th

8 ANKARA INTERNATIONAL AEROSPACE CONFERENCE AIAC-2015-078

10-12 September 2015 - METU, Ankara TURKEY

PARAMETIC STUDY OF DELAMINATION ANALYSIS IN COMPOSITES WITH

COHESIVE ZONE METHOD

Busra Bartan1and Altan Kayran2 Bülent ACAR3

Middle East Technical University Roketsan Missiles Inc.
Ankara, Turkey Ankara, Turkey

ABSTRACT
In this article, parametric study is conducted for delamination analysis in composites using three
dimensional (3D) solid cohesive elements. The main objective of the study is to investigate the effect
of normal mode fracture energy, stiffness of cohesive elements and element size on the damage
initiation and propagation utilizing three dimensional (3D) solid cohesive elements in an implicit finite
element analysis framework. Parametric study is conducted for Double Cantilever Beam (DCB) and
End Notch Flexure (ENF) specimens in ABAQUS with cohesive zone method (CZM) to model the
delamination initiation and propagation. For verification purposes, results of the delamination analyses
for Double Cantilever Beam (DCB) and End Notch Flexure (ENF) specimens are compared with the
experimental and numerical results available in the literature. The composite layups and cohesive
zone are generated with 8 node solid elements. In the literature, 3D solid cohesive elements are
generally used in combination with explicit solution model because of better convergence
characteristics of the explicit finite element analysis. In this study, a step procedure is used in order to
perform the delamination analysis by implicit finite element analysis. The analysis results are seen to
be compatible with the available experimental and the numerical results available in the literature.

INTRODUCTION
The usage of composite materials is very extensive in aerospace industry because of their high
specific stiffness, high specific strength, corrosion and chemical resistance. The widespread failure
behavior of composites are fiber or matrix tension/compression failure, intralaminar failure,
interlaminar failure, matrix cracking, fiber and matrix debonding [Camanho, Davila, and Ambur, 2001;
O’Brien, 2001]. Interlaminar damage (i.e. delamination, debonding) in composite structures is very
critical because it causes progressive failure of layer separation in laminated composites [Johnson,
Louca, Mouring and Calvano, 2006]. There are many examples on the use of cohesive zone method
(CZM) and interlaminar progressive failure behavior in the literature [Saryanarayana, Bogert, and
Chunchu 2007; Alfano, Furgiuele, Leonardi, Maletta, and Paulino 2007; Kyongchan and Davila 2008;
Travesa, 2006; Diehl, 2004].
In this study, Double Cantilever Beam (DCB) and End Notch Flexure (ENF) specimens are modeled in
three dimensions in ABAQUS and 3D cohesive elements are used in the delamination interface. In the
literature, for the delamination analysis in 3D, generally explicit finite element analysis is performed for
better convergence characteristics. In the present study, implicit finite element analysis is used in the
3D models of the DCB and ENF specimens with 3D cohesive elements in the delamination interface.
For the verification study, delamination analysis results of the present study are compared with the
DCB and ENF analysis and test data of Travesa [Travesa, 2006]. After the verification analysis,
parametric study is conducted to investigate the effect of the selected parameters, which are taken as
normal mode fracture energy (GIC) and shear mode fracture energy (GIIC), stiffness of cohesive

1
Graduate Student. in Department of Aerospace Engineering, Email: busra.bartan@metu.edu.tr
2
Prof. in Department of Aerospace Engineering, Email: akayran@metu.edu.tr
3
Manager in Structural Thermal and Dynamic Analysis and Test Department, Email: bacar@roketsan.com.tr
AIAC-2015-078 Bartan, Kayran & Acar

elements (penalty stiffness) and element size, on the damage initiation and propagation utilizing three
dimensional (3D) solid cohesive elements in implicit finite element analysis. In the delamination
analysis using cohesive elements, the results are dependent on the selection of certain analysis
parameters. For accurate prediction of the delamination initiation and growth, analysis model has to be
high fidelity model. Therefore, in the present study, parametric study is conducted using 3D finite
element model with 3D cohesive elements and the effect of normal mode fracture energy, stiffness of
cohesive elements and element size damage behavior is investigated.

METHOD
In the present study, DCB and ENF analysis and test data of Travesa [Travesa, 2006] is taken as
reference to build the finite element model for the delamination analysis and to compare the results of
the present analysis with the results of Travesa [Travesa, 2006]. The ply properties of the
unidirectional AS4/PEEK carbon fiber reinforced composite, used by Travesa, is given in Table 1. In
6 3
his analysis, Travesa took the stiffness of cohesive elements is as 10 N/mm , normal strength of
cohesive elements as 80 MPa and shear strength of cohesive elements as 100 MPa. The specimens
modeled are 102-mm-long, 25.4-mm-wide, with two 1.56-mm-thick arms. Models used by Travesa
have 150 cohesive elements along the length of the specimens, and 4 cohesive elements along the
width [Travesa, 2006].

Table 1 : Ply Properties [Travesa, 2006]

E11 E22=E33 G12=G13 G23 v12=v13 V23 Gıc Gııc The B-K
Parameter
122.7 10.1 5.5 3.7 0.25 0.45 0.969 1.179 2.284
2 2
GPa GPa GPa GPa kJ/m kJ/m

The finite element models of the DCB and ENF specimens, shown in Fig.1, are performed by
ABAQUS v.6.13-1 [Dassault Systemes, 2013] – Standard Implicit Model. Composite layups are
modelled with 8-node linear brick, incompatible mode elements (C3D8I) and the cohesive layer is
modelled with 8-node three-dimensional cohesive elements (COH3D8). In the finite element models,
global element size is taken as 5 mm and the cohesive element size is taken as 0.5 mm. DCB
specimen is exposed to pure Mode I loading and ENF specimen is exposed to pure Mode II loading.
DCB specimen is clamped at the end of the specimen whereas ENF specimen is pinned at the ends of
the specimen and three point bending simulation is performed. The applied displacements at the end
of the DCB specimen are 4.5 mm in top and bottom directions. The applied displacement for ENF is
6.5 mm in the middle of the specimen. Boundary conditions for DCB and ENF are displayed in Fig.2.

Figure 1: Finite Element Model for DCB and ENF Specimens

2
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

Figure 2: Boundary Conditions for DCB and ENF Specimens

ANALYSIS DEMONSTRATION
The step procedure of the delamination analysis is selected as dynamic-implicit step with quasi-static
application because of the convergence that it provides in 3D implicit analyses. The other step
procedures are static general step with default definition and static general step with automatic
stabilization definition. The analysis results do not match the results of Travesa [Travesa, 2006] if
static general steps are used. The experimental data and numerical predictions of Travesa [Travesa,
2006] and present analysis results with three step procedures for the DCB specimen are shown in Fig.
3. In the static general step with automatic stabilization definition, the loss of stiffness occurs before
the damage starts. In the static general step with default definition, analysis does not converge and
stops at damage initiation, so the damage evolution cannot be observed. It can be seen that the step
procedure dynamic-implicit step with quasi-static application predicts the damage initiation and
progression properly and in accordance with the results of Travesa [Travesa, 2006].

Figure 3: Experimental, numerical and present analysis results with different step procedures
–pure mode I loading

The numerical study of Travesa [Travesa, 2006] is performed in ABAQUS utilizing a user-written
element subroutine (UEL). Travesa used, 3D elements and convergence of the analysis is provided by
control parameters. The experimental data and numerical predictions of Travesa [Travesa, 2006] and

3
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

present analysis results in the form of load displacement plots are shown in Fig. 4 and Fig. 5 for DCB
and ENF specimens. It is seen that present analysis results are virtually compatible with the
experimental and numerical results of Travesa. With this exercise, it is shown that if the step
procedure is taken as dynamic-implicit, the analyses converge when 3D finite element model and 3D
cohesive elements are used in combination and true damage behavior can be observed.

Figure 4: Experimental, numerical and analysis results –pure mode I loading

Figure 5: Experimental, numerical and analysis results –pure mode II loading

Damage initiation and propagation for the DCB and the ENF specimens are shown in Figs. 6-9. In
Figs. 6-9, scalar stiffness degradation (SDEG) plots are show. If SDEG is equal to 1, it means that the
4
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

cohesive element is not active and layers are fully separated and if SDEG is equal to 0, the cohesive
element is fully active.

Figure 6: Delamination initiation in the DCB specimen

Figure 7: Delamination propagation in the DCB specimen

Figure 8: Delamination initiation in the ENF specimen

5
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

Figure 9: Delamination propagation in ENF specimen in present analysis

EFFECTIVE PARAMETERS ON THE DAMAGE BEHAVIOUR

Parametric study is conducted to investigate the effect of certain parameters involved in the
delamination analysis on the damage behavior. These parameters are selected as normal mode
fracture energy (GIC), stiffness of cohesive elements (K) and element sizes (global element size-GES
and cohesive element size-CES). A parametric study is required because stiffness of cohesive
elements and element size must be chosen properly for correct finite element analysis. On the other
hand, critical strain energy release rates GIC and GIIC are usually measured by DCB and ENF tests In
DCB and ENF tests, wide scatter of the test data is very common. Therefore, the effect of variation of
the critical strain energy release rate about the nominal values on the delamination initiation and
progression has to be investigated. Critical fracture energy or critical strain energy release rate, GIC
(normal mode fracture energy) and GIIC (shear mode fracture energy) are two most important
parameters which have to be correctly estimated or measured. Strain energy relase rate governs the
amount of the energy release if a crack growth occurs in surface-bonded structures [Diehl, 2004]. The
2 2
nominal values of GIC and GIIC are taken as 0.969 kJ/m and 1.719 kJ/m in this study [Travesa, 2006].
Additionally, a higher and a lower value for the critical strain energy release rates are used in the
analyses to see their effect on delamination initiation and progression. Figure 10 and Figure 11 show
the effect of critical strain energy release rates on the delamination initiation and progression for DCB
and ENF specimens. These figures show that the delamination starts at higher loads if the values of
GIC and GIIC are higher than the nominal values and delamination starts at lower loads if the values of
GIC and GIIC are lower than the nominal values, as expected.

6
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

Figure 10: Experimental, numerical and analysis results for DCB– the effect of the critical
fracture energy or critical energy release rate, GIC

Figure 11: Experimental, numerical and analysis results for ENF – the effect of the critical
fracture energy or critical energy release rate, GIIC

The other important parameter in the delamination analysis is the stiffness of cohesive elements or the
penalty stiffness, K. The initial stiffness of the cohesive elements should be high enough to provide the
general compliance before the damage starts [Khoramishad, Crocombe, Katnam, and Ashcroft, 2010].

7
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

The stiffness of cohesive elements or the interface stiffness or the penalty stiffness, K, is the slope of
the constitutive equation before damage initiation as shown in Figure 12. The effective elastic
properties of the composite will not be affected by the cohesive surface whenever E33<<Kt, i.e.,

K ≥ αE33 / t (1)

Where t is the thickness of composite layups (adjacent sub-laminates) and α is a parameter much
larger than 1 and it is recommended that α should be chosen as 50 [Turon, Davila and Camanho,
3
2007]. In this study, the penalty stiffness is determined as K≈300000 N/mm with Equation 1.
Therefore, the values higher than 300000 should be used for K. Travesa took the stiffness of cohesive
6 3
elements is as 10 N/mm essentially [Travesa, 2006].

Figure 12: Bilinear constitutive equation [Turon, Davila and Camanho, 2007]

Figure 13 and Figure 14 show the effect of penalty stiffness on delamination initiation and progression
5 6
for DCB and ENF specimens. In the literature, penalty stiffness is generally chosen as 10 or 10 kJ/m.
5 6 2
As long as penalty stiffness is taken in the 10 - 10 kJ/m range, the load–displacement curves are
relatively insensitive to the penalty stiffness and interfacial strength, but very sensitive to the mesh
size. If the value of penalty stiffness is chosen as an overly high value, there are numerical problems
in analysis. Damage starts early so analysis results are wrong. On the other hand, damage cannot be
observed thoroughly with an overly low value of penalty stiffness [Turon, Davila and Camanho, 2007].

Figure 13: Experimental, numerical and analysis results for DCB – the stiffness of cohesive
elements or penalty stiffness, K

8
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

Figure 14: Experimental, numerical and analysis results for ENF – the stiffness of cohesive
elements or penalty stiffness, K

One of the most critical parameter in the delamination analysis is the effect of global and cohesive
element size on the delamination initiation and progression. Figures 15-20 shows the effect of global
and cohesive element sizes on the delamination initiation and progression for DCB and ENF
specimens. The experience in the literature is to use a cohesive element size of 0.5 mm or less in
order to predict the delamination initiation and progression accurately [Harper and Hallett, 2008]. If the
element size increases, the damage initiation starts at higher loads and the analysis will not be
correct.
Global element size-GES and cohesive element size-CES should be optimum value for DCB
specimens so analysis converges. If GES is chosen as quite high value - 5 mm, CES should be
chosen as 0.5 mm or lower. Damage starts at lower loads when GES-5 mm and CES-1 mm or GES-5
mm and CES-5 mm (Figure 15). If GES is chosen as high - 1 mm, CES cannot be chosen as 1 mm
because damage starts at lower load (Figure 16). If GES and CES are chosen as 0.5 mm, damage
initiation and propagation has correct behavior (Figure 17). However, it is not necessary that the value
of GES is too small because of increment in solving time of analysis.

9
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

Figure 15: The effects of element sizes on the damage behavior

Figure 16: The effects of element sizes on the damage behavior

10
Ankara International Aerospace Conference
 AIAC-2015-078 Bartan, Kayran & Acar

Figure 17: The effects of element sizes on the damage behavior

Global element size-GES and cohesive element size-CES should be optimum value for ENF specimens to
converge analysis. If GES is chosen as quite high value - 5 mm, CES should be chosen as 1 mm or lower.
Damage starts at higher loads when GES-5 mm and CES-5 mm (Figure 18). If GES is chosen as high - 1
mm, CES can be chosen as 1 mm or lower (Figure 19). If GES and CES are chosen as 0.5 mm, damage
initiation and propagation has correct behavior (Figure 20). However, it is not necessary that the value of
GES is too small because of increment in solving time of analysis.

Figure 18: The effects of element sizes on the damage behavior

11
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

Figure 19: The effects of element sizes on the damage behavior

Figure 20: The effects of element sizes on the damage behavior

CONCLUSION

12
Ankara International Aerospace Conference
AIAC-2015-078 Bartan, Kayran & Acar

In this article, parametric study is performed for Double Cantilever Beam (DCB) and End Notch
Flexure (ENF) specimens in ABAQUS with cohesive zone method (CZM) to model the delamination
initiation and propagation. For verification purposes, results of the delamination analyses for Double
Cantilever Beam (DCB) and End Notch Flexure (ENF) specimens are compared with the experimental
and numerical results available in the literature. In this study, dynamic-implicit step with quasi-static
application step procedure is used in conjunction with 3D global and cohesive finite elements in order
to perform the delamination analysis by implicit finite element analysis. The analysis results are seen
to be compatible with the available experimental and the numerical results in the literature. The results
of parametric study for the DCB and the ENF specimens are presented for the normal and shear
mode fracture energy (GIC, GIIC), stiffness of cohesive elements (K) and global and cohesive element
sizes. The fracture energies (GIC and GIIC) are generally determined by tests. The present analysis
showed that the use of lower or higher fracture energy than the nominal value caused under and over
estimation of delamination initiation, respectively. Proper experimental it is also concluded that the
penalty stiffness should not be low because with low penalty stiffness value delamination initiation
cannot be predicted. In the literature, the recommended value of the penalty stiffness is in the range of
5 6 2
10 or 10 kJ/m . The analyses performed in the present study also confirm the suitability of this range.
Finally, the size of cohesive elements must be 0.5 mm or less in order to predict the delamination
onset and progression accurately.

References
Alfano, M., Furgiuele, F., Leonardi, A., Maletta, C. and Paulino, G.H. (2007) Cohesive Zone Modeling
of Mode I Fracture in Adhesive Bonded Joints, Key Engineering Materials, Vol. 348-349, p; 13-
16, 2007.
Camanho,P.P., Davila, C.G. and Ambur D.R. (2001) Numerical Simulation of Delamination Growth in
Composite Materials, NASA /TP-2001-211041, 2001.
Dassault Systemes, ABAQUS UNIFIED FEA, 2013.
Diehl,T. (2004) Modelling Surface-Bonded Structures with ABAQUS Cohesive Elements: Beam-Type
Solutions, ABAQUS Users’Conference, 2004.
Harper, P.W. and Hallett, S.R. (2008) Cohesive Zone Length in Numerical Simulations of Composite
Delamination, Engineering Fracture Mechanics, Vol. 75, p: 4774-4792, 2008.
Johnson, H.E., Louca, L.A., Mouring S.E. and Calvano, C.N. (2006) Damage Modeling of Large and
Small Scale Composite Panels Subjected to a Low Velocity Impact, 2006.
Khoramishad, H., Crocombe, A.D., Katnam, K.B. and Ashcroft, L.A. (2010) Predicting fatigue damage
in adhesively bonded joints using a cohesive zone model, International Journal of Fatigue, Vol.
32, p: 1146-1158, 2010.
Kyongchan, S. and Davila, C.G. (2008) Guidelines and Parameter Selection for Simulation of
Progressive Delamination, ABAQUS Users’Conference, 2008.
O’Brien, T. K., (2001), Fracture Mechanics of Composite Delamination. ASM Handbook, Vol.21, 2001.
Saryanarayana, A., Bogert, P.B., and Chunchu P.B. (2007) The Effect of Delamination on Damage
Path and Failure Load Prediction for Notched Composite Laminates , 48 AIAA Structures,
Structural Dynamics and Material Conference, 2007.
Travesa. T.A. (2006) Simulation of Delamination in Composites Under Quasi-Static and Fatigue
Loading Using Cohesive Zone Models, Doctoral Thesis Universitat de Girona, 2006.
Turon A, Davila CG, Camanho PP, et al. (2007) An engineering solution for mesh size effects in the
simulation of delamination using cohesive zone models. Engineer Fracture Mechanics;74:1665–82,
2007.

13
Ankara International Aerospace Conference