Accepted Manuscript

Progressive damage analysis of a rate-dependent hybrid composite beam

Giovanni Belingardi, Hadi Mehdipour, Enrico Mangino, Brunetto Martorana

PII: S0263-8223(16)31194-1
DOI: http://dx.doi.org/10.1016/j.compstruct.2016.07.055
Reference: COST 7655

To appear in: Composite Structures

Received Date: 14 July 2016

Accepted Date: 21 July 2016

Please cite this article as: Belingardi, G., Mehdipour, H., Mangino, E., Martorana, B., Progressive damage analysis
of a rate-dependent hybrid composite beam, Composite Structures (2016), doi: http://dx.doi.org/10.1016/
j.compstruct.2016.07.055

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Progressive damage analysis of a rate-dependent hybrid composite beam

Giovanni Belingardi1, Hadi Mehdipour1 , Enrico Mangino2 , Brunetto Martorana2

1
Department of Mechanical and Aerospace Engineering, Politecnico di Torino, Corso Duca degli Abruzzi- 24,
Torino, Italy
2
Group Material Lab., Centro Ricerche Fiat, strada Torino 50, 10043 Orbassano (TO), Italy

Abstract

A hybrid beam combined of a molded woven fabric beam and over injected or compressive molded short fiber
composite material, are manufactured and three point bending tests with different loading rates and two different
boundary conditions are performed. Since the experimental results demonstrate a rate dependent behavior of the
beam, then to predict its mechanical behavior under different loading velocities two strain rate sensitive
progressive damage models have been developed based on Continuum Damage Mechanics approach. For
internal reinforcing constituent of the beam which dominated by matrix, the matrix failure index of Hashin-
Rotem’s failure criterion was chosen and for outer part which governed by fibers the maximum stress theory is
considered as damage initiation indicator. For both models Lapczyk-Hurtado’s approach is followed to capture
damage propagation. To account for rate-dependent material properties, where is needed, equations derived from
a logarithmic formula, are utilized. The models is applied on the commercial finite element software of
ABAQUS using user-written subroutine based on explicit numerical method. The numerical results are verified
by the bending tests and it is shown that the numerical results of the proposed models have shown better
agreement with experimental results of compressive molding technology than back injected one.

Keywords: Continuum damage mechanics, Progressive damage, Fabric thermoplastic sheet, Short fiber
composite, ABAQUS subroutine

Introduction

One of the main obstacles that vehicle companies are facing is the reduction of the vehicle weight without
sacrificing their safety, reliability, durability and riding performances. The substitution of new materials and
modern manufacturing technologies, such as hybrid composite materials, for classic approaches and materials,
such as fully made metallic components, is one of the main trends to achieve relevant weight reduction.
Introduction of complicated hybrid composite materials leads us to going through relatively immature field of
mechanics of material. To deal with these new generations of components and materials, researchers have been
searching for accurate sorts of analytical and numerical procedures to predict their mechanical responses under
different loading and boundary conditions.
Continuum Damage Mechanics (CDM) approach deals with material damage based on stiffness component
degradation[1].Also, CDM method which incorporates damage in the material response functions might be
introduced as a set of internal vector field variables [2]. To define a damage model based on CDM approach, a
constitutive model including damage variables must be presented. Generally, a constitutive model relates stresses
to strains components using a fourth order stiffness tensor. To incorporate damage variables in constitutive
equations, a reasonable way is redefining the stiffness tensor by implementing damage variables. The redefined
stiffness tensor called degraded stiffness tensor. To achieve a physical base damage model and extract the
constitutive equations, thermodynamic laws are considered which needs to define Helmholtz free energy [3]. To
detect a specific damage mode activation a failure criterion, which represent the damage surface in stress or
strain space, must be defined. The damage envelope might be in separated form such as Hashin-Rotem criterion
for unidirectional composite plies[4]. In this type of failure criterion for each modes of damage, in longitudinal
and transverse direction and under compression and tension, damage surfaces defined by different formulas.
Later, Puck and Schurman based on Hashin and Rotem’s hypothesis formulated another separated form of
failure criterion which deals with inter-fiber failures and has a capability of detection of crack direction [5]. Even
though based on World Wide Failure Exercise (WWFE) [6] Puck’s failure theory showed the best agreement
with experimental results, but there are some non-physically based parameters in this theory which make its
application hard for users. To overcome this difficulty, Davila and Camanho developed a failure criterion based
on Hashin and Puck theories, where no nonphysical parameters are needed[7]. Formerly, Chang and Chang
suggested a separated form of failure criterion for notched laminated composites with consideration of material
nonlinearity [8]. In addition to the separated type of failure criteria, one can implement an integrated type in
CDM which are mostly the generalized form of Hill’s criterion for orthotropic materials [9].
To deal with a progressive damage modeling, in addition to a failure criterion, a damage propagation law must
be defined to govern the growth of an existing damage. Lemaitre et al. formulated an anisotropic damage law in
the framework of equivalent strain which the evolution of damage variables governed by plastic strains [10].
Maimi et. al. suggested a general exponential form of damage law which used to represent the cohesive response
of all the failure modes of the ply, except for the longitudinal tension damage [11]. For simulation of
delamination in composite materials Camanho and Davila proposed a new decohesion element with mix-mode
capability and for applying the softening law, they used a displacement-based damage parameter to track the
damage state of interface [12].
Finding a proper failure criterion to active the initial damage, and a constitutive equation to govern stress-strain
relation before and after damage, and also defining sufficient numbers of damage variables, are just the
beginning of adamage problem modelling difficulties. Mesh sensitivity and convergence come out during
numerical solution of damage problem which severely affect the results. Lapczyk and Hurtado proposed an
orthotropic damage model for pre and post damage in brittle laminated composites [13]. In this model LaRC04
failure criterion which proposed by Davila and Camanho [7] is used to predict the onset of damage. In LarC04 is
same as Hashin’s failure criterion, but stresses replaced by effective stress components. Also for damage
evolution they generalized Camanho and Davila method for delamination [12] and used specified fracture
energy. Also to alleviate mesh dependency in numerical implementation they used crack band model which had
been already proposed by Bazant and Oh[14] and to overcome convergence difficulties by following the Duvaut
and Lions regularization method [15], instead of direct use of damage variables, a set of regularized form of
damage variables are utilized .
On the other hand, mechanical properties of reinforced polymeric composite materials including unidirectional,
short fiber and woven reinforcement are drastically rate-dependent [16].In unidirectional composite materials the
module of elasticity in fiber direction is not rate-dependent as much as material tensile strengths at fiber direction
[17]. Also, it has been shown that the module of elasticity and material strength in transverse directions are rate-
dependent[18]. Also, it has been shown that woven fabric composites with different pattern, laminated sequence
and material are rate dependent [19, 20].
Basically modelling of rate-dependent behavior of composite material divided into two major categories:
macromechanical and micromechanical approaches. To model the rate-dependent behavior of polymer matrix
Weeks and Sun followed a macro mechanical approach and developed a rate dependent constitutive equation and
also used a plastic potential function to determine inelastic part of total strains[21]. Later, Thiruppukuzhi and
Sun modified the previous model and incorporated the rate-dependent behavior of material directly into
constitutive equations for unidirectional and woven fabric composites[22]. Aidun and Addesio considered
micromechanical approaches to model rate-dependent response of composite materials by developing a
nonlinear elastic constitutive model for matrix constituent and implementing it through a micromechanics model
to compute nonlinear response of the composite [23]. Goldberg modified Ramaswamy-Stouffer viscoplastic
constitutive equation for metal to model the rate-dependent inelastic deformation of ductile polymers [24].
Tabiei et. al. developed a nonlinear finite element method to model dynamic response of fibrous composites
using Goldberg model[25] .
In this study the back-injection and compression molding those are some of the newest versions of hybrid
manufacturing techniques, has been considered. In particular, an outer shell, with a typical open channel section
made of a molded thermoplastic woven fabric, has been completed by means of a back-injection or compressive
molding of short fiber composite. Due to the complex nature of this type of components, the simulation of its
mechanical behavior under various loadings and boundary conditions, is a challenging issue. The most
reasonable approach for avoiding expensive whole component tests is to find more reliable numerical methods,
but this approach has its own difficulties.
In this study composite beams using two simultaneous manufacturing procedures of hydroforming and injection
or compression molding have been produced and their responses under different bending rates have been
studied. Considering that the internal part made of short fiber composite material, it is supposed that is totally
governed by matrix and the open section is made of a fabric composite which is supposed to be dominated by
fabrics. Therefore, in this research to predict damage initiation of the internal part, the second part of Hashin-
Rotem’s which is dealing with matrix damage initiation has been chosen. For the external part with considering a
same reason, maximum stress theory is chosen which is dealing with fibers damage failure. Based on two failure
criteria two different kinds of damage propagation models have been developed and implemented in ABAQUS
using subroutines so-called VUMAT. Also, since the components have been shown rate dependent behavior,
another model is suggested and developed to dealing with this kind of behavior and implemented in another
VUMAT to investigate the rate-dependent behavior of the component. It should be noted that this rate
dependency is just assigned to the internal part which is completely dominated by matrix. The model is
developed assuming that plastic strain in these composite material are negligible, and also strain and stress are
evenly distributed through an element. CDM approach has been followed to correlate damage variables coming
from Lapczyk-Hurtado’s model[13] and their effects into stiffness matrix and consequently stress-strain
relationship. The rate dependence effect incorporated into modules of elasticity and strength of reinforcing
constituent, using a logarithmic function and following Daniel and Wei et al.’s works [26, 27]. Also, mesh
dependency problem has been alleviated following Lapczyk-Hurtado’s model which strains are replaced by
characteristic length and displacements in damage variables. Convergence problem has been solved following
Duvaut and Lions regularization method [15] which instead of direct use of damage variables, a set of
regularized form of damage variables are utilized. The accuracy of the newly developed model have been
evaluated using performed bending experiments which demonstrate that the presented model is accurate enough
to simulate progressive damage of 2-D orthotropic materials.

Manufacturing process and material properties

Hybrid solution for producing variety of components in automotive industries has been growing in recent years.
Injected plastics over a molded sheet is one of the manufacturing process which intelligently proposed to use the
formability of fabric composite sheets and its appropriate quality as an exterior part, and short fibers composite
with capability of being injected to form an arbitrary shape as a reinforcement part [28].
In the considered case, molded part is made of an organo-sheet which consists of a thermoplastic (Polyamide 66)
as a matrix and 63% plane weave glass fiber as reinforcement. Injected part is made of the combination of
Polyamide 66 as its matrix and 30% of short glass fiber as reinforcement.
The manufacturing process begins with forming of the organo-sheet which must be heated up to 300°C.
Afterwards, it is inserted into the mold and mold is closed. In order to achieve a more uniform heat distribution
all over the sheet, this process is performed by infrared technology. Molding process does not take more than 4
seconds. After completion of organo-sheet shaping process, the injection process is performed with two
distinguished phases, without extracting the part from the mold.
In the considered case, at first, the melted short fiber composite is injected over the molded sheet at 320°C
temperature within 3 seconds. Then, to compensate the volume reduction during the cooling phase, injection
process must be continued which the so-called packing phase. The volume reduction is the result of mold
cooling, at the mean temperature of 70°C and it approximately equals 5% of total injected material volume. The
packing phase takes about 10 seconds and then the final cooling phase, which lasts 60 second, is started. After
cooling phase, by which the mean temperature of the injected material is reduced to about 90°C, the final
product is ejected in 3 seconds. It can be noted that a mean temperature of 90°C seems to be quite enough to
prevent the final shape of the component from distortion and deformation after ejection. The whole production
process is continuous, because during the cooling phase of the process there is enough time to refill the injector
by short fiber composite granules and melting them. The final shape of the considered component shown in
figure 1.

Figure 1: Hybrid component: cross section and upper view

The mechanical properties of constituents of the hybrid component are presented in Table 1, separately.

are Young and shear elastic module, respectively.
is Poisson ratio and X and Y represent strengths of material
Subscripts 1, 2 and 3 refer to longitudinal, transverse and out of plane material directions, respectively. E and G

in the first and second in-plane orthotropic directions, respectively. S is shear strength of the material.
Subscripts C and T refer to material loading condition of compression and tension, respectively.
Table 1: material properties of molded and injected constituents of the hybrid component; all modules of
elasticity presented in GPa and strengths in MPa

Property    

  


  .

.

 Short fiber 8.5 6 2.1 2.3 0.29 0.3 125 200 125 200 50 0.031 0.04
Plane weave
24.3 25.2 1.3 1.2 0.25 0.3 370 280 430 280 58 - -
fabric

According to Table 1, substructure orientation relatively affects material properties of the injected constituent,
while the organo-sheet material properties are rather less affected. In addition, according to the images taken by
electron microscope (see figure 2), to assume a decision about the fibers directions, except for the injection gates
area, is a really controversial issue. For having the orientation of glass fibers during injection process in
industrial areas, researchers usually use numerical simulation, by means of some commercial software such as
Moldflow®, but in this study the substructure of injected molded part has been evaluated through electron
microscope images in order to identify the reinforcement orientation of the most important zones. Finally, during
FEM modelling the geometry of the model has been subdivided in some sub segments and for each segments a
mean value of rotation of the orthotropy axis has been assigned.

(a) (b)
Figure 2: short fiber orientation in injected constituent; picture (a) shows the material orientation at the
injection gate area and picture (b) at one of the randomly selected zones which both taken from damaged
zones

Observation of the pictures in Figure 3 is pointing out that in addition to short fibers orientation, also their
density distribution should be considered on a micro-scale. However, due to the fact that this research deals with
macro mechanical approach, average values of the mechanical properties, neglecting the local variability due to
this fiber density distribution, can be accepted. In any case these microstructural images help to have a vision
close to reality of material principal direction that are of interest for FEM simulation. Besides, based on the
images that have been taken from undamaged injected constitute surface, one can suppose a random distribution
and orientation of glass fibers.

(a) (b)
Figure 3: Short fiber orientation in injected constituent; pictures (a) and (b) show glass fibers distribution
and orientation on the surface of two randomly chosen areas

Continuum damage mechanics (CDM)

The user-defined constitutive equations are based on Continuum Damage Mechanics (CDM) principals which
includes degradation of the material stiffness matrix as proposed by Matzenmiller et al [1]. In this implemented
model, damage initiation is detected based on Hashin and Rothem [4] failure criterion. It is well known that the
Hashin and Rothem failure criterion has been developed for unidirectional reinforced laminates. Damage

release rates, and strains are replaced by characteristic length L , and associated displacements to alleviate mesh
propagation is applied based Lapczyk-Hurtado’s model which damage propagation is governed by energy

sensitivity problem[13].

Continuum Damage Mechanics

It is assumed that before damage initiation the composite material behaves as a fully elastic orthotropic
continuum with a linear response:

 = " # $# (1)

Which for a plane stress conditions and for an orthotropic material gives:


  0

"= )
   0 +


0 0 
%&'( &('
(2)

After damage initiation, virgin stiffness matrix must be updated by taking into account the consequent
degradation of the material characteristics. The degraded stiffness matrix for 2-D problems can be written as
follows:

.1 − 12 3 .1 − 12 341 − 15 6
  0

" = -.1 − 12 341 − 15 6
  41 − 15 6 8


,
0 (3)
0 0 41 − 17 6

 64  6  64  6
where 9 = 1 − .1 − 12 3.1 − 12 341 − 15 1 − 15

 , 17 = 1 − .1 − 12 3.1 − 12 341 − 15 1 − 15
and 12 and 15 are giving the measure of the material degradation and therefore are describing the current values
of fiber (subscript f) and matrix (subscript m) damages, respectively.

Hashin-Rotem failure criterion

Damage initiation must be defined in one of the stress or strain state called failure criterion. Variety of failure
criteria for composite materials have been proposed over the years. Some of them are more considered in real
applications such as Maximum stress, Maximum strain, Hashin-Rotem, Chang-Chang, Puck and LaRC [4, 5, 7,
8]. But, as Hinton et al. mentioned in the conclusive evaluation of the results obtained with the World-Wide
Failure Exercise (WWFE) [6], despite recent improvements of failure criteria for composite material, still the
majority of researchers insist in using classic form such as Maximum stress or Maximum strain theory, Tsai-Wu
and Hashin-Rotem criteria. In this paper, Hashin-Rotem criterion is used, by considering that the stress space is
effective one. This criterion is a separated form criterion which presents a separated criterion for each modes of
failure including matrix and fiber failure in tension and compression as is reported Table 3.
Table 2: Maximum stress theory; failure criterion for fabric part of the component; ;<= are effective stress
components, >? >@ , A? and A@ are strengths of material. X and Y denote warp and woof directions,
respectively. While T and C denote tension and compression, respectively.

Failure criterion for fabric

Failure mode Maximum stress theory

 
Tensile in warp Direction 4 ≥ 06 CD

=E G =1
X
 
Compressive in warp Direction4 < 06 CD

=E G =1
X
  
Tensile in woof Direction 4 ≥ 06 CT

=E G =1
Y
 
Compressive in woof Direction 4 < 06 CT

=E G =1
Y

Table 3: Maximum stress theory; failure criterion for external part of the component; ;<= are effective
stress tensor components, >? >@ , A? and A@ are strengths of material. X and Y denote two perpendicular
directions of the material. While T and C denote tension and compression, respectively. WXY and WYX are
longitudinal and transverse shear strengths .

Failure criterion for internal part

Failure mode 2-D Hashin-Rotem (for matrix)

   
Tensile Matrix in X direction 4 ≥ 06 CD
Z
=E G +E G =1
X 
  X 
  
Compressive Matrix in X direction 4 < 06 CD
Z
=E

G + ]E

G − 1^ +E G =1
2 2 |X | 
   
Tensile Matrix in Y direction 4 ≥ 06 CT =E G +E G =1
Z  

 
 

  
Compressive Matrix in Y direction 4 < 06 CT
Z
=E G + ]E G − 1^ +E G =1
2 2 |
| 

Damage evolution law

Once damage initiation criterion is satisfied in any of the damage modes, damage evolution law is activated and
any further loading will be caused degradation of stiffness matrix. The propagation of damaged governed by
damage evolutions law which is based on damage variables varying between zero and one. Zero amount of
damage variables indicates virgin status of material, while amount of one notes a fully damaged material point.
There are two different approach to represent a damage law to control damage evolution: ply-discount method
and damage laws based on energy release rates. Ply-discount method is the simplest one and degrade material
just are taken a value zero or one which zero one shows virgin material and value of one indicates again fully
damaged material. But, from numerical point of view using value of one for fully damaged material point is not
reasonable because sometimes leads to a stiffness matrix with a complete zero row or columns. So, to implement
ply-discount method to capture damage evolution instead of value of one, one has to set a value such as 0.99 to
avoid numerical instability. The second one is founded on energy release rate concept and leads a gradually
degradation of material properties. This gradually release of energy could be governed by any equation,
providing passes through points A and B and becomes strictly descending according to Figure 4. A strain-
softening equation can be have physical meaning or can be proposed considering to a non-physical data fitting
methods. Mainly this strain-softening equation is chosen linear as presented in [13, 29, 30] or could be
exponential such as proposed in [31, 32]. In this study the linear damage law proposed in [13] which is governed
by energy release rates, equivalent displacements and characteristic lengths of associated elements.
Based on damage model proposed by Lapczyk-Hurtado which also used in this paper for a specific damage
mode, a linear strain-softening leads [27]:

`, = `, c`,

b 2

(4)

Which `, is maximum achievable equivalent stress before damage initiation and c`, is maximum equivalent
b 2

displacement when the material point is failed, while index I indicates a specific damage mode .For an element

stress-equivalent displacement is /e (the segment line of OA in Figure 4), and after damage will be 4/
which is subjected to damage and the damage can be total or partial, before damage the slope of equivalent

e 641 − 16 (the segment line of OB in Figure 4). Referring to Figure 4, the area of the OAB triangle represents
the amount of energy that has been released during damage process while the area of the OBC triangle represents
the remaining deformation energy that can be released until the material is completely damaged.

Eq
uiv

A
ale b
nt 

`
str B
ess

C
c
0
b
c Equivalent displacement
O 2

Figure 4: Linear Damage evolution law

In this model two main variables are equivalent displacement and equivalent stresses calculated regards to the
equations presented in Table 4. The characteristic length is related to the element geometry and formulation; it
can be calculated by means of many different methods [13, 14]. In this paper the calculation of the characteristic
length (e ) is left to ABAQUS’s default which assumed to be equal to the length of a line across the element for
a first-order element and half of that length for a second-order element. If one intends to apply a user-defined
formula to calculate the characteristic length it would be possible by coding in a subroutine called
VUCHARLENGTH. The damage for any of damage modes are calculated as follows:
j l
fg,hi kfhi %fg,hi m
1` = ; c`, ≤ c ≤ c`,
b 2
j l m
fhi kfg,hi %fg,hi
(5)

and c`, describe equivalent displacement for 1` = 0

b 2
Which I indicate different modes of damage. while c`,
(i.e. no damage) and 1` = 1, (I.e. material complete failure), respectively. Equation (5) is a function which varies

irreversible, function (5) must be strictly increasing in its domain. The value of c`, can be calculated based on
between zero and one in accordance with the δeq interval of variation. According to the fact that the damage is
2

the critical energy release rate concept as follows:

c`, = $ e =
2 2 pg,q
l
rg,hi
(6)

where `,
b
is the equivalent stress value at the damage initiation state as mentioned before. Considering the
Hashin-Rothem failure criterion, Table 5 summarizes the relationships to achieve equivalent stress and
displacement in the different modes of damage.

〈 〉indicates Macaulay operator which is defined as 〈u〉 = 4u + |u|6; ∀u ∈ ℝ

Table 4: equivalent values of displacements and stresses for different damage modes of fabric; the symbol
X
Y

Equivalent stresses and displacements for fabric outer shell

Failure mode y z ; z
e 4〈 〉〈$ 〉6
Tensile in warp Direction 4 ≥ 06 e 〈$ 〉
c
D

4L{ − 〉〈−$ 〉6

〈
Compressive in warp Direction4 < 06 e 〈−$ 〉
c
D

e 4〈 〉〈$ 〉6
Tensile in woof Direction 4 ≥ 06 e 〈$ 〉
c
T
 e 4〈− 〉〈−$ 〉6
Compressive in woof Direction 4 < 06 e 〈−$ 〉
c
T

constituent; the symbol 〈 〉indicates Macaulay operator which is defined as 〈u〉 = Y 4u + |u|6; ∀u ∈ ℝ
Table 5: equivalent values of displacements and stresses for different damage modes of internal
X

Equivalent stresses and displacements for internal part

y z ; z
e 4〈 〉〈$ 〉 +  $ 6
Failure mode

Tensile Matrix in X direction 4 ≥ 06 e |〈$ 〉 + $


c
D

4L{ 〈− 〉〈−$ 〉6

Compressive Matrix in X direction 4 < 06 e 〈−$ 〉
c
D

e  $ 〉 +  $ 6

4〈 〉〈
Tensile Matrix in Y direction 4 ≥ 06 e |〈$ 〉 + $

c
T

e 4〈− 〉〈−$ 〉 +  $ 6

Compressive Matrix in Y direction 4 < 06 e |〈−$ 〉 + $

c
T

Equivalent values of stress and displacement at the damage initiation point might be calculated as follows:

c`,
b
= c`, }` (7)

`,
b
= `, }` (8)

where the scaling factors }` are properly presented for different modes of damage in Table 6.

Table 6: scaling factors for different damage modes [13]

Failure mode Scaling Factors (for fabric outer shell)

Tensile in warp Direction 4 ≥ 06 1~|CD



Compressive in warp Direction4 < 06 1~|CD



Tensile in woof Direction 4 ≥ 06 1~|CT



Compressive in woof Direction 4 < 06 1~|CT



Failure mode Scaling Factors (for short fiber internal part)

Tensile Matrix in X direction 4 ≥ 06 1~|CD

Z

Compressive Matrix in X direction 4 < 06 1~|CD

Z

Tensile Matrix in Y direction 4 ≥ 06 1~|CT

Z

Compressive Matrix in Y direction 4 < 06 1~|CT

Z
Rate-dependent material behaviour
According to [33, 34] and [17, 35] strain rate affects material properties of glass-reinforced composite materials
including module of elasticity, strengths and toughness energies. In this study following Daniel’s and later Wei
et. al. works [27, 36] , and for keeping the numerical implementation of the models simple, the effects of strain-
rate have been applied od modules of elasticity and strengths of the material and rate dependent toughness
energies are ignored. Following equations:

|$„ |

 = ,7 ]2 '' €‚ ƒ

… + 1^
$„,7


 = ,7 †2 (( €‚ E‡„ G + 1‰ 496
 |‡„ ((|
((,iˆ

|$„ |

 = ,7 ]2 '( €‚ ƒ … + 1^
p
$„,7

condition. $„,‹Œ, $„,‹Œ and $„,‹Œ are reference values of strain rates in longitudinal normal, transverse normal
The abbreviation of ”qs” mentions to the amount of a specific value of a material constant under quasi-static

and in-plane shear statuses. As Wei et. al.[27] to assure stability of the damage propagation model, it is assumed
that all strain-rate dependence material properties have a uniform type of governing equations. Therefore, for
rate dependent strengths can be written as follows:

|$„ |

 =  ]2  €‚ ƒ … + 1^
7 D
$„,7


 =  †2 q €‚ E G + 1‰
7 D |‡„ '' |
‡„ '',iˆ
(10)

|$„ |


 =
 ]2  €‚ ƒ …+ 1^
7 T
$„,7

|$„ |


=
]2 q €‚ ƒ … + 1^
7 T
$„,7

|$„ |

 =  ]2 '( €‚ ƒ … + 1^
7 
$„,7

|$„ |

 =  ]2 Ž €‚ ƒ … + 1^
7 
$„,7

The constants 2 and all which shown by “qs” indices could vary even case by case, but here it is assumed that
these constants become equals for all the cases. For an instance, here in this study for calculation of  ,  ,

and etc. a same value of 2 are considered. To determine the constant of 2 for each different condition one
has to perform different sorts of basic material property characterization tests with some various loading speeds
to compare that specific material constant to its quasi-static condition. Then by data fitting procedure with each
set of experimental result the related rate-dependency constant will be determined. In this study the above
mentioned rate-dependent material properties have been applied in the subroutines, but since the rate dependency

beam, then in order to determine the constant of 2 the following equations have been suggested for this specific
experiments just have been performed for whole component and have not conducted for constituents of the

boundary conditions:

G + 1‰ 4116
’“”•–—˜™
.
..
= .
‘..
†.

€‚ E
’iˆ
. and .
.. ‘..
indicate the equivalent stiffness of the beam subjected to 3-point bending (Case1) under

graphs of quasi-static and loading velocity of 127 mm/sec has been considered as . and .
dynamic and quasi – static loadings, respectively. In this study, the approximate slope of the force-displacement
‘.. ..

respectively. .
,


special boundary and loading conditions, and š›5 and š7 indicate puncher velocity for dynamic and quasi-
shows an effective coefficient of m for dealing with total stiffness of the beam under this

static cases, respectively. A similar formula is suggested to determine rate-dependent strength of the component
as follows:

’“”•–—˜™
. = . †. €‚ E G + 1‰ 4126
. ‘.. 
’iˆ

. and .

.. ‘..
indicate the equivalent strengths of the internal constituent material without considering

material direction, for dynamic and quasi–static loadings, respectively. . demonstrates an effective
coefficient of m to shift the results of material strength under quasi-static loading to dynamic level. In this study,
since the Eq. (12) which suggests a direct proportionality of quasi-static and dynamic strengths of the material,
then the maximum forces of force-displacement graphs have been considered as representatives of the material
strengths.

Description of the experimental bending tests

In this research the considered hybrid component is made of a composite material (back) injected or compression
molding over a molded composite sheet. Three-point bending tests have been performed submitting the
component to both a quasi-static and velocity of 127 mm/s bending tests. Bending tests are performed in two
possible orientation of the component which are shown in Figure 5. It should be noted that in this paper Case1 is
chosen as the main research case, due to the more complex behavior than Case2. Both in Case1 and 2 tests,
components are subjected to a quasi-static bending loading characterized by a speed of the loading head of
1mm/s, in Case1 test the component has been also subjected to a bending loading with speed of 127 mm/s. The
bending span (distance between two supports) is 338 mm and is constant during all the tests as shown in Figure
Figure 7.

(a) (b)

Figure 5: Bending tests; (a) Case1 (b) Case2

Figure 6 clearly shows the differences between the hybrid beam under quasi static and 127
mm/sec loading velocity for both back injection molding and compression molding. The quasi
static loading condition for the Case 2 is shown by solid lines, while the results related the Case 1
both for the quasi-static and the loading velocity of 127 mm/s for the back-injected reinforcement
are reported using dash lines. Finally the curve related to Case 1 in quasi-static loading condition
for the compression molded reinforcement is reported with dashed line. Based on these
experimental results, one can say that for the Case1 the beam manufactured using back-injection
technology shows better energy absorption capacity than the compression molded solution.
Further it is possible to note that the 127 mm/s loading is giving higher resisting force with
respect to the quasi-static one and thus the strain-rate sensitivity of the beam material is
confirmed.

QS_Case1_Injected
6 127 mm/sec_Case1_Injected
QS_Case2_Injected
5 QS_Case1_Compression

4
Force (kN)

0
0 5 10 15 20
Displacement (mm)
Figure 6: Bending test results; quasi-static and 127 mm/sec velocity loading for Case1 and Case2. For
Case1 the results of compression molding technology are available.

(a) (b)
Figure 7: Bending tests; (a) test starting point (b) end point

Numerical simulations
Two different subroutines have been developed based on two proposed models for internal reinforcing
constituent and outer fabric shell. The VUMAT which developed for internal part is rate sensitive, which means
that by increasing velocity of puncher, the model shows proportional reaction and capture rate changes. The
flowchart shown in Figure 8 explains how the VUMAT works:
 Figure 8: Flowchart of the user-written subroutine algorithm

Using VUMAT, a user can implement a desired material model just by vectorising the solution dependent
variables for updating the stress tensor components. Main trial values of the subroutine are strain inceremnt
tensor components and previous step stress tensor components which generated by ABAQUS itself and
delivered to the subroutine. Then the current stress tensor component are computed based on the trial values and
the previous step components of stress tensor by passing into the integration routine. The strain rate tensor
components for applying in the subroutines are computed by dividing strain increment tensor components by
step-time as a time increment.
For some specific laminated composite materials most of the needed material properties are available [6]. While
day by day new generations of composite materials are developed and become available for industrial
applications, but most of the time for these new materials even some basic data, such as shear strengths, are not
available. Besides, fabric and short fiber composite materials do not follow the classical laminate theory in the
elastic zone and do not behave as the laminated composites after damage. Their responses during damage are as
complicated as those of laminated composites and their damage modes might be different. For instance, for an
injected short fiber composite the delamination term is meaningless, instead the material can accumulate a
different type of damage that is the debonding damage. It should be noticed that even if debonding damage is
present also in the laminated composites, it becomes much more relevant for the case of the injected short fiber

Four different damage parameters, i.e. the fracture energies `, for fiber and matrix under compression and
composite.

tension conditions, govern the material responses during the damage process evolution. Referring to the Lapczyk
and Hurtado [13] work, there is no available source to quantitative data for these values in our application, so for
the numerical problem the required values are assumed as listed in Table 7 by considering simmilar values which
can could be found in Ref.[6] .
The shell type element (S4R), with 4-node, reduced integration and hourglass control is used for all the
simulations. However, taking into account the ratio of wall thicknesses to the rest of the component dimensions,
and considering the ABAQUS manual suggestions, the results achieved from 2-D modeling would be regarded
as properly significant. Also, the number of assigned integration points varies from section to section according
to the section thickness and its distance from beam critical region that is located under the loading punch. The
total number of elements is 13564. The whole simulations have been performed in ABAQUS explicit dynamic
module. To avoid convergence problem the damage variables are not applied directly and following Lapczyk
and Hurtado approach the regularized forms are used as follows:

1„,œ = .1, − 1,œ 3


ɳg
(13)

the equation is strongly dependent to the value of viscous coefficient ɳž which in this study are considered to be
equal to 0.005 for all damage modes (I) and also the Eq. (13) is integrated by using Euler forward integration
method through the subroutines.
Table 7: Assumed energy release rates; f, m, t and c indicate fiber, matrix, tension and compression,
respectively.

Fabric shell (N/mm) Internal constituent (N/mm)

2, 2, 5, 5, 2, 2, 5, 5,
160 160 160 160 0.2 0.2 0.2 0.2

The other material properties are listed in Table 1 based on data received from the material supplier. For internal
reinforcing part it has been supposed that the material properties related to injected and compressed prats have
same values. The thickness assigned to the fabric shell is 1 mm while for the internal constituent varies from 2.2
to 5.4 mm at route.

Discussion

By observation of the experimental results shown in Figure 6, under quasi-static loading, Case1 is appropriate
for energy absorption and Case2 gives higher structural stiffness. Also, for 127 mm/sec velocity of loading,
Case1 shows a notable energy absorption potential. Also, the components reinforced by back-injection
technology show higher amount of energy absorption capability than ones reinforced by compression molding
technology for both quasi-static and 127 mm/sec loading conditions. Both compression molding and back
injection reinforced components show rate dependent behavior. Considering to Figure 6 the components
reinforced by both back-injection and compression molding technology show rate dependent behaviors.
However, all bending tests have been performed by applying 70 mm of deflection at the midspan of the beam.
Whereas, all simulations have been performed for 15-20 mm of punch displacement because of the main part of
damage is happened before 20 mm deflection and the results seem to be almost stable out of this range.
Considering the diagrams shown in Figure 6, Case2 includes minor damages before the major one which is
related to damage of the outer shell in the contact zone under compressive damage, and the major one which
leads to sudden drop down of the force-displacement graph is related to the breakage of the middle column of
reinforcing internal constituent. After full breakage of internal part, which is considered as a reinforcement for
the component, it seems that the outer shell as an open section may not tolerate the loading.
 QS_Case1_Exp2

5 QS_Case1_Exp3

QS_Case1_Numerical
4

Force (kN) 3

0
0 5 10 15 20
Displacement (mm)
Figure 9: Quasi static bending experimental and numerical results for Case1 and back-injection
technology

In Figure 9 experimental and computational results shown together for Case1 reinforced by back-injection
technology and subjected to a quasi-static bending load. The numerical simulation by considering the viscous
coefficient of 0.005 (s) shows a relatively good agreement with experimental results. The difference between the
numerical and experimental results mostly could be due to the mismatching between real orientation of the
internal injected part and the roughly applied material orientation in the FEM model, and also the change of
material properties during injection process. Even, the accumulation of short fibers in some zones near the
injection gates which could make the reinforcing columns more stiffer that the ribs can affect strongly the
experimental results which may not be considered in the macromechanically presented models.

QS_Case1_Numerical
QS_SC_Case1
4 QS_SC_Case1

3
Force (kN)

0
0 5 10 15 20
Displacement (mm)

Figure 10: Quasi static bending experimental and numerical results for Case1 with compression molding
technology

Figure 10 shows two different experimental results for two components of same type (Compression-molding
reinforcement) besides of the numerical results. The numerical parameters and effective constants are same as
used for back-injection reinforced one, but the numerical results show more agreement to experimental results
than back-injection reinforced ones. The main reason could be, on the contrary of the back-injected one, less
changes of the material properties of the reinforcing part respect to the original material properties in comparison
to the back-injection technology.
 6 127mm/se_Case1

5 127mm/sec_Case1

Force (kN)
4

0
0 5 10 15 20
Displacement (mm)

Figure 11: Bending experimental and numerical results of Case1 with back-injected internal part and
under loading velocity of 127 mm/sec

As a consequence of quasi-static numerical result for internally made by back-injected reinforcing subpart, the
numerical result of the component under 127 mm/sec loading velocity shows less agreement with experimental
results than the component reinforced by compression molding technology (compare Figure 11 and Figure 12).
Also, comparing the experimental bending test results of the components made of two different reinforcing
procedure of injection and compression moldings, the latter one illustrate less energy absorption capacity than
the first one.

4
Force (kN)

2
127mm/sec_Case1_Numerical

1 127mm/sec_SC_Case1

0
0 5 10 15 20
Displacement(mm)
Figure 12: Bending experimental and numerical results of Case1 with compression molding internal part
and under 127 mm/sec of loading

According to Figure 13 the numerical results predicts less total stiffness of the component under quasi static
bending than experimentally achieved results, but numerically predicted maximum achievable force before total
failure of the structure is almost happened between total displacement of 7 to 8 mm for both numerical and
experimental results.
 6
QS_Case2_Comp2_Injected

5 QS_Case2_Comp1_Injected

QS_Case2_Numerical
4
Force (kN)

0
0 2 4 6 8 10
Displacement (mm)
Figure 13: Bending experimental and numerical results of Case2 with back-injected internal part and
under quasi static loading

Conclusion
Hybrid composite beams using two simultaneous manufacturing procedures of hydroforming and injection or
compression molding have been produced and their responses under different bending rates have been studied.
Since the internal reinforcing part and external shell are dominated by matrix and fibers, respectively, then to
predict damage initiation of the internal and external parts, the second part of Hashin-Rotem’s failure criterion
and the maximum stress theory were chosen, in the order already mentioned. Two different kinds of damage
propagation models were derived and implemented in ABAQUS using subroutines so-called VUMAT for
internal and external parts. Since the developed model for the internal part is rate dependent, then rate
dependence effect incorporated into modules of elasticity and strength of reinforcing constituent, using a
logarithmic function. Mesh dependency problem was alleviated by incorporating characteristic length and
displacements in damage variables and convergence problem was solved using of regularized form of damage
variables. According the experimental results which have been conducted, it has been shown that the hybrid
composite beam reinforced by back-injection modling exhebits higher energy absorption capacity than the beam
reinforced by compression modling technology. The developed models for modelling of damage initiation and
propagation for internal and external parts have been applied using subroutine into ABAQUS, while the defined
model for the internal part has been considered to be rate dependent. The numerical results for both quasi-static
and 127 mm/sec loading speed show better agreement with experimental results of the beam reinforced by back-
injection technology than that one reinforced by compression molding. This might be related to the bigger
changes in material properties, respect to the original material properties, of the component reinforced by back –
injection technology than compression molded one.

Acknowledgements

This study was supported by Scholarship for doctoral training as part of the project “ITALY 2020”, of the
National Technological Cluster “Mobilità”. Authors would like to thank to Andrea Pipino for his kind support
during the project, Davide Gabellone for performing mechanical tests and Paolo Cacciabue for manufacturing of
the specimens. Also, authors would thank to Luca Belforte and Valentina Pau for taking electron microscope
images.

1. Matzenmiller, A., J. Lubliner, and R. Taylor, A constitutive model for anisotropic damage in
fiber-composites. Mechanics of materials, 1995. 20(2): p. 125-152.
2. Talreja, R. A continuum mechanics characterization of damage in composite materials. in
Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering
Sciences. 1985. The Royal Society.
3. Talreja, R. and C.V. Singh, Damage and failure of composite materials. 2012: Cambridge
University Press.
4. Hashin, Z. and A. Rotem, A fatigue failure criterion for fiber reinforced materials. Journal of
composite materials, 1973. 7(4): p. 448-464.
5. Puck, A. and H. Schürmann, Failure analysis of FRP laminates by means of physically based
phenomenological models. Composites Science and Technology, 1998. 58(7): p. 1045-1067.
6. Hinton, M.J., A.S. Kaddour, and P.D. Soden, Failure criteria in fibre reinforced polymer
composites: the world-wide failure exercise. 2004: Elsevier.
7. Dávila, C.G. and P.P. Camanho, Failure criteria for FRP laminates in plane stress. NASA
TM, 2003. 212663(613).
8. Chang, F.-K. and K.-Y. Chang, A progressive damage model for laminated composites
containing stress concentrations. Journal of composite materials, 1987. 21(9): p. 834-855.
9. Hill, R., The mathematical theory of plasticity. Vol. 11. 1998: Oxford university press.
10. Lemaitre, J., R. Desmorat, and M. Sauzay, Anisotropic damage law of evolution. European
Journal of Mechanics-A/Solids, 2000. 19(2): p. 187-208.
11. Maimí, P., et al., A continuum damage model for composite laminates: Part II–Computational
implementation and validation. Mechanics of Materials, 2007. 39(10): p. 909-919.
12. Camanho, P.P. and C.G. Dávila, Mixed-mode decohesion finite elements for the simulation of
delamination in composite materials. 2002.
13. Lapczyk, I. and J.A. Hurtado, Progressive damage modeling in fiber-reinforced materials.
Composites Part A: Applied Science and Manufacturing, 2007. 38(11): p. 2333-2341.
14. Bažant, Z.P. and B.H. Oh, Crack band theory for fracture of concrete. Matériaux et
construction, 1983. 16(3): p. 155-177.
15. Duvant, G. and J.L. Lions, Inequalities in mechanics and physics. Vol. 219. 2012: Springer
Science & Business Media.
16. Ryou, H., K. Chung, and W.-R. Yu, Constitutive modeling of woven composites considering
asymmetric/anisotropic, rate dependent, and nonlinear behavior. Composites Part A: Applied
Science and Manufacturing, 2007. 38(12): p. 2500-2510.
17. Shokrieh, M.M. and M.J. Omidi, Tension behavior of unidirectional glass/epoxy composites
under different strain rates. Composite Structures, 2009. 88(4): p. 595-601.
18. Shokrieh, M.M. and M.J. Omidi, Investigating the transverse behavior of Glass–Epoxy
composites under intermediate strain rates. Composite Structures, 2011. 93(2): p. 690-696.
19. Naik, N., et al., High strain rate tensile behavior of woven fabric E-glass/epoxy composite.
Polymer Testing, 2010. 29(1): p. 14-22.
20. Foroutan, R., et al., Experimental investigation of high strain-rate behaviour of fabric
composites. Composite Structures, 2013. 106: p. 264-269.
21. Weeks, C. and C. Sun, Modeling non-linear rate-dependent behavior in fiber-reinforced
composites. Composites Science and Technology, 1998. 58(3): p. 603-611.
22. Thiruppukuzhi, S.V. and C. Sun, Models for the strain-rate-dependent behavior of polymer
composites. Composites Science and Technology, 2001. 61(1): p. 1-12.
23. Aidun, J.B. and F. Addessio, An enhanced cell model with nonlinear elasticity. Journal of
composite materials, 1996. 30(2): p. 248-280.
24. Goldberg, R.K. and D.C. Stouffer, Strain rate dependent analysis of a polymer matrix
composite utilizing a micromechanics approach. Journal of composite materials, 2002. 36(7):
p. 773-793.
25. Tabiei, A., W. Yi, and R. Goldberg, Non-linear strain rate dependent micro-mechanical
composite material model for finite element impact and crashworthiness simulation.
International Journal of Non-Linear Mechanics, 2005. 40(7): p. 957-970.
26. Daniel, I.M., B.T. Werner, and J.S. Fenner, Strain-rate-dependent failure criteria for
composites. Composites Science and Technology, 2011. 71(3): p. 357-364.
27. Wei, X., et al., A new rate-dependent unidirectional composite model–Application to panels
subjected to underwater blast. Journal of the Mechanics and Physics of Solids, 2013. 61(6): p.
1305-1318.
28. Wedgewood, A.R., P. Granowicz, and Z. Zhang, Multi-Scale Modeling of an Injection Over-
Molded Woven Fabric Composite Beam. 2014, SAE Technical Paper.
29. Iannucci, L. and M. Willows, An energy based damage mechanics approach to modelling
impact onto woven composite materials—Part I: Numerical models. Composites Part A:
Applied Science and Manufacturing, 2006. 37(11): p. 2041-2056.
30. Donadon, M., et al., A progressive failure model for composite laminates subjected to low
velocity impact damage. Computers & Structures, 2008. 86(11): p. 1232-1252.
31. Maimí, P., et al., A thermodynamically consistent damage model for advanced composites.
2006.
32. Tijssens, M.G., B.L. Sluys, and E. van der Giessen, Numerical simulation of quasi-brittle
fracture using damaging cohesive surfaces. European Journal of Mechanics-A/Solids, 2000.
19(5): p. 761-779.
33. Daniel, I.M., et al., Interfiber/interlaminar failure of composites under multi-axial states of
stress. Composites Science and Technology, 2009. 69(6): p. 764-771.
34. Daniel, I., et al. Mechanical behaviour and failure criteria of composite materials under static
and dynamic loading. in Proceeding of 17th international conference on composite materials
(ICCM17), Edinburgh, UK. 2009.
35. Ochola, R., et al., Mechanical behaviour of glass and carbon fibre reinforced composites at
varying strain rates. Composite Structures, 2004. 63(3): p. 455-467.
36. Daniel, I., B. Werner, and J. Fenner, Strain-rate-dependent failure criteria for composites.
Composites Science and Technology, 2011. 71(3): p. 357-364.