Thin-Walled Structures 122 (2018) 52–63

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

Progressive failure analysis of thin-walled Fibre Metal Laminate columns T

subjected to axial compression
⁎
D. Banat , R.J. Mania
Department of Strength of Materials, Lodz University of Technology, Stefanowskiego 1/15, 90-924 Lodz, Poland

A R T I C L E I N F O A B S T R A C T

Keywords: The subject of the paper is the progressive failure analysis (PFA) of thin-walled Z-shape cross section members
Composite FML subjected to axial compression. This study concerns angle-ply multi-layered Fibre Metal Laminates (FMLs) which
Finite element analysis (FEA) consist of alternating thin layers of aluminium and glass fibre-reinforced unidirectional prepreg. Laboratory
Failure modes damage tests were performed by the static testing unit that provided displacement control loading. Experimental
Damage mechanics
results were compared with FEA wherein based on the profile's nonlinear stability investigations the failure
Progressive failure analysis (PFA)
analysis was performed. Nonlinear FE simulation combined with available progressive failure mechanics al-
Failure criteria
lowed to predict the initiation and propagation of the multi-failure modes within composite material. Hashin
failure criterion was used to monitor the initiation of damage, whereas material degradation method (MPDG)
was applied in FRP layer to define the damage evolution law. Damage variables were specified according to FEM
modelling procedures in order to control material stiffness reduction after damage initiation. For aluminium
layers the J2 plasticity model was employed. Progressive failure assessment by FEM allowed to estimate the post-
buckling equilibrium paths and damage modes with particular regions of laminate's fracture that were found to
be in a good agreement with experimental evidences.

1. Introduction
Composite structures have been extensively used in recent years
with important developments in the aviation, automotive and wind
energy industries. A wide group of composites is represented by Fibre
Metal Laminates (FMLs) which are hybrid composites constructed by
binding fibre-reinforced laminates with metallic layers [1]. Fur-
thermore, particularly in aviation industry most of the FMLs appli-
cations are based on unidirectional glass fibre-reinforced prepregs
combined with aluminium alloy sheets (Glare type) [2]. Ply combi-
nations such as this provide high bearing strength, large impact re-
sistance and improved damage tolerance [3]. Fibre-reinforced ma-
terial guarantees also improved strength and stiffness, particularly
when compared to other structural materials on a unit weight basis.
What is more important, depending on fibre alignment, FRP can be
designed to be stiffer in a specific direction [4]. This makes them the
material of choice for multiple applications. Such achievements
within the FML concept have inspired the research community over
the years to investigate its mechanical behaviour under specific
loading conditions [5,6].
Nevertheless, FMLs still need to meet the stringent requirements of
the aerospace applications in components such as fuselage sections or
cargo interiors [3]. Apart from high damage tolerance and fatigue re-
sistance, lowest possible weight of the structure needs to be maintained.
Therefore, one can observe the accelerating adoption of the thin-walled
Glare structures within the aviation industry [7]. However, study
aiming to reduce the thickness of the specific wall's section gives rise to
the stability considerations of the thin-walled structures [8,9]. Parti-
cularly for relatively slender and thin-walled components, the stability
phenomenon and structural optimisation becomes the interest of to-
day's investigations [10]. For that reason, numerous buckling problems
with specific stability constraints were posed in order to determine
solutions for critical load and buckling mode shape of the compressed
members [11,12].
Another stage of stability considerations includes the post-buckling
analysis to observe the structure's behaviour in full load range [13]. It
requires the implementation of semi-analytical and numerical methods
to verify the analytical solutions [14,15]. Based on the stability solu-
tions, the failure analysis can be applied to investigate the damage in-
itiation in the post-buckling range. In FMLs, aluminium constituent is
claimed to dominate the material response and the failure modes vary
with ductility [16]. However, due to the presence of FRP, the stacking
sequence and particular fibre alignment greatly influence the damage
pattern [4,17].

⁎
Corresponding author.
E-mail address: dominikbanat@gmail.com (D. Banat).

http://dx.doi.org/10.1016/j.tws.2017.09.034
Received 18 April 2017; Received in revised form 26 September 2017; Accepted 29 September 2017
Available online 13 October 2017
0263-8231/ © 2017 Elsevier Ltd. All rights reserved.
D. Banat, R.J. Mania Thin-Walled Structures 122 (2018) 52–63

For failure analysis of angle-ply multi-layered FML columns the
experimental tests and numerical simulations need to be performed
simultaneously. Particularly for composite laminates, advanced con-
cepts and developments in failure and damage analysis need to be
considered [18]. Numerical computation requires FEM implementation
wherein the earliest and most simplified modelling technique to address
material failure is the First Ply Failure (FPF) analysis. Such preliminary
analysis to FML study is given in the paper [19], where different failure
criteria are implemented to assess either matrix or fibre failure in the
composite material layers. Further study requires material degradation
model by means of ply discount method [20] or progressive damage
analysis (PDA) [21,22]. The latter includes the failure initiation and
material degradation algorithm that are defined in FEM software by
specific damage evolution law.
Separate failures modes of the material constituents of FMLs are
already identified [23] but their specific interaction has not been
proven yet [16]. In the literature one can find various examples of the
PFA application to the failure analysis of the composite structures
[24,25]. However, according to the World Wide Failure Exercise
(WWFE), to date there is no particularly satisfying method in FEM to
model the progressive damage analysis in order to assess the propa-
gation of the multi-failure modes in the composite material [26]. There
are also relatively few papers devoted to failure analysis of thin-walled
laminated profiles in the post-buckling state where local buckling ef-
fects are considered [14]. Hence, there is a need for fundamental un-
derstanding of the various failure mechanics and its impact on the
structure stability. The purpose of this study is to reveal the initial FE
results of progressive failure analysis for FML Z-section columns sub-
jected to compression.
2. Progressive failure analysis
A large variety of failure criteria and degradation models have
been developed over the years with different applications depending
on loading conditions and material properties [23,26]. For ductile,
isotropic material the Huber-Mises-Hencky criterion (also referred as
J2 plasticity condition) is widely applied, regarded as most reliable
when dealing with ductile materials. The issue of failure criteria is
more ambiguous for fibre-reinforced layered materials of orthotropic
and anisotropic nature. For that reason, Classical Laminate Plate
(CLP) theory is generally applied to determine the properties of the
entire composite together with its stress and strain distributions when
based on the stress tensor elements in plain stress case. Such assessment
allows one to predict the failure initiation within the aluminium layers
in the area where equivalent stress reaches the yield limit.
2 2 2
σEQV = σ11 − σ11 σ22 + σ22 + 3σ12 (1)
For fibre-reinforced laminate various failure criteria can be con-
sidered to indicate failure initiation. These available in many FEM
codes and commonly used in research for orthotropic plies analysis
are as follows: Tsai-Wu (1971), Hashin (1980), and Alfred Puck
(1996) criteria. Each of them is applied to predict material failure
and the relative differences result from the way in which stress and
strength components participation is defined in the failure function.
In the case of Alfred Puck criterion, additional inclination para-
meters need to be accepted as constants according to specific
guidelines [29]. Note, that Hashin and Puck criteria consider various
failure mechanisms by developing four different damage initiation
modes: fibre tension (rupture), fibre compression (kinkling), matrix
tension (cracking), and matrix compression (crushing). Hence, these
criteria allow one to track the matrix and fibre failure that can occur
either separately or sequentially. Certainly, several researchers have
also proposed modifications to Hashin criterion in order to improve
its predictive capabilities [30]. However, some of them can be simple
reduced to Hashin criterion providing certain assumptions. There are
works in which the shear stress component in the failure function is
associated with the additional weight factor that is usually denoted
as α. That provides different failure factor by solely increasing or
decreasing the shear stress effect onto the fibre failure [31]. Fur-
thermore, according to the World Wide Failure Exercise (WWFE)
most criteria were unable to capture some of the trends in the failure
envelopes of the experimental results [26]. However, various studies
claim that Hashin criterion provides sufficient predictive capabilities
and there is an increasing adoption of this criterion especially for the
purpose of FRP's failure prediction [17,28]. Therefore, for the pur-
pose of PFA model the authors decided to use Hashin criterion to
assess failure initiation in the composite layers.
Different failure modes developed by Hashin criterion can be de-
fined in terms of nominal Cauchy stress components and material
strength to recognize different failure modes (Eqs. 2–5)[32]:
• Fibre tension (σ 11≥0 )
2 2
σ σ
f ft = ⎛ 11 ⎞ + ⎛ 12 ⎞
⎜ ⎟

subjected to loading. Further GFRP strength analysis requires failure ⎝ Xt ⎠ ⎝ S ⎠ (2)

criteria application, wherein stress and strain components are com-
bined to allow a direct failure assessment. By means of specific failure
criteria the strength analysis of the laminate can be carried out at each • Compression (σ 11 < 0 )

ply separately. The weakest layer, which fails first according to σ

f fc = − 11
adopted criterion, determines the First Ply Failure (FPL) strength of Xc (3)
the laminate [1]. However, it is claimed that the load corresponding
to FPF does not always correspond to the loss of carrying capacity of
the entire composite. Most of the advanced structural materials such
• Matrix tension (σ 22≥0 )
2 2
as GFRP can still carry further loading after the failure initiation and σ σ
fmt = ⎛ 2 ⎞ + ⎛ 12 ⎞
⎜ ⎟

for that reason the progressive failure analysis with material de- ⎝ Yt ⎠ ⎝ S ⎠ (4)
gradation model is required to predict the laminate's ultimate strength
[27,28].
• Matrix compression (σ 22 < 0 )

2 2
2.1. Failure initiation σ σ Y 2 σ
f mc = ⎛ 22 ⎞ + ⎛ 12 ⎞ + ⎡ ⎛ c ⎞ −1⎤ 22
⎝ 2S ⎠ ⎝ S ⎠ ⎢
⎣ ⎝ 2S ⎠ ⎥
⎦ Yc (5)
The strength analysis of the laminate is carried out in order to
predict first failure initiation. The methodology includes failure criteria where f ft , f ft , f ft , f ft
- are the failure factors (FF); σij - the nominal stress
implementation that could assess the strength of each composite ply tensor components; Xt , Xc - the longitudinal tensile and compressive
separately. In case of Fibre Metal Laminate different criteria need to be strength limits; Yt , Yc - the transverse tensile and compressive strength
adopted for aluminium and GFRP layers. limits; S - the shear strength limit. Similarly, as for other criteria, failure
In the aluminium layer, the aforementioned Huber-Mises-Hencky factor indexes indicate whether a damage initiation criterion is satisfied
(HMH) criterion is used to determine the equivalent stress σEQV (Eq. (1)) or not, for value greater or lower than 1 respectively.

53
D. Banat, R.J. Mania
2.2. Material property degradation method (MPDG)
Once the failure initiation is observed the progressive failure ana-
lysis includes the introduction of damage evolution law. For that
reason, two damage evolution methods are available in FEA i.e. mate-
rial property degradation method (MPDG) and continuum damage
mechanics method (CDM). First assumes the material stiffness is in-
stantly reduced based on the explicitly specified damage variables. This
approach is still very often applied as the generalized form of con-
tinuum damage modelling in the fibre-reinforced composites proposed
by Camanho et al. [33]. The more detailed information of considered
degradation model and the procedures of selecting the damage vari-
ables one can find in the work of Barbero et al. [28]. In the second
method damage variables increase gradually based on the energy
amounts dissipated for the various damage modes [32]. However, CDM
also requires the definition of additional material properties such as
energies dissipated per unit area and viscous damping coefficients,
which are not available at this stage of FML columns investigation.
Hence, for the purpose of this study, MPDG method is used for pro-
gressive failure model. Thus, material stiffness is gradually reduced as
deformation continues to accumulate after the initiation criterion is
met. The constitutive relation for post-damage degraded material is
given in Eq. (6).
σ = [D]d ε (6)
Stress σ is the averaged value of effective stresses in the entire
material domain; [D]d represents damaged elasticity matrix and ε cor-
responds to total elastic strain. The damaged elastic matrix [D]d for a
transversely isotropic material with plane stress state can be expressed
as (Eq. (7)) [32]:
⎡ (1−df ) E1 (1−df )(1−dm) v21 E1 0 ⎤
1⎢
[D]d = (1−df )(1−dm) v12 E1 (1−dm) E2 0 ⎥
A⎢ ⎥ experimental and numerical results (below 10%) were achieved for
⎢
⎢ 0 0 A (1−ds ) G12 ⎥
⎥ (7)
where: A = 1−v12 v21 (1−df )(1−dm) .
Elastic properties of undamaged material (E1, E2, G12,v12,v21) need to
be adopted according to material characteristic. MPDG requires also
damage variables (df , dm ) which need to be defined for fibre and matrix
in tension and compression separately. Hence, four different modes
need to be determined: dft - tensile fibre stiffness reduction; dfc - com-
pressive fibre stiffness reduction; dmt - tensile matrix stiffness reduction;
dmc - compressive matrix stiffness reduction. Additionally, the shear
damage variable ds is expressed in terms of the remaining damage
variables (Eq. (8)) [28].
t damage state). Hence, for each engineering case it is recommended to
⎧ df for σ11≥0
df =
⎨ dfc for σ11 < 0
⎩ mental evidences.
d t for σ22≥0
dm = ⎧ m c
⎨
⎩ m for σ22 < 0
d
ds = 1 − (1 − dft )(1 − dfc )(1 − dmt)(1 − dmc) (8)
The stiffness reduction coefficients are mainly responsible for the
effect of material damage after the failure initiation in MPDG method.
Note, that for each damage variable the statement d = 0 indicates no
reduction in material stiffness in the affected mode after damage in-
itiation, and d = 1 complete stiffness loss in the affected mode.
2.3. Damage models
For fibre reinforced composites, it is not possible to determine un-
ambiguously the appropriate values of damage variables (df / m ) since
each engineering case needs to be approached separately. Therefore, in
the literature, one can find various studies that investigate the con-
tinuum damage mechanics theories in order to understand the physical
54
Thin-Walled Structures 122 (2018) 52–63
meaning of FRP damage [24,33]. First attempts included the definition
of statistical model which described the strength of disordered material
and the approximate load distribution after the initiation of local failure
in the composite. In tensile stress state, it took into consideration the
composite damage that can be studied in terms of fibre ineffective
length. This fact was recognized and proposed by Rosen, who suggested
that the fibre ineffective length (approx. 90%) is the value over which a
fibre recovers a large percentage of its load [34]. Therefore, damage
variable df ~0.9 was used to predict the composite ultimate strength in
its longitudinal direction.
Further studies included the interaction between strength statistical
theories and the continuum damage mechanics. For that reason, the
damage measure has been originally suggested by Kachanov [35] and
Rabotnov [36], who introduced the concept of damage variables that
must be within the interval [0,1]. Nevertheless, the initial argument
against the damage measure was judged by the experimental evidences
according to which the damage variables cannot reach the value 1
(df / m≠1). Providing linear damage law, it was claimed that the max-
imum value of damage variable cannot exceed 0.5 [37,38]. Ad-
ditionally, damage variables were found to give unsatisfying results for
flat and planar cracks [39] which required introducing the weighting
factor to define the influence of the cracks on the cross-sectional da-
mage in a function of their proximity [40,41]. It allowed one to analyse
the gradual failure accumulation and propagation in creep and fatigue.
The modern literature argues also that CDM solutions cannot be
treated unambiguously and there are specific failure criteria and la-
minate stacking sequences that affects the performance of FR composite
[42]. Therefore, it is recommended to verify progressive failure models
with different damage variables for specific loading conditions. The
paper [17] shows results of PFA applied in Ansys for thin-walled eight-
layer channel section of GFRP laminate subjected to uniaxial com-
pression. The paper argues that the acceptable relative differences of
values of the damage variables greater than or equal to 0.9 (df / m = 0.9 ).
Authors of that paper draw also the conclusion that an identification of
the numerical model allowed to establish that the highest level of
correlation of experimental and numerical results is achieved with the
declaration of all damage parameters equal to unity (df / m = 1). In an-
other work [28], authors present methodology for determination of
material damage parameters (df / m ) for the progressive damage analysis
in Abaqus software. Similarly, the effect of material stiffness degrada-
tion and composite damage is based on the explicitly specified damage
variables. It is claimed in this work that for considered failure modes,
the damage variables control the reduction of the stiffness coefficients,
and may assume values between zero (undamaged state) and one (fully
apply various damage variables and validate results against experi-
3. Subject of the study
Experimental and numerical investigations were carried out for
thin-walled Fibre Metal Laminate column of approx. 300 mm length
and 2 mm walls thickness. Profiles cross sections were of Z-shape type
with the following dimension: the width of the web − 80 mm, flange −
40 mm and the radius (R) of the web and the flange junction of ap-
proximately 1.75 mm (Fig. 1). The laminate included the alternating
plies of 2024 T3 aluminium alloy and fibre-reinforced composite. The
2024 T3 aluminium is a heat-treatable alloy with the copper as the
primary alloying element. For composite plies the specific glass-epoxy
unidirectional fibre-reinforced prepreg TVR 380 M12 26% R-glass was
applied with 60% of the fibre nominal volume fraction. The thickness of
a single aluminium and composite layer after curing equals to 0.3 mm
and 0.25 mm, respectively. The FML columns were made by autoclave
technique with specific curing parameters and manufacturing proce-
dures (Refs. [43,44].)
D. Banat, R.J. Mania Thin-Walled Structures 122 (2018) 52–63

Fig. 1. Overall dimensions of Z-shape section column (a) and the

3/2 stacking sequence (b).

Table 1
Mechanical characteristics of the GFRP composite.

E1 [GPa] E2 [GPa] G12 [GPa] G23 [GPa] ν12 Xt [MPa] Xc [MPa] Yt [MPa] Yc [MPa] S [MPa]

53.90 14.92 5.49 5.33 0.27 1534 800 75 500 58

The 7-layered composite with the 3/2 stacking were considered

(Fig. 1b), where ‘3’ indicates three aluminium layers and ‘2’ refers to
two embedded FRP layers between aluminium sheets. Due to the GFR
epoxy presence the structure is considered as thin-walled orthotropic
plate but of transversely isotropic symmetry in plane perpendicular to
fibres. Depending on fibre alignment 7 various stacking sequences of
FML were considered and compared [45]. However, for the purpose of
this study, failure assessment was carried out on Al/0/0/Al/0/0/Al lay-
up configuration.
Mechanical properties of aluminium and glass-epoxy prepreg con-
stituents are adopted from laboratory tests conducted according to
tensile and three-point bending test standards for composite materials
[46]. For the isotropic aluminium layer, the following mechanical
properties were defined: Young's modulus E = 77 GPa, Poisson's ratio ν
= 0.33 and the yield limit R0.2 = 359 MPa. For the FRP layer the
elastic properties and strength limits (see Table 1) were determined
according to specific rules [31] in the main orthotropic directions
connected with fibre orientation (principal 1-axis). Material properties
of aluminium and glass-epoxy prepreg constituents were declared by
specimens’ manufacturer and verified during own experiments
[43,44,46].
The considered profiles were subjected to damage compressive tests
that were carried out by means of the double column electromechanical Fig. 2. Experimental test rig of the damage compressive test.
strength testing machine by 200 kN Instron. Applied screw type testing
machine provided displacement control loading set to 1 mm/min. The
crosshead and the base of the testing column were equipped with the shell element is generally suggested. Particularly, for the purpose
special plate rigs designed to guarantee uniform axial compression. of this study, SHELL 181 from Ansys library was applied as this four-
Additionally, in order to prevent lateral displacement of the loaded node element with six degrees of freedom at each node is suitable for
edges, samples were positioned in flat bottom grooves (see Fig. 2). That nonlinear applications. SHELL 181 is dedicated to modelling layered
solution guaranteed better agreement with the simply supported applications of composite shells or sandwich constructions [32].
boundary conditions assumed in analytical and numerical approach SOLID type element was also investigated for the purpose of FML
[47]. The surface deformations were measured by ARAMIS 3D non- modelling, however it resulted in faulty buckling mode while com-
contact optical system which allowed to observe the specimen's failure paring to experimental evidences (for more details see Ref. [45].).
behaviour in full load range until fracture. That measuring system in- Shell element and section options in FE application allowed to define
cluded a sensor with two digital cameras of 1280 × 1024 px resolution independently individual plies of the laminate i.e. thickness, material
and a maximum full frame rate of 485 images/sec. It allowed to cover at properties and main axes of the orthotropic. Specific 3/2 FML stacking
least one flange and the whole web surface of the tested column. The of alternating thin layers of aluminium and glass fibre-reinforced
gathered data was used for comparison with FEM computations. prepreg was defined separately by means of the Lay-up method. Alu-
minium plies were modelled as the bilinear material with isotropic
4. FEM computations hardening whereas for GFRP the linear orthotropic model was applied
in the full range of loading. Particular fibre alignment in composite
The material response of the open cross-section columns subjected plies allowed to define various angle-ply laminate stacking i.e., AL/0/
to axial compression was simulated by commercial software package 90/AL/90/0/AL, AL/90/0/AL/0/90/AL, AL/45/0/AL/0/45/AL, AL/
ANSYS®. In literature one can find various approaches of modelling 0/45/AL/45/0/AL and AL/0/0/AL/0/0/AL. Subsequently, the geo-
fibre-reinforced Glare structures [6], however for thin-walled profiles metry of Z-shape section columns was modelled with dimensions

55
D. Banat, R.J. Mania
Fig. 3. Comparison of post-buckling equilibrium paths for Z-shape columns.
corresponding to real specimens (see Fig. 1a). The implementation of
the symmetry in the modelling process was also considered to reduce
the computation time. However, such simplification in the GFRP
model might also disrupt deformation modes, particularly in post-
buckling analysis [45]. Therefore, the full geometry of FML columns
was modelled in FEM. Subsequently, the appropriate displacements of
simply supported type were constrained on columns loaded edges in
order to provide boundary conditions consistent with the experi-
mental tests. The detailed analysis of FMLs boundary conditions so-
lutions is given in Ref [47].
Fig. 4. Aluminium layers judged by H-M-H criterion a) L1, b) L4, c) L7.
56
Thin-Walled Structures 122 (2018) 52–63
4.1. Stability considerations
The columns’ stability was assessed by linear buckling analysis
(LBA) which allowed one to determine buckling load and the corre-
sponding buckling mode shapes [48]. Then, the geometrically non-
linear analysis was required to account for the geometrically nonlinear
response of the structure and to find the solution in post-buckling
range until fracture. In the nonlinear analysis, load factor was nor-
malized with respect to first buckling load obtained in LBA. Such so-
lution provided gradual loading that corresponded to experimental
procedures where loading was controlled by a screw type testing
machine. Slender, thin-walled members are also claimed to be sensi-
tive to wall imperfections that could affect significantly the post-
buckling behaviour [49]. Thus, the first local mode shape obtained in
LBA was mapped on to the nodal mesh of the finite element model to
guarantee walls imperfections. The magnitude of the imperfection was
referred to 1% of the column wall thickness. Such analysis allowed
one to find the solution in post-buckling range until fracture - post-
buckling equilibrium path and load carrying capacity were de-
termined.
4.2. Failure assessment
Based on the laminate's stability considerations the failure analysis
was performed at each layer separately. Aluminium layers were judged
by Huber-Mises-Hencky (HMH) criterion which allowed to determine
equivalent stresses (EQVS) in the whole material domain. EQVS were
further compared with aluminium yield strength limit (R0.2 =
359 MPa). For composite layers, material strength limits (see Table 1)
and specific failure criteria were introduced. Criteria which are applied
for the purpose of this study allow to determine following failure factors
(FF):
• TWSR - inverse of Tsai-Wu strength ratio index,
D. Banat, R.J. Mania Thin-Walled Structures 122 (2018) 52–63

Fig. 5. Composite layers judged by failure criteria a) TWSR, b) HMAT, c) HFIB d) PMAT, e) PFIB and compared with damaged specimen (f).

• HFIB - Hashin fibre failure criterion, Failure assessment was performed for compressed FML members in
• HMAT - Hashin matrix failure criterion, full load range where fibre and matrix fractures were considered se-
• PFIB - Puck fibre failure criterion, parately. Failure factors were determined based on stress distributions
• PMAT - Puck inter-fibre (matrix) failure criterion. adopted from particular load-step of nonlinear loading process.

57
D. Banat, R.J. Mania
Fig. 6. Comparison of post-buckling equilibrium paths for flange (a) and web (b) de-
flections.
Mapping FF onto the profile geometry indicated regions greatly ex-
posed to damage which was further compared with experimental evi-
dences.
Second stage of failure analysis in GFRP layers included the pro-
gressive damage analysis (PDA) with a material degradation model. Its
FEM application required the definition of material damage initiation
criteria (DMGI) and the damage evolution law (DMGE). Damage in-
itiation defined the criterion type for determining the onset of material
damage. For that reason, Hashin criterion was used as a baseline to
establish failure initiation threshold in the progressive damage model.
The damage evolution law was performed by the material property
degradation method (MPDG). Hence, based on the damage variables
the stiffness was gradually reduced in the area where failure was in-
itiated. Impact of damage variables on the laminate's failure assessment
and post-buckling equilibrium paths was considered. The failure ana-
lysis allowed also to map the progressively damaged status onto the
profile geometry in order to expose the regions particularly susceptible
to failure.
5. Results and discussion
5.1. Failure monitoring in FML layers
Geometrically nonlinear analysis in FEA that included aluminium
real material characteristic, allowed one to estimate the post-buckling
equilibrium paths which was further compared with experimental tests
in order to validate the numerical model. Results of the comparison
between flange deflection for Z-shape profiles with various fibre
58
Thin-Walled Structures 122 (2018) 52–63
Fig. 7. ARAMIS experimental measurements compared with the most accurate PFA
models for flange (a) and web (b) deflections.
alignments are presented in Fig. 3. Different fibre orientations in FRP
layer are marked in the legend as follows:
• Z1 - AL/0/90/AL/90/0/AL
• Z2 - AL/90/0/AL/0/90/AL
• Z3 - AL/45/0/AL/0/45/AL
• Z4 - AL/0/45/AL/45/0/AL
• Z5 - AL/0/0/AL/0/0/AL
• ARAMIS measurement data
Post buckling equilibrium paths allowed to estimate relationship
between compressive load applied to the FML structure as the function
of wall deflection. Buckling load determined experimentally is slightly
lower than the one estimated with FEM application. That results from
the influence of flange imperfections in the tested specimen that are not
modelled directly in FEA. Such simplification for the laminates could
slightly affect local wall stiffness but provide only minor deviation
when compared to real specimens [50]. Furthermore, the curve ob-
tained with the digital correlation system ARAMIS was calculated as the
maximal deflection measured from the outer profile surface whereas in
FEM the nodes attaining the maximum displacement amplitudes were
related to the composite mid-plane. Nevertheless, both methods pre-
dicted buckling and ultimate loads for compared stacking sequences of
Z-shape columns in acceptable agreement where the mutual deviations
do not exceed 10%. Such results assessment and model verification for
buckling and post-buckling range approved the FEM model for further
analysis.
Subsequently, various failure tests were performed for Z-shape
section column with different fibre alignment in the composite layer.
However, in each case the damage initiation was observed in similar
manner and therefore results for only one specimen are presented and
compared with FE simulation. The corresponding numerical model al-
lowed to estimate the potential material damage in a buckling and post-
D. Banat, R.J. Mania Thin-Walled Structures 122 (2018) 52–63

Fig. 8. Damage status of progressive model PFA_1 in all composite layers a) L2, b) L3, c) L5, d) L6 compared with damaged specimen (e).

buckling state at each layer separately. The loading conditions and the damage initiation was tracked by failure criteria implementation
nominal stress state for specific substep were based on the previous and none of material degradation methods was applied. Herein, iso-
nonlinear buckling analysis, where load factor was normalized with tropic aluminium layers were judged by Huber-Mises-Hencky criterion
respect to first buckling load. At the initial stage of failure assessment, that allowed to estimate failure in the regions where equivalent stresses

59
D. Banat, R.J. Mania
exceed the material yield limit. For orthotropic composite layers Tsai-
Wu, Hashin and Puck criteria were applied to track failure initiation
based on failure factors assessment. The analysis was performed in the
post-buckling range for compressive load significantly higher than load
carrying capacity of the entire laminate. Such analysis that provided
maps of FF are presented for aluminium and composite layers in Figs. 4
and 5, respectively.
Results for the aluminium layers show regions greatly exposed to
fracture where EQVS (H-M-H equivalent stress) exceed significantly the
adopted yield limit R0.2 = 359 MPa. Considerable stress concentration
in some areas results from the bending effect under high compressive
load. Similar modes of fracture for outer layers (L1, L7) and more
uniform stress distribution for inner layer (L4) confirm the symmetry of
the shell model in FEA. The equivalent stress patterns in aluminium
layers indicate the damage occurrence along profile's web-flange joint,
same as failure initiation observed in experimentally damaged speci-
mens. Similar results are confirmed in numerical simulations by Tsai-
Wu (TWSR), Hashin (HMAT, HFIB) and Puck (PMAT, PFIB) criteria that
confirm that a fracture occurs along the profile's mid-length at the web-
flange joint. Maps of FF were determined for loads higher than struc-
ture's carrying capacity in order to determine the collapsed mode
shapes after the column fracture. Further comparison with experi-
mental evidences (Fig. 5f) proved that modes of fracture exhibit similar
patterns which suggests good compliance within numerical and ex-
perimental results. For more detailed analysis of failure criteria appli-
cation and stress tensor elements participation into FML failure see
Refs. [19].
5.2. Progressive damage analysis
In progressive damage analysis with material degradation model,
the Hashin criterion was applied to predict the onset of material da-
mage in composite layer. The damage evolution law after failure in-
itiation was performed by the material property degradation method
(MPDG). Thus, according to specific rules described in Section 2.2. and
based on the damage variables (df , dm) the stiffness was gradually re-
duced once the failure was initiated. For comparison reasons, various
PFA models were defined in FE code to observe the influence of damage
variables onto the FML failure mechanism:
Fig. 9. Damage modes a) fibre tensile, b) fibre compressive, c) matrix tensile, d) matrix compressive.
60
Thin-Walled Structures 122 (2018) 52–63
• PFA1: d = 0.25, d = 0.25
f m
• PFA2: d = 0.25, d = 0.5
f m
• PFA3: d = 0.5, d = 0.5
f m
• PFA4: d = 0.5, d = 0.75
f m
• PFA5: d = 0.75, d = 0.75
f m
• PFA6: d = 0.75, d = 1
f m
• PFA7: d = 1, d = 1
f m
• NO_PFA: d = 0, d = 0 (geometrically nonlinear analysis without
f m
stiffness reduction)
Note, that df / m = 1 indicates complete stiffness reduction after
failure initiation. Models PFA2, PFA4 and PFA6 assume different da-
mage variables for fibre and matrix in order to simulate various sus-
ceptibility to failure of these constituents. For aforementioned PFA
models the post-buckling equilibrium paths were determined and
compared with experimental results for flange and web separately (see
Fig. 6a,b). As it can be noticed, depending on damage variables adopted
to progressive model the ultimate load in FEA varies in the range of
44.3 kN (PFA_7) to 48.6 kN (PFA_1). Highest ultimate load in the value
of 52.9 kN was predicted for model without stiffness reduction
(NO_PFA). The yield strength from ARAMIS experimental measure-
ments is 47.3 kN and 46.8 kN determined from flange and web de-
flection respectively. For flange analysis, the best correlation between
experimental test and PFA simulation was obtained for model PFA_7
(df = 1, dm = 1) , whereas for web it was achieved in model PFA_4
(df = 0.5, dm = 0.75). Such comparison is presented for flange and web
separately in the Fig. 7.
For obvious reasons the equilibrium path for flange and web differs
from each other. One of the flanges’ longitudinal edges are free of
constrains and therefore the deflection is slightly ‘deeper’ for corre-
sponding compression force level. For the PFA models there is a no-
ticeable bend of the curve that similarly as in the experiment it corre-
sponds to the maximum compressive force that can be carried by the
structure. In the case of geometrically nonlinear analysis without the
implementation of progressive failure algorithm (NO_PFA), the deflec-
tion increased gradually with the load. The comparative analysis of
various PFA models for flange deflections indicates that the complete
stiffness reduction (df / m = 1) gives the best agreement with experi-
mental results (see Fig. 7a). Similar results were obtained for PFA of
eight-layered channel section composite columns subjected to uniform
D. Banat, R.J. Mania Thin-Walled Structures 122 (2018) 52–63

Fig. 10. Maps of failure criteria (a) TWSR, (b) HMAT, (c) HFIB, (d) PMAT, (e) PFIB with progressive stiffness reduction.

compression [17]. Nevertheless, in case of the web deflections analysis damage variables for fibre and matrix failure in order to simulate var-
the model PFA_4 (df = 0.5, dm = 0.75) corresponds most accurately with ious susceptibility to damage of these two constituents.
ARAMIS measurements. It suggests that the progressive damage ana- Presented in that manner PFA comparative analysis implies that
lysis of composite material should take into consideration different both PFA models (PFA_7 and PFA_4) can be applied with a good degree

61
D. Banat, R.J. Mania
of success to analyse the progressive failure of composite structures
subjected to compression. However, it is recommended to verify PFA
models with different damage variables for specific loading conditions.
Further analysis of progressive failure models included the con-
siderations of the damage status (0 - undamaged, 1 - damaged, 2 -
completely damaged). Results for all composite layers (L2, L3, L5, L6)
with stiffness reduction factor df / m = 0.25 (model PFA_1) are presented
in Fig. 8. The larger fracture is observed in layers L2, L6 (Fig. 8a,d)
since outer layers are greatly exposed to bending effect in post-buckling
range. For the same reason, layers L3, L5 (Fig. 8b,c) that are in fact
closer the shell's mid-plane, exhibit smoother damage status. Note, that
due to the symmetry of profile lay-up, the behaviour of GFRP layers is
symmetrical with respect to profile mid-plane. Damage status at each
layer confirms also previous results where the failure of the laminate
occurred along profile's mid-length at the web-flange joint. Therefore,
the material property degradation method (MPDG) can be effectively
used for progressive failure analysis in FRP.
Various damage statuses were also developed to observe matrix and
fibre failure separately. Results for progressive model (PFA_1) are
shown in Fig. 9. Such analysis allowed to compare different failure
mechanisms within GFRP layer that are as follow: (a) FT - fibre tensile
damage, (b) FC - fibre compressive damage, (c) MT - matrix tensile
damage, (d) MC - matrix compressive damage. Herein, FF = 0.25 in-
dicates complete damage of the material structure. These results prove
also the First Ply Failure analysis of the FMLs [19] where fibre failure
mechanism was initiated in the corners and propagated towards the
web and the flange of the column. However, for angle-ply multi-layered
laminates it is not possible to unambiguously assess the type of either
fibre and matrix failure. Literature surveys argues that the most
common form of damage initiation is the matrix cracking on the verge
of fibre-matrix joint but the loss of the load carrying capacity is mainly
affected by the failure initiation within specific fibres [17].
Detailed progressive analysis was also performed for model PFA_7
with stiffness reduction factor df / m = 1. Failure criteria were again ap-
plied to analyse maps of failure factors in FRP layer wherein stiffness
was progressively reduced in the loading process. Results for Tsai-Wu,
Hashin and Puck criteria are shown in Fig. 10. That allowed to analyse
the impact of highest stiffness reduction factor onto the maps of failure
factors. According to progressive model definition, the fibre and matrix
failure initiation was assessed by Hashin criterion. Hence, the effect of
stiffness reduction is mainly noticeable for maps of FF determined by
this criterion (Fig. 10b,c). However, similar results are obtained for
other criteria where the highest FF is approx. 1 or lower, which con-
firms the concept of material stiffness reduction in FEM application.
6. Conclusions
This paper concerns the failure analysis of the thin-walled Z-shape
section FML profiles subjected to axial compression. Initial stability
considerations in FEM included linear buckling and geometrically
nonlinear analysis. Determined buckling mode shape and equilibrium
paths were found to be in good agreement with experimental mea-
surements. Further analysis with the failure criteria implementation
allowed one to predict the members collapsed mode shapes that cor-
responded to the experimentally damaged specimens.
Detailed failure assessment in FEM included the implementation of
progressive damage analysis with the material degradation model.
Herein, Hashin criterion was used to indicate the onset of material
failure in GFRP layer. The damage evolution law was performed by the
material property degradation method (MPDG). Therefore, based on
the damage variables the stiffness was gradually reduced in the area
where failure was initiated. That allowed one to estimate the impact of
stiffness reduction onto the FML's failure modes. A comparative analysis
of the numerical results suggested that the implementation of the high
values of the damage variables PFA_7 (df / m = 1) and PFA_4 (df = 0.5,
dm = 0.75) provides the best agreement with the experimental tests. For
62
Thin-Walled Structures 122 (2018) 52–63
PFA models, the equilibrium paths were predicted to correspond better
with experimental curves than ones determined by geometrically non-
linear analysis without stiffness degradation method. Nevertheless, the
implementation of PFA in the modelling process increases the compu-
tation time required to solve the appropriate equilibrium equations for
post-buckling analysis.
Damage statuses displayed for PFA models at each layer separately
confirmed also previous FML failure analysis where the laminate failure
occurred along profile's mid-length at the web-flange joint. Therefore, it
can be claimed that MPDG method can be applied with a good degree of
success to analyse the progressive failure of GFRP subjected to com-
pression. However, authors recommend verifying PFA models with
different damage variables for specific loading conditions. The im-
plementation of progressive damage model proved also that the First
Ply Failure (FPL) strength of a laminate does not indicate the structure's
loss of carrying capacity. The laminate can still carry loads higher load
before fracture and therefore the PFA is recommended to predict the
laminate's ultimate strength. Such analysis identifies the various failure
modes in full load range which allows one to truly understand the ad-
vantage of hybrid composites over other conventional materials.
The future analysis might include the detailed comparison of PFA
application for the various laminate stacking sequences of plies with
different fibre alignment. That could provide important insights for
composite's manufacturers in the designing process. Furthermore, for
the purpose of this study the progressive failure analysis was performed
with the use of material property degradation method and the im-
plementation the Hashn's criterion to indicate damage initiation.
Nevertheless, one should be also aware that once the damage initiation
criterion is satisfied, further loading will cause not only degradation of
material stiffness coefficients but also energy release of specific rate.
Hence, an attempt should be also taken in the future to perform PFA by
means of continuum damage mechanics method (CDM). However, it
requires energy dissipation constants and viscous damping coefficients
implementation that are not readily available for most unidirectional
composites (including FML).
Acknowledgment
This study was supported by the Ministry of Science and Higher
Education of Poland - National Science Centre Grant no. DEC-2012/07/
B/ST8/04093.
References
[1] R.M. Jones, Mechanics of Composite Materials, 2nd ed., Taylor & Francis,
Philadelphia, 1999, http://dx.doi.org/10.1007/BF00611782.
[2] R. Mania, Comparative static buckling study of FML thin-walled profiles, in:
Proceedings of the ECCM16 – 16th European Conference on Composite Materials,
Seville, 2014.
[3] A. Vlot, J.W. Gunnink, Fibre Metal Laminates: An Introduction, Springer
Science & Business Media, Dordrecht, the Netherlands, 2001 (Accessed 5 April
2015), https://books.google.com/books?id=LxJUKNSiOEMC&pgis=1.
[4] P. Czapski, T. Kubiak, Influence of fibre arrangement on the buckling load of
composite plates - analytical solution, Fibres Text. East. Eur. 5 (2015) 92–97,
http://dx.doi.org/10.5604/12303666.1161764.
[5] R.J. Mania, C.B. York, Buckling strength improvements for Fibre Metal Laminates
using thin-ply tailoring, Compos. Struct. 159 (2017) 424–432, http://dx.doi.org/
10.1016/j.compstruct.2016.09.097.
[6] P. Soltani, M. Keikhosravy, R.H. Oskouei, C. Soutis, Studying the tensile behaviour
of GLARE laminates: a finite element modelling approach, Appl. Compos. Mater. 18
(2011) 271–282, http://dx.doi.org/10.1007/s10443-010-9155-x.
[7] N.E. Prasad, R.J.H. Wanhill, Aerospace Materials and Material Technologies, 1st
ed., Springer, Singapore, 2017.
[8] M. Urbaniak, J. Swiniarski, P. Czapski, T. Kubiak, Experimental investigations of
thin-walled GFRP beams subjected to pure bending, Thin-Walled Struct. 107 (2016)
397–404, http://dx.doi.org/10.1016/j.tws.2016.06.022.
[9] H. Debski, A. Teter, T. Kubiak, Numerical and experimental studies of compressed
composite columns with complex open cross-sections, Compos. Struct. 118 (2014)
28–36, http://dx.doi.org/10.1016/j.compstruct.2014.07.033.
[10] P. Wysmulski, H. Dębski, P. Różyło, K. Falkowicz, A study of stability and post-
critical behaviour of thin-walled composite profiles under compression, Eksploat. I
Niezawodn. - Maint. Reliab. 18 (2016) 632–637 (doi:10.17531).
D. Banat, R.J. Mania
[11] P. Czapski, T. Kubiak, Numerical and experimental investigations of the post-
buckling behaviour of square cross-section composite tubes, Compos. Struct. 132
(2015) 1160–1167, http://dx.doi.org/10.1016/j.compstruct.2015.07.039.
[12] K. Falkowicz, P. Mazurek, P. Różyło, P. Wysmulski, W. Smagowski, Experimental
and numerical analysis of the compression thin-walled composite plate, Adv. Sci.
Technol. 10 (2016) 177–184, http://dx.doi.org/10.12913/22998624/64063.
[13] K. Falkowicz, H. Dębski, Postbuckling behaviour of laminated plates with a cut-out,
Adv. Sci. Technol. 11 (2017) 186–193, http://dx.doi.org/10.12913/22998624/
68283.
[14] T. Kubiak, Z. Kolakowski, J. Swiniarski, M. Urbaniak, A. Gliszczynski, Local buck-
ling and post-buckling of composite channel-section beams – numerical and ex-
perimental investigations, Compos. Part B Eng. 91 (2016) 176–188, http://dx.doi.
org/10.1016/j.compositesb.2016.01.053.
[15] Z. Kołakowski, R.J. Mania, Semi-analytical method versus the FEM for analysis of
the local post-buckling of thin-walled composite structures, Compos. Struct. 97
(2013) 99–106.
[16] F. Moriniere, Low-Velocity Impact on Fibre-metal Laminates (Ph.D. thesis), TU
Delft, 2014.
[17] A. Gliszczynski, T. Kubiak, Progressive failure analysis of thin-walled composite
columns subjected to uniaxial compression, Compos. Struct. (2016), http://dx.doi.
org/10.1016/j.compstruct.2016.10.029.
[18] H. Altenbach, T. Sadowski (Eds.), Failure and Damage Analysis of Advanced
Materials, Springer, Udine, Italy, 2014(Accessed 12 March 2015), https://books.
google.com/books?id=h9OlBQAAQBAJ&pgis=1.
[19] D. Banat, R.J. Mania, Comparison of failure criteria application for FML column
buckling strength analysis, Compos. Struct. 140 (2016) 806–815.
[20] D. Wang, W. Wen, Three-dimensional progressive damage analysis of composite
laminates containing a central hole subjected to compressive loading, in:
Proceedings of the 2012 International Conference of Mechanical Engineering
Research Materials Science, 2012, pp. 367–371. doi: 〈http://doi.org/10.2991/
mems.2012.19〉.
[21] Z. Zhang, H. Chen, L. Ye, Progressive failure analysis for advanced grid stiffened
composite plates/shells, Compos. Struct. 86 (2008) 45–54, http://dx.doi.org/10.
1016/j.compstruct.2008.03.037.
[22] Y. Pan, Stiffness and Progressive Damage Analysis on Random Chopped Fiber
Composite Using FEM (Ph.D. Thesis), The State University of New Jersey, New
Brunswick, USA, 2010.
[23] M. Hinton, A.S. Kaddour, S. P.D, A further assessment of the predictive capabilities
of current failure theories for composites laminates: comparison with experimental
evidence, Compos. Sci. Technol. 64 (2004) 549–588.
[24] L. Wang, C. Zheng, H. Luo, S. Wei, Z. Wei, Continuum damage modeling and
progressive failure analysis of carbon fiber/epoxy composite pressure vessel,
Compos. Struct. 134 (2015) 475–482, http://dx.doi.org/10.1016/j.compstruct.
2015.08.107.
[25] M.L.P. Tonatto, M.M.C. Forte, V. Tita, S.C. Amico, Progressive damage modeling of
spiral and ring composite structures for offloading hoses, Mater. Des. 108 (2016)
374–382, http://dx.doi.org/10.1016/j.matdes.2016.06.124.
[26] A.S. Kaddour, M.J. Hinton, P.A. Smith, S. Li, The background to the third world-
wide failure exercise, J. Compos. Mater. 47 (2013) 2417–2426, http://dx.doi.org/
10.1177/0021998313499475.
[27] N. Knight, Factors influencing progressive failure analysis predictions for laminated
composite structures, in: Proceedings of the 49th AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference, 2008, p. 92047.
[28] E.J. Barbero, F.A. Cosso, R. Roman, T.L. Weadon, Determination of material para-
meters for Abaqus progressive damage analysis of E-glass epoxy laminates, Compos.
Part B Eng. 46 (2013) 211–220, http://dx.doi.org/10.1016/j.compositesb.2012.09.
069.
63
Thin-Walled Structures 122 (2018) 52–63
[29] A. Puck, J. Kopp, M. Knops, Guidelines for the determination of the parameters in
Puck's action plane strength criterion, Compos. Sci. Technol. 62 (2002) 371–378.
[30] M. Knops, Analysis of Failure in Fiber Polymer Laminates: the Theory of Alfred
Puck, Springer Berlin, Heidelberg, 2008.
[31] E.J. Barbero, Finite Element Analysis of Composite Materials, 2nd ed., CRC Press,
Boca Raton, 2007https://books.google.com/books?id=tWx_NLLD14wC&pgis=1.
[32] User's Guide ANSYS® Academic Research, Help System, (n.d.).
[33] P. Maimí, P.P. Camanho, J.A. Mayugo, C.G. Dávila, A continuum damage model for
composite laminates: part I - constitutive model, Mech. Mater. 39 (2007) 897–908,
http://dx.doi.org/10.1016/j.mechmat.2007.03.005.
[34] B.W. Rosen, Tensile failure of fibrous composites, AIAA J. 2 (1964) 1985–1991,
http://dx.doi.org/10.2514/3.2699.
[35] L.M. Kachanov, Time of the rupture process under creep conditions, Izv. Akad.
Nauk. SSSR Otd. Teckhnicheskikh Nauk. 8 (1958) 26–31.
[36] Y.N. Rabotnov, On the equation of state of creep, in: Proceedings of the Institution
of Mechanical Engineers Conference, 1963, pp. 117–122.
[37] D. Krajcinovic, J. Trafimow, D. Sumarac, Simple constitutive model for a cortical
bone, J. Biomech. 20 (1987) 779–784.
[38] D. Krajcinovic, Damage mechanics, Mech. Mater. 8 (1989) 117–197, http://dx.doi.
org/10.1016/0167-6636(89)90011-2.
[39] S. Jansson, U. Stigh, Influence of cavity shape on damage parameter, J. Appl. Mech.
52 (1985) 609, http://dx.doi.org/10.1115/1.3169109.
[40] D. Krajcinovic, Continuous damage mechanics revisited: basic concepts and defi-
nitions, J. Appl. Mech. 52 (1985) 829, http://dx.doi.org/10.1115/1.3169154.
[41] D. Krajcinovic, D. Fanella, A micromechanical damage model for concrete, Eng.
Fract. Mech. 25 (1986) 585–596, http://dx.doi.org/10.1016/0013-7944(86)
90024-X.
[42] A.P.C. Duarte, A. Díaz Sáez, N. Silvestre, Comparative study between XFEM and
Hashin damage criterion applied to failure of composites, Thin-Walled Struct. 115
(2017) 277–288, http://dx.doi.org/10.1016/j.tws.2017.02.020.
[43] J. Bieniaś, R.J. Mania, P. Jakubczak, K. Majerski, The issues of manufacturing of
geometrically complicated elements by using FML laminates, Compos. Theory
Pract. 15 (2015) 243–249.
[44] J. Bieniaś, Fibre Metal Laminates - some aspects of manufacturing process, structure
and selected properties, Compos. Theory Pract. 11 (2011) 39–43.
[45] R.J. Mania, Z. Kolakowski, J. Bienias, P. Jakubczak, K. Majerski, Comparative study
of FML profiles buckling and postbuckling behaviour under axial loading, Compos.
Struct. 134 (2015) 216–225, http://dx.doi.org/10.1016/j.compstruct.2015.08.093.
[46] M. Kamocka, M. Zglinicki, R.J. Mania, Multi-method approach for FML mechanical
properties prediction, Compos. Part B Eng. 91 (2016) 135–143, http://dx.doi.org/
10.1016/j.compositesb.2016.01.014.
[47] D. Banat, Z. Kolakowski, R.J. Mania, Investigations of fml profile buckling and post-
buckling behaviour under axial compression, Thin-Walled Struct. 107 (2016)
335–344, http://dx.doi.org/10.1016/j.tws.2016.06.018.
[48] D. Banat, R.J. Mania, Failure assessment of thin-walled FML profiles during buck-
ling and postbuckling response, Compos. Part B Eng. 112 (2017) 278–289, http://
dx.doi.org/10.1016/j.compositesb.2017.01.001.
[49] T. Kubiak, M. Urbaniak, G. Zucco, A. Madeo, Imperfection sensitivity analysis of the
nonlinear stability of composite beams e Numerical and experimental investiga-
tions, Compos. Part B. 94 (2016) 360–369, http://dx.doi.org/10.1016/j.
compositesb.2016.03.067.
[50] W. Węglewski, K. Bochenek, M. Basista, T. Schubert, U. Jehring, J. Litniewski,
S. Mackiewicz, Comparative assessment of Young's modulus measurements of
metal-ceramic composites using mechanical and non-destructive tests and micro-CT
based computational modeling, Comput. Mater. Sci. 77 (2013) 19–30, http://dx.
doi.org/10.1016/j.commatsci.2013.04.007.