Sei sulla pagina 1di 12

XML Template (2014) [3.3.

20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

JOURNAL OF
COMPOSITE
Article M AT E R I A L S
Journal of Composite Materials
0(0) 112
! The Author(s) 2014
Efficient 3D modeling of damage Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
in composite materials DOI: 10.1177/0021998314525983
jcm.sagepub.com

Alaa El-Din A El-Sisi1, Hesham M El-Emam1, Hani A Salim1 and


Hossam El-Din M Sallam2

Abstract
Due to the complexity of composite material, numerical methods are generally utilized in their analysis and design.
Commercial finite element (FE) codes, such as ANSYS and ABAQUS, allow the implementation of user subroutines in
the program, which provides the advantage of using high meshing and solving technologies besides the improvement of
materials and/or elements models. Nonlinearities arise for many engineering problems, for example, the progressive
damage of a composite element that contains sources of stress concentration or damage localization such as holes, bolts,
and/or flaws causes nonlinear material behavior. In order to simulate this nonlinear behavior, especially in 3D, an accurate
material constitutive model is required. Therefore, the objective of this paper was to simulate the 3D progressive
damage model of composite materials by using simple numerical models. In this paper, ANSYS user subroutine
(USERMAT) was used to simulate the progressive damage behavior of a composite plate containing holes using
simple models. Three different material models were used: ply discount model (PDM), simple progressive damage
model (SPDM) by adding an empirical progressive damage criteria to the PDM, and continuum damage mechanics
model (CDMM). Good agreements were observed between SPDM, CDMM, and published experimental results.
Furthermore, CDMM showed the least dependence on mesh size. Three different damage evolution laws, linear,
quadratic, and degradation laws, were adjusted and tested. It was found that there was no significant difference in the
predicted failure load between these selected laws.

Keywords
3D progressive damage, degradation, composites, finite element, ANSYS, USERMAT

Introduction
With these inputs, either the maximum stress or max-
Laminated ber reinforced polymer (FRP) composite imum strain failure criteria could be used to make rea-
materials are being used in greater frequency mainly sonably accurate predictions of the strength for similar
due to their specic strength and specic stiness tests in the future.14 Under multi-loading conditions,
advantages over conventional materials and since the the prediction of failure becomes more complicated,
ber architecture can be tailored to meet specic design even for homogeneous, isotropic materials. For compo-
considerations. The progressive failure analysis of com- site materials, the tendency for the components of stress
posite laminates is required to predict their mechanical to interact due to the orthotropic material properties
behavior under various loading conditions, and thus and the various failure modes of the constituent
the use of appropriate material constitutive models
plays a crucial role. Continuum damage mechanics
(CDM) provides a physically based concept for simu- 1
Department of Civil and Environmental Engineering, University of
lating damage activation and evolution, which occurs Missouri, Columbia, MO USA
2
in conjugation with stiness reduction.14 Civil Engineering Department, Jazan University, Kingdom of Saudi Arabia
The simplest method for obtaining the stiness and
Corresponding author:
strength properties of a new material is to perform a Hani A Salim, University of Missouri, Columbia, E2509 Lafferre Hall,
direct tensile, compression, and shear tests which will Columbia, MO 65211, USA.
yield the stiness constants and ultimate strengths. Email: SalimH@missouri.edu

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

2 Journal of Composite Materials 0(0)

materials have led to the development of several other from the three techniques, CDMM shows less depen-
classes of failure criteria as in references.512 dence on mesh size and is capable to capture the 3D
Several works were adopted to analyze the progres- localization.
sive failure of composite materials. The simplest tech-
nique to model degradation of stiness is by applying
an instantaneous degradation factor if the failure cri-
teria are met, which is called the ply discount techni-
Numerical modeling
que.1316 While simple to implement, the sudden, Three simple models were used to simulate the progres-
complete failure of the elements does not compare sive damage in FRP; PDM, SPDM, and CDMM. In
well with experimental observations.13 This technique this section, the three models and their implementations
was followed by a group of models at which damage are described in detail.
state variables associated with the residual stiness of a
lamina were used.1719 The stiness terms of an element
Ply discount model
were degraded according to these internal damage state
variables upon satisfying the corresponding failure cri- This model is the simplest way to model the damage
terion. The aforementioned methods of progressive process; it was used in several previous works.1316 In
damage modeling have led to the development of this model, failure criteria Fm given in Table 1 were
more phenomenal and physically based approaches, checked for the integration points for all elements,
including the damage mechanics approach.2026 where m ft1 for ber tension, c1 for ber
The main objective of this paper was to develop and Compression, t2 for matrix tension, c2 for matrix com-
implement a simple technique for modeling of 3D pro- pression, s12 ber matrix shear, and t3 for thickness
gressive damage in a composite element that contains a direction tension}. For all modes, t1, c1, t2, c2, and
source of stress concentration or damage localization s12, Hashins failure criteria were used. Although the
such as holes, bolts, or cracks. To do this, three simple thickness direction tension is not completely equivalent
models were used to simulate the progressive damage in to delaminating, for simplicity the delamination onset
FRP, which are the ply discount model (PDM), the criterion was used for mode t3.14,16,27 For each integra-
simple progressive damage model (SPDM), and the tion point, if the failure criterion is met, an instanta-
continuum damage mechanics model (CDMM). neous degradation factor (km) is applied to the material
Dierent modeling parameters, such as the mesh size properties at that integration point. The degraded
and the damage evolution laws were studied. The material properties (DMP) for each mode are given in
results of all the proposed models were compared Table 1, and in this model km was taken as 104, where
with experimental test results published in the litera-  i is the nominal stress tensor, in which i f1, 2, 3, 12,
ture. Although good correlation could be obtained 23, 13}, and Sm is the lamina strength parameter.

Table 1. Failure criteria.

Failure mode Loading function (Failure Criteria)5,6,14 DMP


   2  2
1 2 12 13
Fiber tension 1  0 Ft1 All
St1 Ss12 Ss12
1
Fiber compression 1 5 0 Fc1 
Sc1
   2   2 
2 3 2  2 23  2 3
Matrix tension 2 3 4 0 Ft2 12 2 13 E2 ,G12
St2 Ss12 S2s23
 2   2 
Matrix compression 2 3 4 0 Fc2 mc1 2 3 mc2 2 3 2 mc3 23  2 3 mc4 12 2
13
Coefficients mc1, mc2, mc3, and mc4 are material constants.
 2  2
1 12
Fiber matrix Shear 2 3 5 0 Fs12 E2 ,G12
St1 Ss12
 2  2  2
3 23 13
Delamination onset Ft3 E3
St3 Ss23 Ss13

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

El-Sisi et al. 3

It is important to note that ANSYS 14.528 proposes other for compression mode. For simply adjusting the
three built-in materials that can simulate the damage damage law, a damage activation function (DAF)
initiation and evolution process; the rst one is the based on the maximum strain criteria was used.
MPDG model. In this model, a degradation factor Linear, quadratic, and exponential damage laws were
can be multiplied directly by the material stiness com- used to describe the damage evolution.
ponent if the damage initiation criterion is met. This
model supports 3D solid layered elements, such as Secant stiffness matrix. Six damage variables were
SOLID185 and SOLID186. The second model is adopted to describe the degradation of material proper-
based on damage mechanics, which supports plane ties of di, where i f1, 2, 3, 4, 5, 6}. Among the six
and shell elements only,28 whereas the third model is variables, three independent variables are found, as
a general composite failure model that is available only the damage for shear stiness component is taken
for explicit analysis. Since this paper is mainly focused from the normal damage component, equation (2)
on 3D domain with implicit analysis, MPDG was stu-
died by developing an element test to investigate this d4 1  1  d1 1  d2
material model. It was found that the behavior of this d5 1  1  d2 1  d3 2
material model is a pure ply discount mechanism. d6 1  1  d1 1  d3

The damage tensor d relates the nominal stress  to


Simple progressive damage model ^ i.e.  I  I : d : ^ where I is
the eective stress ,
This model is similar to the PDM model, except that the identity tensor. The stressstrain relation is derived
instead of applying 104 factors to all the material from Matzenmiller20 complementary free energy den-
properties, a gradually decreased factor between 1 sity of a damaged orthotropic lamina , equation (3)
and 0 was applied. The integration subroutine steps
are listed below: 12 22 12 2
12
1 2
21  d1 E1 21  d2 E2 E1 21  d4 G12
(a) Retrieve the degradation factors from the previous 3
step and apply as factors to the material properties.
(b) Calculate the nominal stresses r C : e, where where Ei , Gi , and i are the Youngs modulus, shear
r, e, and C are the stress, strain, and stiness ten- modulus, and Poissons ratio for the undamaged elastic
sors, respectively. material, respectively. The partial derivative of equa-
(c) Check the failure criteria in Table 1 for rst mode. tion (3) gives the corresponding two-dimensional (2D)
(d) If the failure is detected, slightly increase the degra- compliance tensor H, equation (4)
dation factor k.
2 3
1 21
0
ki1 ki ki 1 6 1  d1 E1 E2 7
6 7
@ 6 12 1 7
6 0 7
" 6 7 H
(e) Loop the steps a, b and c until the failure @ 6 E1 1  d2 E2 7
6 7
disappears. 4 1 5
(f) Repeat all the previous steps for the other failure 0 0
1  d4 G12
modes of Table 1. 4
(g) Save the degradation factors and calculate the sti-
27
ness matrix. From equations (3) and (4), Leone proposed an
expression for that can be available for the 3D case
For a small degradation factor increment, k, the as given in equation (5)
solution error decreases while the solution time
increases. 12 22 32 12
1 2
21  d1 E1 21  d2 E2 21  d3 E3 E1
13 23 12 23
Continuum damage mechanics model 1 3 2 3
E1 E2 21  d4 G12 21  d5 G23
This model is based on the theory of continuum 13
damage mechanics (CDM). In this paper, a simple
21  d6 G13
damage mechanics model was adopted with only 5
three independent damage variables, i.e. di were used
to describe the energy dissipation, for which each vari- Like equation (4), the partial derivative of equation
able takes two values, one for tension mode and the (5) gives the corresponding 3D compliance tensor H.

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

4 Journal of Composite Materials 0(0)

The stiness tensor is the relation between the eective state is one of unloading or neutral loading. If the gra-
stress  and the strain ", and can be obtained from the dient g_^ m is positive, there is damage evolution, and the
inverse of the 3D compliance tensor,  H1 : ". consistency condition has to be satised, i.e.
g_m g_^ m  _^ m 0
Damage activation functions. DAF is the function that
denes the elastic domain,3 and it can be dened as .
Under tension loading, the produced cracks do
gm g^m  ^m  0 6 not aect the compression response; the compression
elastic domains are unaected by ^m , where
where g^ m is the positive loading function (normal) that m ft1 , t2 , t3 g. The following expressions can be
depends on the stress states, and ^m is the updated used to represent the compression damage thresholds
damage threshold function, and m refers to the failure  
mode. The aim of this paper was to develop the sim- ^m max 1, max g^ m 7
plest 3D CDM model and compare it with other simple
models such as PDM and SPDM, and thus a simple where  0 : T, and m fc1 , c2 , c3 g.
loading function was used to describe the damage On the other hand, under compression loading the
initiation process based on the maximum strain cri- produced cracks completely aect the tension response;
teria.14 Table 2 describes the proposed criteria; it can the tension elastic domains are aected by ^m , where
be observed that shear strength was neglected. This was m fc1 , c2 , c3 g (Figure 1). The following expressions
explained in the Secant stiness matrix section since the can be used to represent the tension damage thresholds
degradation of shear stiness components was consid-  
ered to be a function of normal stress damage variables ^m max 1, max g^m , g^ n 8
d1, d2, and d3. This means that the activation of damage
in normal stress components would also activate the where  0 : T, m ft1 , t2 , t3 g, and n fc1 , c2 , c3 g.
damage in the shear stress components. According to Bazants crack band theory,29,30 the
For example, the ber tension failure strain is calcu- damage energy dissipated per unit volume Gm for uni-
lated from t1 SEt11 . The damage development in the axial or shear stress condition is related to the critical
material initiates when the value of g^ m exceeds the strain energy release rate GC along with the character-
initial damage threshold of ^m 1. Further damage istic length LC of the nite element as follows
growth occurs when the value of g^ m in the current GCm
stress state exceeds the value of ^m in the previous load- Gm 9
LC
ing history.25
where m is the failure mode number.
Damage evolution laws. The evolution of the damage This condition was used to adjust the laws for the
threshold values ^m is expressed by the three Kuhn damage variables or the damage evolution laws. In this
Tucker conditions, i.e. _^ m  0, gm  0 and gm  paper, an analytical procedure was performed to iden-
^ m 0,3,22 where _^ m @@T ^m
, in which T is time. tify a suitable degradation law. First, a two-coecient
Neglecting viscous eects, the DAFs of equation (6)
always have to be non-positive.24,25 While gm is negative,
the material response is elastic. When the strain state
St

activates the criterion gm 0, it is necessary to evaluate St: Tension Strength


: Tension Failure Strain
the gradient g_^m .24,25 If the gradient is not positive, the Sc: Compression Strength
Stress

: Compression Failure Strain

Table 2. Loading Functions g^ m :

Damage direction Tension Compression


Strain

"1 "1
Fiber direction 1 g^ t1 For "1 4 0  For "1 5 0
t1 c1
"2 "2
Matrix direction 2 g^ t1 For "2 4 0  For "2 5 0
t2 c2
Sc

"3 "3
Matrix direction 3 g^ t3 For "3 4 0  For "3 5 0
t3 c3
Figure 1. Proposed stressstrain relation for CDMM.

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

El-Sisi et al. 5

material law is proposed for the stressstrain relation propagation is aligned with the mesh lines, the charac-
i "i . To nd these coecients, the following two equa- teristic length must be the square root of the area cor-
tions were used responding to an element integration point.29,30 Unlike
Z ABAQUS,3 in ANSYS USERMAT subroutine28 there
1
GCm is no direct variable to determine the characteristic
m m Sm ; Gm m "i @"i 10
0 Lc length. To solve this problem, an ANSYS function
called get_ElmData128 was used. This function returns
where i is the direction corresponding to the mode m. the volume associated with each integration point VIP.
By solving the produced material law, m m , with the In this study, as 3D elements are used, the cubic root of
proposed damage threshold function ^m , another func- this volume VIP p was
assumed to be the characteristic
tion in terms of dm and ^m instead of i and "i can be length, i.e. Lc 3 VIP .
^ The proposed material
found, where  I  I : d : .
laws are listed in Table 3 along with the corresponding
damage activation laws. For the exponential damage Implementation of user subroutines
law case, Camanho and Matthews19 found the same
damage evolution law. The positive m adjustment
ANSYS User Material Subroutine
parameter can be dened as follow The user material subroutine (USERMAT) is an
ANSYS user-programmable feature.28 Its function is
2S2m Lc to allow users to write their own material constitutive
m 11 equations within a newly developed general material
2Ei Gm  Lc S2m
framework. This subroutine is called at all material
where i is the direction corresponding to the mode m. integration points of these elements during the solution
Therefore, the maximum size for the nite element phase. The input parameters for the USERMAT sub-
for each damage law m is Lc 5 2ESi2Gm .23 If an FEM routine are dened in the modeling stage before the
m
contains elements with LC larger than 2ESi2Gm , the para- solution. In the eld of damage mechanics and compo-
m
site materials modeling, there is very limited research
meter m will be negative and the mesh becomes unfea-
that uses ANSYS USERMAT in the modeling.
sible. This can be avoided byq reducing
the Berbero3 developed a group of 2D models based on
corresponding strength form Sm 5 2ELi Gc m ,29 and in damage mechanics to simulate the matrix failures in
this case the damage variable will take its maximum composites; all the implementations were done using
value of 1 if ^m 4 1.23 USERMAT. Elisa16 used USERMAT in the damage
For a 2D rectangular element with an unknown model of a plate with a hole.
direction of crack propagation, the average
p
of this
expression can be used, Lc 1:12 AIP , where AIP is
the area associated with each integration point. When
Tangent stiffness operator
the crack propagation path can be estimated in To obtain a fast convergence rate of the solution algo-
advance, it is recommended to align the mesh with rithm for the nonlinear problem, it is necessary to cal-
the direction of crack propagation because cracks culate correctly the material tangent constitutive tensor
tend to evolve along the mesh lines. If the crack TM.23 The following equations were used to derive the
tangent stiness operator.3
Table 3. Material degradation law and corresponding damage C:"
evolution law. @C @d @^
_ C_ : " C : "_ "_ : " C : "_
Degradation @d @^ @" 12
 
law Material law Damage law @ @d @^
C : "_ TM : "_
  @d @^ @"
m 1
Linear m B  A"m dm 1 1
2 ^m The previous expression for TM was used in the
 2
2 2m 3 CDMM model only, while for PDM and SPDM
Quadratic  m BA  "m dm 1  1  ^m models, TM was used as equal to C.
9^m m
1
Exponential  m BeA"m dm 1  em 1^m
^m Finite element modeling
Note: A & B are material law constants,  is the adjustment parameter of A 20-node layered solid element28 was used to perform
damage law. all the FE modeling. For each layer, ve integration

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

6 Journal of Composite Materials 0(0)

Table 4. Time elapsed for each analysis and solution warnings and errors.

Time (minutes)

Model NOE 3200 5000 7200 Solution warning and errors

PDM 480 780 960 Negative Pivot


Too large deformation
Unconverged solution
SPDM 1380 2040 2460 Negative Pivot
Too large deformation
Unconverged solution
CDMM 720 1140 1260 Too large deformation
Unconverged solution

Figure 2. Plate with-hole-model for mesh evaluations, dimensions are in mm.

Table 5. Lamina material properties.

Stiffness parameters Strength parameters Fracture toughness


1632
T300/1034-C carbon/epoxy composite
E1 148 GPa St1 2000 MPa Gt1 89.83 N/mm
E2 9.5 GPa Sc1 1500 MPa Gc1 78.27 N/mm
G12 4.5 GPa St1 50 MPa Gt2 0.43 N/mm
 12 0.3 Sc2 150 MPa Gc2 0.76 N/mm
Ss12 100 MPa Gs12 0.46 N/mm
T800H/3633-C carbon/epoxy composite17,23
E1 146.8 GPa St1 1730 MPa Gt1 89.83 N/mm
E2 11.4 GPa Sc1 1370 MPa Gc1 78.27 N/mm
G12 6.1 GPa St1 66.5 MPa Gt2 0.23 N/mm
 12 0.3 Sc2 268.2 MPa Gc2 0.76 N/mm
Ss12 58.2 MPa Gs12 0.46 N/mm

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

El-Sisi et al. 7

LR=57%

LR=63%

LR=71% The red box


represents the site
and the extension of
Section I

the localization

Figure 3. Development of damage localization for CDMM at different load ratio LR values.

points through the layer thickness were used. The thick- the element type used is described in the Finite element
ness was divided to more than one element, and one or modeling section. The laminate is composed of a single
two layers were assigned to each element by using the unidirectional layer with a thickness of 1 mm. One-
layup section option in ANSYS.28 quarter mesh of the model is given in Figure 2, which
shows the dimensions, constrains, and loading direc-
tion. The ber orientation is perpendicular to the load-
Validation and discussion ing direction, so that the matrix is subjected to tension
failure. The material properties are given in Table 5 for
Solution and computational cost
the two laminates T300/1034 and T800H/3633 used in
The solution was performed under windows 64 bit plat- this study.
form, the CPU was Intel Core i7 and the random access Figure 3 shows the development of damage localiza-
memory (RAM) was 6 GB. The solution was per- tion. It can be seen that the crack began at 52% of total
formed using ANSYS version 14.5 default setting for applied load (P 30 MPa) and propagated until it
memory and CPU management.28 Table 4 represents reached the entire width of the specimen, section I.
the time elapsed to reach the failure load. As mentioned Localization is a natural phenomenon, which is often
in the Tangent stiness operator section, there are no dicult to obtain with a numerical model. An
algorithms available for the tangent stiness tensor for objective model was found to capture localization
PDM and SPDM models, which could aect the better, as it has been shown in Figure 3, which is cap-
elapsed time for both of these models. Many errors able of capturing the 3D localization. Figure 4 shows
appeared which led to too much bisection to be able the amount of damage at dierent loading levels. When
to reach the predicted failure load. the load reached 71% of the total applied load,
most of the elements that pass through section I were
completely damaged. Figure 5 shows the relation
Localization and mesh size effect
between the failure load (FL) and the number of ele-
To study the mesh size eect on each material model, a ment (NOE) for each type of damage modeling. It can
simple composite plate-with-hole model was created; be seen that increasing the NOE increases the FL for

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

8 Journal of Composite Materials 0(0)

LR=57% LR=67%

0.77 LR=71%

0.22
0.44
1.00

0.11

0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.88 1.0

Figure 4. Development of damage parameter d2 for CDMM at different load ratio LR values.

1.00 1.00
Sudden Softening

0.75
0.75
Force, kN

PDM 3200
0.50
FL,kN

5000
0.50 SPDM 7200

CDDM 0.25

0.25
0.00
0.00 0.05 0.10 0.15 0.20

0.00 Displacement, mm
3200 5000 7200
Figure 6. Mesh size effect on CDMM.
NOE, #

Figure 5. Failure load (FL) as a function of number of elements


(NOE) for different damage models.

PDM by about 26% and for SPDM by about 19%, was provided for dierent mesh densities from very
while the eect is less pronounced for CDMM by coarse to very rened,31 as shown in Figure 6. It can
about 2.5%. To evaluate objectivity, or mesh depen- be seen that the force vs displacement trends are very
dency behavior, a force (F) vs displacement relation similar in the elastic zone for all mesh sizes. After the

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

El-Sisi et al. 9

1.00 1.00
EXPONENTIAL
LINEAR

0.75 QUADRATIC
0.75

Force, kN
Force, kN

0.50 0.50

CDMM

SPDM 0.25
0.25
ANSYS (MPDG)

PDM
0.00
0.00 0.00 0.02 0.04 0.06 0.08 0.10
0.00 0.05 0.10 0.15 0.20
Displacement, mm
Displacement, mm
Figure 10. Forcedisplacement relation for displacement
Figure 7. Forcedisplacement relation for different models.
loaded model for the three damage activation laws.

1.00
ANSYS (MPDG) K=0.5

ANSYS (MPDG) K=0.995


0.80
Failure Index, Fm

0.60

0.40

0.20
Figure 11. Configuration of the open-hole test specimen.

0.00
0.00 0.50 1.00 1.50 2.00 2.50

Normalized Displacement

Figure 8. Behavior of MPDG elements.

20
1.00 Experimental[17]
EXPONENTIAL Present
16
LINEAR Maimi[23]
0.75 QUADRATIC
12
Force, kN

0.50 8

0.25 4

0
0.00 0 0.4 0.8 1.2 1.6
1.00 1.50 2.00 2.50 3.00
Displacement, mm

Figure 12. Comparison between CDMM numerical and


Figure 9. Damage activation laws. experimental results for the tension test of T300/1034 laminate.

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

10 Journal of Composite Materials 0(0)

elastic zone, a nonlinear behavior began, followed by a the normalized displacement, relative to the failure dis-
sudden softening behavior at which large displacement placement. It can be seen that when the failure criteria
increments were achieved with very small load incre- was met, the stiness was decreased directly with the
ments up to failure. value of the degradation factor.
Figure 7 shows the force vs displacement relation for
the present three models (CDMM, SPDM, and PDM)
in addition to ANSYS (MPDM). The degradation
factor for all the failure modes was 0.995. It can be
Effect of damage evolution law
seen that the results of ANSYS MPDM and PDM Figure 9 shows the relation between dt2 and the damage
are close. Furthermore, it is observed that after the threshold function ^ for the three proposed damage
rst ply damage point at about 1.7 kN, the stiness of evolution laws, while Figure 10 shows the forcedispla-
ANSYS MPDM and PDM was aected sharply more cement relation under displacement control. It can be
than the other models. This can be attributed to the seen from Figure 10 that changing the damage evolu-
degradation factor being almost zero. Figure 8 shows tion law does not have a signicant eect on the load
a relation between the ber tension failure index and displacement relation. Displacement control loading
mechanism can capture the softening behavior of a
damaged material very well. In addition, the model
could still be solved even with partial damage and con-
30
tinue to total failure without instability errors that are
Experimental [16]
25
normally found in the case of force-control loading
PDM
SPDM
mechanism.
20 CDMM
Force, kN

15
Validation of finite element models
10 Two dierent laminates were considered in this section
from the existing literature.17,16,32 The rst laminate
5 was tested in tension, and is composed of plies made
from T300/1034-C carbon/epoxy composite (Table 5)
0 with a stacking sequence of [0/(45)3/903]s. The dimen-
0.00 0.40 0.80 1.20 1.60
sions of the test specimen were length L 203.2 mm,
Displacement, mm
thickness H 2.616 mm, and width W 25.4 mm.17,23
Figure 13. Comparison between numerical and experimental The corresponding diameter of the central hole for
results for the compression test of T800H/3633 laminate. this laminate was D 6.35 mm. The second rectangular

(a) (b)

0.88 0.44
1.00

0.77

0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.88 1.0

3D damage Localization Damage Parameter dt2 Contours

Figure 14. Damage Parameter of T800H/3633 laminate CDMM damage model. (a) 3D damage localization and (b) damage
parameter dt2 contours.

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

El-Sisi et al. 11

laminate was tested in compression, and is composed of References


T800H/3633 composite plies (Table 5), with length, 1. Ochoa O and Reddy J. Finite element analysis of compo-
width, and thickness of 118 mm, 38 mm, and 1.1 mm, site laminates (Solid mechanics and its applications). New
respectively.16,32 Figure 11 shows the dimensions of the York, USA: Springer, 1992.
second specimen. 2. Matthews F, Davies G, Hitchings D, et al. Finite element
Figure 12 shows a comparison between the present modelling of composite materials and structures. CRC
numerical results, experimental results, and numerical Press, 2000.
results proposed by Maimi et al.23 for the T1300/1034 3. Barbero E. Finite element analysis of composite material
laminate. The failure stress of the present model was by using ABAQUS. Cambridge, England: CRC Press,
2013.
222 kN compared to the experimentally measured value
4. Barbero E. Introduction to composite materials design,
of 235.8 kN, which represents a good correlation
2nd ed. Cambridge, England: CRC Press, 2010.
between the present model and experiment with an 5. Hashin Z and Rotem A. A fatigue criterion for fiber
error of about 5.5%. Maimi et al.22,23 closely over- reinforced materials. J Compos Mater 1973; 7: 448464.
estimated the experimental load with about 11%. The 6. Hashin Z. Failure criteria for unidirectional fiber compo-
present model does not include the eect of plasticity sites. J Appl Mech 1980; 47: 329334.
which might enhance the results if it were included, 7. Soden PD, Hinton MJ and Kaddour AS. A comparison
since the plasticity eect enhances the solution of the predictive capabilities of current failure theories for
convergence.25 composite laminates. Compos Sci Technol 1998; 58:
For the compression test on T800H/3633 laminate, 12251254.
the experimental result obtained by Suemasu16,32 was 8. Puck A and Hurmann H. Failure analysis of frp lami-
compared with the numerical results obtained from the nates by means of physically based phenomenological
three numerical models as shown in Figure 13. It can be models. Compos Sci Technol 1998; 58: 10451067.
9. Hinton M, Kaddour A and Soden P. A comparison of
seen that CDMM model gives the best solution of the
the predictive capabilities of current failure theories for
three damage models evaluated. PDM underestimates composite laminates. 2002. Compos Sci Technol 1998; 62:
the peak value due to a solution convergence problem, 17251797.
whereas SPDM overestimates the peak response. 10. Daniel I. Failure of composite materials. Strain 2007; 43:
Improved results might be obtained with a ner mesh, 412.
because both PDM and SPDM depend on the size of 11. Tay T, Liu G and Tan V. Progressive failure analysis of
the mesh. Figure 14 shows the deformed shape and the composites. J Compos Mater 2008; 42: 19211966.
distribution of damage parameter dt2 for the T800H/ 12. Catalanotti G, Camanho P and Marques A. Three-
3633 laminate model at failure load. dimensional failure criteria for fiber-reinforced laminates.
Compos Struct 2013; 95: 6379.
13. Hahn H and Tsai S. On the behavior of composite lami-
nates after initial failures. J Compos Mater 1974; 8:
Conclusions 288305.
14. Tserpes K, Labeas G, Papanikos P, et al. Strength pre-
In this paper, a simple, ecient 3D progressive
diction of bolted joints in graphite/epoxy composite lami-
damage model for composite materials was developed nates. Compos: Part B 2002; 33: 521529.
using eective numerical models. A model based on ply 15. LIN W. Nonlinear analysis of fiber-reinforced composite
discount technique, which is the simplest model, was laminates subjected to uniaxial tensile load. J Compos
rst developed followed by another simple model simi- Mater 2002; 36: 14291450.
lar to the ply discount technique combined with an 16. Elisa P. Progressive failure analysis of composite struc-
empirical progressive damage criterion to each of the tures using a constitutive material model (USERMAT)
plies. A 3D progressive damage model using CDM was developed and implemented in ANSYS. Appl Compo
also developed. The developed models were compared Mater 2012; 19: 657668.
with existing models and existing experimental data. 17. Tan SC. A progressive failure model for composite lami-
CDDM provided the best comparison of all models nates containing openings. J Compos Mater 1991; 25:
developed. The eect of mesh size on each of the 556577.
18. Nuismer R and Tan S. Constitutive relations of a
models presented was studied, and it was shown that
cracked composite lamina. J Compos Mater 1988; 22:
of the three models CDMM was the least sensitive to
306321.
mesh size. The eect of the proposed progressive 19. Camanho P and Matthews F. A progressive damage
damage criteria using three damage activation laws, model for mechanically fastened joints in composite lami-
linear, quadratic, and exponential degradation laws nates. J Compos Mater 1999; 33: 22482280.
was investigated in this paper. It was shown that the 20. Matzenmiller A, Lubliner J and Taylor RL. A constitu-
degradation laws had little eect on the displacement tive model for anisotropic damage in fiber-composites.
stress response. Mech Mater 1995; 20: 125152.

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014


XML Template (2014) [3.3.20141:59pm] [112]
//blrnas3/cenpro/ApplicationFiles/Journals/SAGE/3B2/JCMJ/Vol00000/140030/APPFile/SG-JCMJ140030.3d (JCM) [PREPRINTER stage]

12 Journal of Composite Materials 0(0)

21. Lonetti P, Barbero E, Zinno R, et al. Inter-laminar 31. Oliver J. Modeling strong discontinuities in solid
damage model for polymer matrix composites. mechanics via strain softening constitutive equations.
J Compos Mater 2003; 37: 14851504. Part I: fundamentals. Int J Num Meth Eng 1996; 39:
22. Maim P, Camanho P, Mayugo J, et al. A continuum 35753600.
damage model for composite laminates: Part I consti- 32. Suemasu H, Takahashi H and Ishikawa T. On failure
tutive model. Mech Mater 2007; 39: 897908. mechanisms of composite laminates with an open hole
23. Maim P, Camanho P, Mayugo J, et al. A continuum subjected to compressive load. Compos Sci Technol
damage model for composite laminates: Part II 2006; 66: 634641.
Computational implementation and validation. Mech
Mater 2007; 39: 909919. Appendix
24. Chen J, Morozov E and Shanka K. Plastic damage model
for progressive failure analysis of composite structures. Notation
In: 18th international conference on composite materials,
m Failure Mode ft1 for fiber tension, c1
Edinburgh, Scotland, 27 July31 July 2009.
for fiber compression, t2 for matrix ten-
25. Chen J, Morozov E and Shankar K. A combined elasto-
plastic damage model for progressive failure analysis of
sion, c2 for matrix compression, s12
composite materials and structures. Compos Struct 2012; fiber matrix shear, and t3 for thickness
94: 34783489. direction tension, i.e. delamination}
26. Kim E, Rim M, Lee I, et al. Composite damage model ri or r Nominal stress tensor, in which i f1,
based on continuum damage mechanics and low velocity 2, 3, 12, 23, 13}
impact analysis of composite plates. Compos Struct 2013; r^ Effective stress tensor, in which i f1,
95: 123134. 2, 3, 12, 23, 13}
27. Leone F. Full-scale testing and progressive damage model- di Damage variable, in which i is the
ing of sandwich composite aircraft fuselage structure. PhD direction f1, 2, 3, 4, 5, 6}
Thesis, Drexel University, 2010. e or ei Strain tensor, in which i f1, 2, 3, 12,
28. ANSYS Release 14.5 Documentation, ANSYS Inc., 23, 13}
2012. Ei Youngs Modulus where i f1, 2, 3}
29. Bazant Z and Oh B. Crack band theory for fracture of
Gi Shear Modulus where i f12, 23, 13}
concrete. Mater Struct 1983; 16: 155177.
ti Poisson ratio where i f12, 23, 13, 21,
30. Turon A, Davila C and Camanho P. An engineering
solution for mesh size effects in the simulation of delami- 21, 31}
nation using cohesive zone models. Eng Fract Mech 2007; s Time period
74(10): 16651682. T Time value

Downloaded from jcm.sagepub.com at UNIVERSITE DE SHERBROOKE on December 18, 2014

Potrebbero piacerti anche