zone
model (CZM) is usually applied to describe the delamination phenomenon (Liu, Gu, Peng
& Zheng, 2015).
In most of the published works, the authors focus only on open cross-sections. The
current study is based on a comparison of thin-walled columns with a square cross-section
(closed cross-section) prepared in 3 different laminate layer configurations. Furthermore, the
numerical simulations have been carried out in a more detailed way than in previously
published works. Both PFA and CZM numerical models were used simultaneously.
Furthermore, the cohesion zones were used globally. Previous work uses cohesive zones
only at delamination locations on real specimens during experimental tests. This definitely
simplifies the model and may lead to the omission of delamination phenomena in other areas
of the elements. The present work is devoted exclusively to numerical analysis as a preliminary
to further research on closed sections. In order to validate the simulation results obtained,
experimental tests are planned to be carried out in the next stages.

44
2. METHOD

2.1. Object of research

In this study, columns made of a closed square profile with a height equal to 250 mm
were examined. The internal side length of the square was 40 mm. For a better representation
of the real specimens, a corner rounding was used. The composite was made of 4 layers
of laminate, each 0.1 mm thick. The exact dimensions of the model are shown in Figure 1.
In this paper, 3 different configurations of the arrangement of the laminate layers were
analysed:
 P1 – [0/90/90/0],
 P2 – [90/0/0/90],
 P3 – [45/0/0/45].

Fig. 1. Test specimen: a) specimen height, b) cross-sectional parameters

The specimens were made of carbon-epoxy laminate (CFRP). The used mechanical and
strength properties are similar to those found in the literature (Rozylo, Debski, Wysmulski
& Falkowicz, 2018) and shown in Table 1.

45
 Tab. 1. Material properties of CFRP

Symbol Property Value Unit

E1 Young's modulus (along fibres) 130000
E2 Young's modulus (perpendicular to fibres) 6500 MPa
G12 Kirchhoff modulus 5000
ν1 Poisson’s coefficient 0.3 –
FT1 Tensile Strength (along fibres) 2000
FC1 Compressive Strength (along fibres) 100
FT2 Tensile Strength (perpendicular to fibres) 1500 MPa
FC2 Compressive Strength (perpendicular to fibres) 50
F12 Shear Strength 100

2.2. Numerical analysis

The simulations were carried out based on the finite element method (FEM) using the
Abaqus software. The first part of the research consisted in the analysis of the buckling of
the structure. The Progressive Failure Analysis (PFA) and Cohesive Zone Model (CZM)
were used to analyse post-buckling, loss of load-carrying capacity and delamination.
The first stage of the research was to solve the eigenproblem based on the minimum
potential energy criterion. This allowed to obtain the buckling form and the critical load
value. The critical load value was defined using the equation:

(𝐾0𝑁𝑀 + 𝜆𝑖 𝐾∆𝑁𝑀 )𝑣𝑖𝑀 = 0 (1)

where: K0NM represents the stiffness matrix (corresponding to the base state), which includes
the effects of preloads (PN), KΔNM denotes the differential initial stress as well as load
stiffness matrix due to the incremental loading pattern (QN), λi represent the eigenvalues, viM
constitute the buckling mode shapes – eigenvectors, M and N refer to degrees of freedom M
and N (of the whole model), i refers to the ith mode of buckling. The critical buckling loads
are PN + λi QN. Furthermore, viM constitute normalized vectors (and do not represent real
magnitudes of deformation at critical load).
Then, simulations of non-linear loss of stability and load-carrying capacity of the structure
were performed. The study also included the phenomenon of delamination occurring
between the composite layers. In order to better represent the real phenomena, imperfections
of the model from the form of buckling obtained in the first stage of research were used.
The preparation of the computational models involved making each laminate layer
separately and then adding contact relations between them with the properties of cohesive
layers. This made it possible to observe the delamination phenomenon in the whole model.
During the test, thin-walled composite columns were subjected to compressive loads over
the full range up to failure. Two non-deformable plates were added to the ends of each
column in order to best represent real conditions. Contact relationships were established
between the composite columns and the plates in the normal and tangential directions. One
of the plates (bottom plate) was fully fixed by removing all degrees of freedom. The upper
plate was fixed in all directions except the direction along the height of the test specimen
(Z axis). A compressive force was applied to this plate (fig. 2).

46
 Fig. 2. Test specimen: a) constraints and loading, b) discrete model

The prepared models were discretised. Non-deformable, four-node elements with a linear
shape function (R3D4) were used for the support plates. SC8R elements (eight-node shell
elements with linear shape function) were used for the laminate layers. The numerical model
for each layer layout consisted of 200 R3D4 elements, 8000 SC8R elements and 16562
nodes. A view of the discrete model is shown in Figure 2.

3. RESULTS

3.1. Buckling analysis

The analysis of the buckling of the structure showed significant differences between the
studied arrangements of composite layers. The differences can be seen both in the values of
critical loads causing buckling of the composite columns and in the form of buckling of the
structure.
As shown in the graph (Fig. 3.), the obtained buckling load for the K3 configuration was
the highest and amounted to about 804 N. The loss of stability for K1 and K2 occurred at
similar load values of 703 N and 694 N, respectively. The load for the K3 configuration was
therefore 14% and 16% higher respectively. This suggests that the use of layers at 45 degrees
to the compressive force reinforces the structures and increases the buckling strength.

47
 Fig. 3. Critical load values

The buckling forms, shown in Figure 4, are completely different depending on the layup.
For configuration K1, 4 half-waves occur on each wall. The appearance of the half-waves is
symmetrical with respect to the planes passing through the centres of the opposite edges of
the cross-section. For the K2 configuration there is also symmetry, but the number of
occurrence of half-waves has significantly increased to 9 on each wall. A completely different
form of buckling is visible on the specimen in the K3 configuration. There are 5 half-waves
on two of the walls and 7 half-waves on the other two walls. There is no symmetry as seen
in the previous specimens. The half-waves are arranged at an angle of 45 degrees, according
to the arrangement of fibres in the outer and inner layer of the composite.

Fig. 4. Buckling view for layer arrangements: a) K1, b) K2, c) K3

3.2. Damage initiation

Damage initiation was analysed using two criteria: Tsai-Wu and Hashin. The results
obtained with these criteria did not show much difference from each other. The biggest
deviation was recorded for the K2 configuration, but it did not exceed 3.5%. Achieved values

48
of damage initiation forces and their differences between the used criteria are summarised
in Table 2. In K1 and K2 configurations, damage initiation occurred at similar loads (for
selected criteria). Opposite to buckling, configuration K3 was found to be the least strong at
the time of damage initiation. The damage initiation load for K3 compared to the other
configurations was about 16% lower according to the Tsai-Wu criterion and about 18%
lower according to the Hashin criterion. It is important to emphasise that the failure initiation
loads are as much as 5 to 7 times higher than the critical loads causing buckling of the
structure. According to the Hashin criterion, the damage initiation occurred as a result of
matrix tension – in the case of K1 configuration, and matrix compression in the case of K2
and K3 configurations.
Tab. 2. Load values for damage initiation (Tsai-wu and Hashin criteria)

Composite Damage initiation forces [N] Difference

layup Tsai-Wu criterion Hashin criterion [N] [%]
5129
K1 4995 134 2.68
(matrix tension)
5128
K2 4905 168 3.43
(matrix compression)
4218
K3 4218 0 0
(matrix compression)

The distributions of damage initiation obtained with the Tsai-Wu criterion are shown in
Figure 5. The analysis by using this criterion shows that initiation for each of the systems
occurs on the inner layer, and only for the K1 configuration also on the outer layer.

Fig. 5. Distribution maps of damage initiation on layer 1 – Tsai-Wu criterion: a) K1, b) K2, c) K3

For the Hashin criterion, for all configurations the damage initiation is visible on the inner
layer. A view of the damage initiation distribution according to this criterion is shown in
Figure 6.

49
 Fig. 6. Distribution maps of damage initiation on layer 1 – Hashin criterion: a) K1, b) K2, c) K3

In general, the areas exposed to damage initiation are the corners of the tested columns
and the buckling half-wave hinge locations at the edges of the element. The areas of damage
initiation are identical except the layup configuration K1. For this layup, according to the
Tsai-Wu criterion, the corners are the locations of damage initiation. In the case of analysis
using the Hashin criterion, damage initiation occurs at the edges of the element at the half-
wave inflection point. Low values for this criterion are observed in the corners.

3.3. Damage evolution

The application of progressive failure analysis (PFA) has allowed the study of the areas
of occurrence, the failure mechanism and the loads at the time of damage evolution. The
damage evolution occurred in 1 layer of the composite for each of the investigated config-
urations. The areas were identical to those observed at failure initiation (Fig. 7).

Fig. 7. Areas of damage evolution in different layups: a) K1 – DAMAGEMT, b) K2 – DAMAGEMC,

c) K3 - DAMAGEMT

For the K1 and K3 layer configurations, the damage occurred by tension of the composite
matrix (DAMAGEMT), while for K2 it occurred by compression of the matrix
(DAMAGEMC). During the tests carried out for all configurations, matrix damage occurred
in both tension and compression before the tested structure lost its load-carrying capacity.
For configurations K2 and K3 the fibre damage (DAMAGEFC) was also observed in
compression but after the loss of load-carrying capacity. The fibre failure due to tension
(DAMAGEFT) and, in the case of K1, due to compression were not reached during the tests.

50
The loads that caused the reach of the individual parameters are summarised in the table
below (Tab. 3).

Tab. 3. Load values at damage evolution

Composite Load values [N]

layup DAMAGEMC DAMAGEMT DAMAGEFC DAMAGEFT
K1 5602 5518 – –
K2 5557 5826 5116* –
K3 6471 5486 6321* –
* – evolution has taken place after the loss of load-carrying capacity

3.4. Delamination

Observation of the delamination phenomenon was possible by using the cohesive zone
model (CZM). Both the initiation (CSMAXSCRT) and evolution of delamination (CSDMG)
occurred in areas close to the initiation and evolution of failure. The main areas exposed to
delamination were the corners of the elements and the half-wave inflection points at their
edges. The locations of occurrence between layers 3 and 4 for the K2 configuration are
shown in Figure 8.

Fig. 8. Locations of delamination between layers 3 and 4 - K2 configuration

The applied numerical model allowed for excellent visualisation of the delamination
phenomenon for the tested samples. It is especially visible in the corners of the investigated
composite columns. The following figure (Fig. 9) shows a view of delamination for the K2
configuration.

51
 Fig. 9. Delamination view for K2 configuration

For each of the tested configurations, delamination initiation between at least two layers
occurred before damage initiation. Delamination initiation was achieved for all specimens
between each layer during the tests. Each of them occurred before the loss of load-carrying
capacity of the tested structures. The delamination evolution was not achieved only between
layers 2 and 3 in the K1 configuration. For each configuration, at least between two of the
layers delamination occurred before the loss of load-carrying capacity of the structure. The
occurrence of the phenomenon in the load-carrying area of composite structures confirms
the importance of its analysis. This may allow better prediction of the behaviour of real
structures made of laminates.

3.5. Equilibrium paths and loss of load-carrying capacity

By carrying out the study over the full load range, the moment of loss of bearing capacity
was obtained. In order to analyse this phenomenon, equilibrium paths were determined.
They allow for a convenient analysis of the sequence of occurrence of particular phenomena
during compression of the tested specimens. The equilibrium path for the K3 configuration
is shown in Figure 10.
In the presented configuration, delamination initiation occurred at forces of 3083 N and
3257 N. Subsequently, damage initiation occurred at a force of 4218 N (according to
Tsai_Wu and Hashin's criterion). Although this point was exceeded, the structure still carried
the load, which increased continuously. Successively, the matrix damage evolution took
place by tension, compression and then the delamination between layers 1 and 2, and layers
3 and 4 occurred. Loss of load-carrying capacity occurred only when a load of 6642 N was
reached. In the same calculation step, delamination evolved between layers 2 and 3.

52
 Fig. 10. Equilibrium path – K3 layup

The load causing loss of load-carrying capacity for all the tested composites was
significantly higher than the load causing damage initiation. For the K3 configuration, this
difference was as high as 2424 N, which is about 57% of the value of the damage initiation
load. For K1 and K2 configurations the differences were smaller but also significant. They
were 622 N (12%) and 922 N (19%), respectively. The K3 configuration was characterized
by the highest load-carrying capacity (Fig. 11).

Fig. 11. Load-carrying capacity for different configurations

53
4. CONCLUSIONS

The numerical analyses carried out allowed comparison of composite thin-walled

columns with square cross-sections and various arrangements of laminate layers. They also
provided a preliminary study of the stability and load-carrying capacity of compression
composite profiles with closed cross-sections. The conducted research results in the following
conclusions:
 the buckling analysis of thin-walled structures with square cross-section is possible
by using FEM and solving the eigenproblem,
 applying the PFA model allows to thoroughly analyse the post-buckling behaviour
of composite columns with square cross-section,
 the Cohesive Zone Model (CZM) is a useful method to analyse the phenomenon of
delamination in composite structures with closed cross-section,
 the arrangement of composite layers has a significant influence on buckling,
initiation and evolution of damage, delamination, as well as the load-carrying
capacity of composite structures,
 the sample with layers arranged at the angle of 45 degrees was characterised by the
lowest stability of the structure, but at the same time the highest load-carrying
capacity,
 for all tested configurations the phenomenon of delamination occurred before the
loss of load-carrying capacity, which confirms the importance of its investigation,
 each of the tested specimens lost its load-carrying capacity at a load significantly
higher than the load occurring at the damage initiation.
Future experimental research using a universal testing machine is planned as part of the
project No. 2021/41/B/ST8/00148 (National Science Centre, Poland). The study of the
composite structure damage will be performed based on acoustic emission method and
microscopic analysis. A digital microscope with a mobile head will be used to record the
forms of failure during compression of the specimens.

Funding
The publication was financed by Grant – Lublin University of Technology Doctoral
School – Błażej Czajka.

Acknowledgments
The research has been conducted under project No. 2021/41/B/ST8/00148 (National
Science Centre, Poland)

REFERENCES

Abrate, S. (1998). Impact on Composite Structures. Cambridge University Press.

Berardi, V. P., Perrella, M., Feo, L., & Cricri, G. (2017). Creep behavior of GFRP laminates and their phases:
Experimental investigation and analytical modelling. Composites Part B: Engineering, 122, 136–144.
https://doi.org/10.1016/j.compositesb.2017.04.015

54
Camanho, P. P., & Matthews, F. L. (1999). A Progressive Damage Model for Mechanically Fastened Joints in
Composite Laminates. Journal of Composite Materials, 33(24), 2248–2280.
https://doi.org/10.1177%2F002199839903302402
Campbell, F. C. (2004). Manufacturing Processes for Advanced Composites. Elsevier B.V.
Campbell, F. C. (2006). Manufacturing Technology for Aerospace Structural Materials. Elsevier Ltd.
Chung, D. D. L. (1994). Carbon Fiber Composites. Elsevier Inc.
Debski, H., Rozylo, P., Gliszczynski, A., & Kubiak, T. (2019). Numerical models for buckling, postbuckling and
failure analysis of pre-damaged thin-walled composite struts subjected to uniform compression. Thin-
Walled Structures, 139, 53–65. https://doi.org/10.1016/j.tws.2019.02.030
Debski, H., Teter, A., Kubiak, T., & Samborski, S. (2016). Local buckling, post-buckling and collapse of thin-
walled channel section composite columns subjected to quasi-static compression. Composite Structures,
136, 593–601. https://doi.org/10.1016/j.compstruct.2015.11.008
Falkowicz, K., Mazurek, P., Rozylo, P., Wysmulski, P., & Smagowski, W. (2016). Experimental and numerical
analysis of the compression thin-walled composite plate. Advances in Science and Technology Research
Journal, 10(31), 177–184. https://doi.org/10.12913/22998624/64063
Fascetti, A., Feo, L., Nisticò, N., & Penna, R. (2016). Web-flange behavior of pultruded GFRP I-beams: A lattice
model for the interpretation of experimental results. Composites Part B: Engineering, 100, 257–269.
https://doi.org/10.1016/j.compositesb.2016.06.058
Freeman, W. T. (1993). The use of composites in aircraft primary structure. Composites Engineering, 3(7–8),
767–775. https://doi.org/10.1016/0961-9526(93)90095-2
Koiter, W. (1963). Elastic Stability and Post Buckling Behavior in Nonlinear Problems. University of Wisconsin
Press.
Kubiak, T., Kolakowski, Z., Swinarski, J., Urbaniak, M., & Gliszczynski, A. (2016). Local buckling and post-
buckling of composite channel-section beams – Numerical and experimental investigations. Composites
Part B: Engineering, 91, 176–188. https://doi.org/10.1016/j.compositesb.2016.01.053
Lapczyk, I., & Hurtado, J. A. (2007). Progressive damage modeling in fiber-reinforced materials. Composites Part A:
Applied Science and Manufacturing, 38(11), 2333–2341. https://doi.org/10.1016/j.compositesa.2007.01.017
Liu, P. F., Gu, Z. P., Peng, X. Q., & Zheng, J. Y. (2015). Finite element analysis of the influence of cohesive law
parameters on the multiple delamination behaviors of composites under compression. Composite
Structures, 131, 975–986. https://doi.org/10.1016/j.compstruct.2015.06.058
Paszkiewicz, M., & Kubiak, T. (2015). Selected problems concerning determination of the buckling load of channel
section beams and columns. Thin-Walled Structures, 93, 112–121. https://doi.org/10.1016/j.tws.2015.03.009
Rozylo, P., Debski, H., Wysmulski, P., & Falkowicz, K. (2018). Numerical and experimental failure analysis of
thin-walled composite columns with a top-hat cross section under axial compression. Composite
Structures, 204, 207–216. https://doi.org/10.1016/j.compstruct.2018.07.068
Singer, J., Arbocz, J., & Weller, T. (1998). Buckling Experiments: Experimental Methods in Buckling of Thin-
Walled Structures, Volume 1: Basic Concepts, Columns, Beams and Plates. John Wiley & Sons Inc.
Wysmulski, P., Debski, H., Rozylo, P., & Falkowicz, K. (2016). A study of stability and post-critical behaviour
of thin-walled composite profiles under compression. Eksploatacja i Niezawodnosc-Maintenance and
Reliability, 18(4), 632–637. http://dx.doi.org/10.17531/ein.2016.4.19

55