Damage Modeling in Random

Hicham Mir
Research Professional
Short Glass Fiber Reinforced
e-mail: [email protected]

Mario Fafard
Composites Including Permanent
Professor
e-mail: [email protected] Strain and Unilateral Effect
Benoı̂t Bissonnette This paper presents the development of a theoretical damage mechanics model applicable
e-mail: [email protected] to random short glass fiber reinforced composites. This model is based on a macroscopic
approach using internal variables together with a thermodynamic potential expressed in
Marie-Laure Dano the stress space. Induced anisotropic damage, nonsymmetric tensile/compressive behavior
e-mail: [email protected] (unilateral effect) and residual effects (permanent strain) are taken into account. The
anisotropic damage is represented with second-order tensorial internal variables D. The
Department of Civil Engineering, unilateral effect due to microcrack closure in compression is introduced by generalizing
Université Laval, the hypothesis of the complementary elastic energy equivalence. In the case of the per-
Québec City, manent strain, a new term related to frozen energy, which is a function of the damage
Quebec G1K 7P4, Canada variable, the stress tensor, and some materials constants to be identified, is added to the
basic thermodynamic potential. Using laboratory test results, parameter identification has
been performed to illustrate the applicability of the proposed model.
关DOI: 10.1115/1.1839593兴

1 Introduction elastic modulus decrease with progressive repeated tensile loading

tests. Permanent strains have been also observed in these tests.
The industry of utility and recreative vehicles uses random
However, loading and unloading in compression show that there
short glass fiber reinforced composites to construct a wide variety is neither stiffness loss nor permanent strain due to damage, indi-
of parts due to its light weight and the ease with which complex cating that the material behavior is linear elastic. Complementary
shapes can be achieved. For example, it is used in the manufac- tests have been performed; they are the pseudo-biaxial tensile
turing of seats for subway trains, shells for watercraft and electri- tests 关8兴. After damage has been induced in one direction, denoted
cal vehicles, side panels of buses, etc. To optimize those parts, 1-direction in Fig. 1, the material has been loaded in an elastic
both the elastic properties of the material and its behavior under domain in some other selected direction 共45 and 90 deg兲 in order
heavy loads inducing progressive damage up to failure have to be to estimate Young’s modulus, Poisson’s ratio, and the shear modu-
characterized. lus. These tests have been performed for different damage levels.
For two decades, many papers have been published on the mod- Results show clearly that the transverse Young’s modulus 共modu-
eling of damage mechanics using a thermodynamic approach with lus in the 2-direction兲 is reduced by the damage induced in the
internal variables 关1–5兴. Although those models were developed 1-direction.
using a macroscopic formulation, microscopic considerations and This paper presents a new model based on TIP with internal
observations were used to justify them. The framework of the variables to take into account the unilateral effect of the composite
thermodynamic of irreversible processes 共TIP兲 using internal vari- and permanent strain 关9兴. The data published in 关7,8兴 and an iden-
ables is probably one of the best approaches to model damage tification procedure were used to determine the parameters of the
mechanics. However, there still remain many open questions for proposed model and the tensile and shear tests were simulated.
the modeling of induced anisotropy 共in particular, in the case of
nonproportional loading兲, desactivation of damage due to micro-
crack closure 共unilateral effect兲, and residual effects. 2 Description of the Proposed Model
The aim of this paper is to present a theoretical damage model 2.1 Modified Hypothesis of Complementary Elastic En-
able to predict the behavior of glass fiber reinforced composites ergy Equivalence. In the case of an isotropic material, the
and the progressive degradation preceding failure. The material is complementary elastic energy is defined as
made of polyester resin and short glass fibers distributed randomly
in the plane of the part using a robotized technology. In Ref. 关6兴, 1 ␯
U e0 ⫽ ␴:␴⫺ 共共 ␴:I2兲 2 ⫺ ␴:␴兲 (1)
some results about the characterization of the tensile behavior of 2E 2E
this material have been published. Those tests results also pro-
where I2 is a second-order unit tensor, ␴ is the stress tensor, E is
vided some information pertaining to the different damage mecha-
Young’s modulus, and ␯ is Poisson’s ratio. After a tensile loading
nisms. In Refs. 关7,8兴, detailed information related to the charac- causing damage, when the material is loaded in compression, as
terization of this material can be found. Damage is revealed by the observed on test specimens, the damage or the 共smeared兲 cracks
seem to disappear. The explanation is relatively simple. In the
Contributed by the Applied Mechanics Division of THE AMERICAN SOCIETY OF
MECHANICAL ENGINEERS for publication in the ASME JOURNAL OF APPLIED ME-
general case, a crack is closed 共in part or totally兲 when a stress
CHANICS. Manuscript received by the Applied Mechanics Division, October 30, normal to its plane is a compressive stress. To take into account
2003; final revision, August 12, 2004. Associate Editor: K. M. Liechti. Discussion on this phenomenon in the expression of complementary elastic en-
the paper should be addressed to the Editor, Prof. Robert M. McMeeking, Journal of ergy of the damage material, a new hypothesis is introduced that
Applied Mechanics, Department of Mechanics and Environmental Engineering, Uni-
versity of California–Santa Barbara, Santa Barbara, CA 93106-5070, and will be
is a modification of the hypothesis of complementary elastic en-
accepted until four months after final publication in the paper itself in the ASME ergy equivalence proposed by Sidoroff 关10兴. It is assumed that the
JOURNAL OF APPLIED MECHANICS. complementary elastic energy of the damaged material has the

Journal of Applied Mechanics Copyright © 2005 by ASME MARCH 2005, Vol. 72 Õ 249

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 1 Tensile testing on rectangular plates and specimens cut from it along specific
directions

same form as an equivalent undamaged material by replacing the ␦ ik ␦ jl

usual stress variable by an effective one, except for the energy ᠪ 共 D兲 ⫽
M ni 丢 n j 丢 nk 丢 nl
linked to the compression that is responsible for the crack closure. 冑共 1⫺D i 兲共 1⫺D j 兲
In the principal coordinate system of damage and using a clas-
sical indicial notation, the complementary elastic energy using the 1
⫽ ni 丢 n j 丢 ni 丢 n j (6)
postulate described previously can be written as 冑共 1⫺D i 兲共 1⫺D j 兲
U e 共 ␴,D兲 ⫽
1 ⫹ ⫹ 1 ⫺ ⫺ 1
˜␴ ˜␴ ⫹ ␴ ␴ ⫹ ˜␴ ˜␴
2E ii ii 2E ii ii 2E i j i j
冏 i⫽ j where D i is the damage eigenvalue in the principal direction i.
The canonical form is used because the fourth-order tensor M ᠪ (D)
␯ is a real symmetric tensor and its second-order eigentensors asso-
⫺ ␴ ˜␴ ⫺ ˜␴ i j ˜␴ i j 兲
共˜ (2)
2E ii j j ciated with the real positive eigenvalues 1/ 冑(1-D i )(1-D j ) are
ni 丢 n j .
where ˜␴ ii⫹ is the tensile stress 共positive兲, and ˜␴ ii⫺ is the compres- Using Eqs. 共3兲, 共5兲, and 共6兲, Eq. 共2兲 becomes
sive stress 共negative兲 normal to the plane of the microcrack sys-
tem. In other words, the compressive stress ␴⫺ is defined as
␴⫺ ⫽H 共 ⫺n i ␴n i 兲共 n i ␴n i 兲 n i 丢 n i ⫽Pᠪ ( ␴,D)⫺ : ␴, U e 共 ␴,D兲 ⫽
1
␴:共 Pᠪ ( ␴,D)⫹ :M ᠪ 共 D兲 :Pᠪ ( ␴,D)⫹ 兲 :␴
ᠪ 共 D兲 :M
2E
Pᠪ ( ␴,D)⫺ ⫽H 共 ⫺n i ␴n i 兲共 n i 丢 n i 丢 n i 丢 n i 兲 (3)
1
where n i are the principal directions of the damage and H is the ⫹ ␴:共 Pᠪ ( ␴,D)⫺ :Pᠪ ( ␴,D)⫺ 兲 :␴
Heaviside function defined by 2E

H共 x 兲⫽ 再 1,
0,
x⬎0
x⬍0
(4) ⫹
1
2E
␴:共 Eᠪ ijij:M
ᠪ 共 D兲 :M ᠪ ijij兲 i⫽j:␴
ᠪ 共 D兲 :E

␯
The quantity Pᠪ ( ␴ ,D)⫺ is a fourth-order tensor corresponding to a ⫺ ᠪ 共 D兲 :␴兲兲 2 ⫺ ␴:共 M
关共 tr 共 M ᠪ 共 D兲兲 :␴兴
ᠪ 共 D兲 :M
negative projection operator. In the same way, the positive projec- 2E
tion operator can be defined as (7)
( ␴,D)⫹
Pᠪ ⫽H 共 n i ␴n i 兲共 n i 丢 n i 丢 n i 丢 n i 兲 (5)
ᠪ (D) has a canonical form
The fourth-order damage operator M where Eᠪ ijij⫽n i 丢 n j 丢 n i 丢 n j . The second term of Eq. 共7兲 can also
关11兴 be written as

250 Õ Vol. 72, MARCH 2005 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1 1
␴:共 Pᠪ ( ␴,D)⫺ :Pᠪ ( ␴,D)⫺ 兲 :␴ ⌬U e 共 ␴,D兲 ⫽ ␴:关 Pᠪ ( ␴,D)⫺ :共 M ᠪ ⫺Iᠪ4 兲 :Pᠪ ( ␴,D)⫺ 兴 :␴
ᠪ :M (10)
2E 2E
1 Substituting Eqs. 共5兲 and 共6兲 in Eq. 共10兲, one obtains
⫽ ␴:共 Pᠪ ( ␴,D)⫺ :M ᠪ :Pᠪ ( ␴,D)⫺ 兲 :␴
ᠪ :M
2E
1
1 ⌬U e 共 ␴,D兲 ⫽ ᠪ̆ „D…:␴
␴:M (11)
⫺ ␴:关 Pᠪ ( ␴,D)⫺ :共 M ᠪ ⫺Iᠪ4 兲 :Pᠪ ( ␴,D)⫺ 兴 :␴
ᠪ :M (8) 2E
2E
where Iᠪ4 is a fourth order unit tensor. Using Eq. 共8兲, Eq. 共7兲 ᠪ̆ (D) is defined as
The fourth-order operator M
becomes
D i 共 2⫺D i 兲
1 ␯ ᠪ̆ 共 D兲 ⫽H 共 ⫺n i ␴n i 兲
M n 丢 ni 丢 ni 丢 ni (12)
U 共 ␴,D兲 ⫽
e
␴:共 M ᠪ 兲 :␴⫺
ᠪ :M 关共 tr 共 M:␴兲兲 2 ⫺ ␴:共 M ᠪ 兲 :␴兴
ᠪ :M 共 1⫺D i 兲 2 i
2E 2E
and thus, the complementary elastic energy defined by Eq. 共9兲 can
1
⫺ ␴:关 Pᠪ ( ␴,D)⫺ :共 M ᠪ ⫺Iᠪ4 兲 :Pᠪ ( ␴,D)⫺ 兴 :␴
ᠪ :M (9) now be written as
2E
The first two terms of Eq. 共9兲 are identical to the one obtained by 1 1
ᠪ̃ ⫺1 :␴⫺
U e 共 ␴,D兲 ⫽ ␴:C ␴:M
ᠪ̆ :␴ (13)
Sidoroff 关10兴 when applying the original postulate of complemen- 2 2E
tary elastic energy equivalence. The last term represents the res-
toration of the system rigidity due to crack closure. In that case, where C ᠪ̃ ⫺1 ⫽M ᠪ ⫺1 :M
ᠪ :C ᠪ is the fourth-order elastic compliance ten-
the microcrack system is said to become inactive. The definition sor of the damaged material which could be written in the princi-
corresponding to the last term of Eq. 共9兲 is written as follows: pal coordinate system of damage as

1 ⫺␯ ⫺␯
0 0 0
E 共 1⫺D 1 兲 2 E 共 1⫺D 1 兲共 1⫺D 2 兲 E 共 1⫺D 1 兲共 1⫺D 3 兲

l m
⫺␯ 1 ⫺␯
0 0 0
E 共 1⫺D 2 兲共 1⫺D 1 兲 E 共 1⫺D 2 兲 2 E 共 1⫺D 2 兲共 1⫺D 3 兲
⫺␯ ⫺␯ 1
0 0 0
E 共 1⫺D 3 兲共 1⫺D 1 兲 E 共 1⫺D 3 兲共 1⫺D 2 兲 E 共 1⫺D 3 兲 2
⫺1
ᠪ̃
C ⫽
1
0 0 0 0 0
G 共 1⫺D 2 兲共 1⫺D 3 兲
1
0 0 0 0 0
G 共 1⫺D 1 兲共 1⫺D 3 兲
1
0 0 0 0 0
G 共 1⫺D 1 兲共 1⫺D 2 兲
(14)

It can be observed from Eq. 共13兲 that if all components of the material constants to be identified. This energy can be viewed as
ᠪ̆ (D) is inactive 共microcracks
stress tensor are positive, the term M frozen by the microcracks. The thermodynamic potential can be
opening兲, and the classical form defined by Sidoroff 关10兴 is rewritten as
obtained. 1 1
U 共 ␴,D兲 ⫽ ␴:C ᠪ̃ ⫺1 :␴⫺ ␴:Mᠪ̆ :␴⫹ ␴:A
ᠪ :D (16)
2.2 Residual Effect. When loads 共tensile stress兲 are applied 2 2E
on a composite structure, cracks and thus damage are induced in 2.3 State Laws
the material. The level of degradation is quantified through the
second-order damage tensor D. On the contrary, during the un- Elastic Constitutive Law. The elastic constitutive law of the
loading phase, the microcracks progressively close up to a certain damaged material is obtained by differentiating the dual potential
extent 共at the end of the process, the microcracks are not neces- with respect to the stress tensor. The strain is defined in two parts
sarily entirely closed兲. This phenomenon has microscopic justifi- due the Heaviside function in the potential expression:
cations related to the nature and geometry of the crack, which fall 1
beyond the scope of this paper. Explanations of the phenomenon ␧⫽C ᠪ̃ ⫺1 :␴⫺ M ᠪ̆ :␴⫹A ᠪ :D (17)
E
can be found in Refs. 关4,11兴. In the present paper, this residual
effect 共residual strain兲 is attributed to the state of damage, and the The total strain is composed of an elastic part ( ␧e ) and a perma-
existence of a potential, noted U p , which is a function of the nent part ( ␧P ):
damage tensor D and the stress tensor is postulated as
␧e ⫽ C 冉 1
ᠪ̃ ⫺1 ⫺ M
E 冊
ᠪ̆ :␴ (18a)

U 共 ␴,D兲 ⫽ ␴:A
p
ᠪ :D (15) ␧p ⫽A ᠪ :D (18b)
The elastic compliance tensor of the damaged material taking into
ᠪ is a symmetric fourth-order tensor whose coefficients are
where A account the unilateral effect has the following form:

Journal of Applied Mechanics MARCH 2005, Vol. 72 Õ 251

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 1
ᠪ̃ T⫺1 ⫽C
C ᠪ̃ ⫺1 ⫺ ᠪ̆
M
E
1 D i 共 2⫺D i 兲
ᠪ̃ ⫺1 ⫺
⫽C H 共 ⫺n i ␴n i 兲 n 丢 n i 丢 n i 丢 n i (19)
E 共 1⫺D i 兲 2 i
In the case of a compressive stress normal to the mean surface of
the microcrack system n i , the elastic modulus C̃ T⫺1n i n i n i n i becomes

C̃ T⫺1n i n i n i n i ⫽n i 丢 ni :C
ᠪ̃ T⫺1 :n i 丢 ni

1 1 D i 共 2⫺D i 兲 1
⫽ ⫺ ⫽ (20)
E 共 1⫺D i 兲 2 E 共 1⫺D i 兲 2 E
This way, the elastic modulus in the direction normal to the mi-
crocrack system loaded in compression is restored. In addition,
the strain–stress relation defined in Eq. 共17兲 is continuous and the
elastic compliance tensor is symmetric and positive definite. The
thermodynamic model is thus physically consistent.
Associated Damage Forces. The thermodynamic force, which
is known as the damage strain energy release rate, must be asso- Fig. 2 Cyclic tensile stress–strain curves „Test †8‡…
ciated with the damage tensor. In the case wherein the principal
directions of the damage should not change during loading 共pro-
portional loading兲, for each eigenvalue of the damage D i , there is an estimation process using a calibrating technique. It is necessary
an associated thermodynamic force Y i , defined as in the course of the test to load and unload the specimen in order

冋冉 冊册
to estimate the damage internal variables D 1 , D 2 and to measure
⳵U 1 ⳵M
ᠪ ⳵M
ᠪ the elastic and permanent strains ( ␧e, ␧p). For further details on
Y i⫽ ⫽ ␴: ᠪ ⫺1 :M
:C ᠪ ⫹M ᠪ ⫺1 :
ᠪ :C :␴
⳵Di 2 ⳵Di ⳵Di the experimental procedure, the reader is referred to Refs. 关7,8兴.

冋 册
Essentially, the results of the uniaxial cyclic tensile tests 共Fig. 2兲
1 are used here for parameter identification purposes.
⫺ ␴: H 共 ⫺n i ␴n i 兲 n 丢 n i 丢 n i 丢 n i :␴
E 共 1⫺D i 兲 3 i Uniaxial tensile tests 关6 – 8兴, have shown, if one neglects
the permanent strain, that the following relationship between the
⫹ ␴:A
ᠪ :n i 丢 n i (21a)
principal damage function and the thermodynamic force is

冋 冉
Y i⫽ ␴: sym
⳵M
ᠪ
⳵Di
ᠪ ⫺1 :M
:C 冊
ᠪ ⫺H 共 ⫺n i ␴n i 兲
1
n
E 共 1⫺D i 兲 3 i
appropriate:
f 共 Y s 兲 ⫽a 共 Y s ⫺Y 0 兲 (26)

丢 ni 丢 ni 丢 ni 册 :␴⫹ ␴:A
ᠪ :n i 丢 n i (21b)
where Y is the thermodynamic force defined in Eq. 共25兲, Y 0 is the
s

initial value at which damage begins, and a is a material param-

eter. These two parameters must be identified.
In fact, to take into account the possibility of interaction between In the principal coordinate system of damage 共2-D case兲, the A
two principal damage directions 共two-dimensional 共2-D兲 case兲, a operator 共Eq. 共15兲兲 is reduced to the following expression using
correction must be done to the expression of the thermodynamic Voigt’s notation 共due to symmetry of ␴, D, and A兲:

冋 册
forces. A weighted sum of the two thermodynamic forces is thus
used: ␣ ⫺␤
A⫽ (27)
⫺␤ ␣
Ȳ i ⫽Y i ⫹bY j , i, j⫽1,2, i⫽ j (22)
The residual 共permanent兲 strains due to damage are obtained as
where the parameter b is a material constant bounded between 0 follows 共Eq. 共18b兲兲:

再
and 1.
␧ 11
p
⫽ ␣ D 1⫺ ␤ D 2
Damage Evolution Law. The thermodynamic forces drive the (28)
evolution of the internal variable characterizing the damage up to ␧ 22
p
⫽⫺ ␤ D 1 ⫹ ␣ D 2
failure. Those forces must satisfy the Clausius–Duhem inequality where ␣ and ␤ are unknown material parameters to be identified
due to damage: using loading tests results.
Equations 共21兲 and 共22兲 can also be simplified in the case of
Y:Ḋ⭓0 (23) uniaxial loading:
The evolution 共quasi-static兲 law satisfying this inequality is cho-
sen to be of the following form: ␴ 21
Ȳ 1 ⫽ ⫹ 共 ␣ ⫺b ␤ 兲 ␴ 1 ,
E 共 1⫺D 1 兲 3
D i ⫽ f 共 Y is 兲 , i⫽1,2 (24) (29)
␴ 21
Y is 共 t 兲 ⫽max兵 Y 0 ,sup共 Ȳ i 共 ␶ 兲兲 其 (25) Ȳ 2 ⫽ 共 b ␣ ⫺ ␤ 兲 ␴ 1 ⫹b
␶ ⭐t
E 共 1⫺D 1 兲 3

where Y 0 designates the initial damage threshold and i is one of Table 1 summarizes the unknown parameters to be identified us-
the principal direction. The function f is a growing positive func- ing an appropriate technique together with test results. In Table 2,
tion intrinsic to the material. the test results to be used for the identification are displayed. A
loading/unloading test allows the estimation of the secant damage
3 Model Identification Young’s modulus. The damage value can thus be estimated using
Eq. 共14兲:
To identify the parameters of the proposed model, progressive
repeated tensile loading tests must be performed and followed by E 1 ⫽E 共 1⫺D 1 兲 2 (30)

252 Õ Vol. 72, MARCH 2005 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 1 Definition of the unknown parameters

Unknown
parameters a Y0 b ␣, ␤
Mathematical Scalar Scalar Scalar Fourth-order
nature tensor
Definition Material Thermodynamic Parameter Parameters to
coefficient force beyond necessary to take estimate
necessary to which damage into account cross permanent strains
estimate the begins effects in the due to damage
damage level thermodynamic
force

and then: NC

兺 储共 ␧ m ⫺␧ t 兲 i 储
p p 2

冑
i⫽1
E1 ERR – STR⫽ NC
D 1 ⫽1⫺
兺 储共 ␧
(31)
m兲i储
p 2
E
i⫽1

NC
In this unique test, it is difficult to estimate the damage parameter
in the transverse direction (D 2 ). Information about the damaged 兺 关共 ␧ m11⫺␧ t11 兲 i ⫹ 共 ␧ m22⫺␧ t22 兲 i 兴
p p 2 p p 2

material Poisson’s ratio ␯ 12 allows us to estimate its values using ⫽

i⫽1
(35)
NC
Eq. 共14兲:
兺
i⫽1
关共 ␧ m11
p
兲 i2 ⫹ 共 ␧ m22
p
兲 i2 兴

1⫺D 1 ␯ Subject to the following constraint:

␯ 12⫽ ␯ ⇒D 2 ⫽1⫺ 共 1⫺D 1 兲 (32)
1⫺D 2 ␯ 12 NC

However, in most cases, this expression may not give reliable

兺 共D
i⫽1
1 ⫺D 1 兲 i
m t 2

damage values because the damaged Poisson’s ratio ␯ 12 measure- ERR – EVO⫽ NC ⭐Tol (36)
ment is not very precise. Instead, D 2 is evaluated using Eqs. 共24兲–
共26兲 and 共29兲:
兺
i⫽1
1 兲i
共Dm 2

where i is the cycle index, NC is the total number of cycles in

冋
D 2 ⫽a 共 Ȳ 2 ⫺Y 0 兲 ⫽a 共 b ␣ ⫺ ␤ 兲 ␴ 1 ⫹b
␴ 21
E 共 1⫺D 1 兲 3
⫺Y 0 册 (33)
axial cyclic tensile test (⫽6, Fig. 2兲 and ␧ mp and ␧ tp are the mea-
sured and predicted 共Eq. 共34兲兲 permanent strains, respectively. D m
and D t1 are the measured and predicted 共using Eqs. 共24兲–共26兲 and
1

共29兲兲 damage variable values, respectively, in the tensile stress

Now, Eq. 共28兲 can be used to express the permanent strain as a direction. A value of 6.9⫻10⫺4 was chosen for the constraint
function of the five unknown parameters Y 0 , a, b, ␣, and ␤: parameter Tol. The parameters have been calibrated by means of
the Lagrangian operator and the Uzawa’s method 关12兴.
Table 3 gives the values obtained from the calibration proce-
␴ 21

冦
␧11
p
⫽ ␣ D 1 ⫺ ␤ a 共 b ␣ ⫺ ␤ 兲 ␴ 1 ⫺b ␤ ⫹ ␤ aY 0 dure. The precision obtained with Eqs. 共34兲 and 共35兲 was 7.7
E 共 1⫺D 1 兲 3 ⫻10⫺3 and 6.899⫻10⫺4 , respectively. Table 4 gives a compari-
␴ 21 son between the measured and the calculated values of permanent
␧ 22
p
⫽⫺ ␤ D 1 ⫹ ␣ a 共 b ␣ ⫺ ␤ 兲 ␴ 1 ⫹b ␣ ⫺ ␣ aY 0 strain and damage D 1 . A very good correlation between those
E 共 1⫺D 1 兲 3 values is observed except for the first two lines of the table cor-
(34) responding to applied stresses of 24.50 and 39.43 MPa. The per-
manent strain and damage parameters generally have low values,
Parameter identification is done using a constrained optimization particularly in the first few cycles 共2 to 3 cycles, see Table 4兲. It is
technique. The following equation is used as the objective func- not easy to obtain reliable measured values and consequently, it
tion to be optimized: may be an error source for correlation and fitting. However, in the
last three cycles, the percentage differences of the expected and
the measured values of strain and damage parameters are within
the ranges of 关 ⫺5.6%, 9.2%兴 and 关 ⫺0.5%, 0.8%兴, respectively.
Table 2 Experimental results from Ref. †7‡ Therefore, the proposed model with the values of parameters
listed in Table 3 is applicable in representing the damage mechan-
Maximum Permanent axial Permanent Damage in the ics behavior of random short glass fiber reinforced composites
stress level strain ␧ 11 transverse strain ␧ 22 direction of loading under relatively large strain.
共MPa兲 共␮m/m兲 共␮m/m兲 D1
24.50 72.24 ⫺126.7 0.0134649
39.43 240.8 ⫺267.6 0.025708 Table 3 Numerical values of the unknown parameters
51.41 337.1 ⫺352.1 0.043811
61.34 550.7 ⫺450.6 0.066579 Y0 a b ␣ ␤
70.27 722.4 ⫺549.2 0.091765
77.30 987.2 ⫺647.8 0.115231 0.06819 0.07884 0.5878 0.010071 0.008371

Journal of Applied Mechanics MARCH 2005, Vol. 72 Õ 253

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 4 Comparison between estimated and measured values

D 1 (10⫺3 ) Permanent strain ␧ 11 共␮m/m兲 Permanent strain ␧ 22 共␮m/m兲

Measured Calculated Error % Measured Calculated Error % Measured Calculated Error %
13.47 10.98 18.5 72.24 135.7 ⫺87.8 ⫺126.7 ⫺112.7 11.0

25.71 27.85 ⫺8.3 240.8 258.9 ⫺7.5 ⫺267.6 ⫺215.2 19.6

43.81 46.46 ⫺6.0 337.1 417.1 ⫺23.7 ⫺352.1 ⫺337.7 4.1
66.58 66.91 ⫺0.5 550.7 581.7 ⫺5.6 ⫺450.6 ⫺450.4 0.0
91.77 90.65 1.2 722.4 750.8 ⫺3.9 ⫺549.2 ⫺559.6 ⫺1.9
115.2 114.4 0.8 987.2 895.9 9.2 ⫺647.8 ⫺646.2 0.2

4 Numerical Simulation Figure 5 illustrates the result for monotonic tensile loading. A
very good agreement between numerical and test results is ob-
4.1 Numerical Implementation. The proposed model has tained. Monotonic and cyclic loading have been simulated in
been implemented in the ABAQUS finite-element code using the shear. Figures 6 and 7 illustrate both the numerical and the experi-
UMAT functionality. The numerical integration is done by dis- mental results, which again are very close. References 关7,8兴 give
cretizing the loading using load increments and thus strain incre- all the information on the test procedure.
ments. From the mechanical state ␧ n , ␴ n , D n , at time t n , an The evolution of the degradation of the elastic properties in
estimation is made of the mechanical state ␴ n⫹1 , D n⫹1 , at time function of tensile stress level in three different directions 共0, 45,
t n⫹1 corresponding to the strain increment ⌬␧. This local integra- and 90 deg兲 is presented in Figs. 8–10. In Figs. 8 and 9, the
tion is done using an implicit integration algorithm and a predicted damaged Young’s modulus and Poisson’s ratio can be
Newton–Raphson technique to estimate the mechanical state. For
this purpose, a consistent tangential matrix has been derived. The
reader can find more information in Mir 关9兴.
4.2 Simulation and Model Validation. First, to verify how
consistent the identified set of parameters is, the uniaxial cyclic
tensile test has been simulated using the numerical model im-
planted in ABAQUS. Figures 3 and 4 illustrate, respectively, the
numerical cyclic tensile test results obtained from finite-element
simulation and the corresponding experimental results superposed
to the predicted curves. A good agreement is found. Now, to vali-
date the model, different experimental tests 关7,8兴 have been simu-
lated: monotonic tensile tests, monotonic and cyclic shear tests,
and tensile tests on rectangular plates and specimens cut from it
共Fig. 1兲. In these last tests, the tensile and shear specimens were
cut from the plates pre-loaded in tension to produce different de-
grees of damage by uniaxial tension 关7,8兴. Afterwards, the speci-
mens were tested to measure their damaged-material elastic coef-
ficients: Young’s modulus and Poisson’s ratio along three
directions 共0, 45, and 90 deg with respect to the initial loading
direction兲 and the shear modulus 共Fig. 1兲. The objective was to
predict their evolution and degradation versus the tensile stress
level.
Fig. 4 Comparison of the predicted cyclic tensile stress–
strain curves with corresponding experimental data

Fig. 3 Predicted cyclic tensile stress–strain curves „Model… Fig. 5 Tensile stress–strain curves

254 Õ Vol. 72, MARCH 2005 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 6 Comparison of predicted shear stress–strain curves
with the corresponding experimental data

compared to the test results reported by Dano et al. 关8兴. The pre-
diction of Young’s modulus and Poisson’s ratio is relatively good
in the 1- and 2-directions. In the 45 deg direction, the discrepancy
is more accentuated, though the numerical prediction still remains
fair.
In the case of the shear modulus presented in Fig. 10, it can be
observed that the numerical results are in good agreement with the
experimental data.
4.3 Simulation of a Uniaxial Tensile–Compressive Load-
ing Test. In this simulation, we show how reversing a tensile
load to compressive load restores the apparent elastic modulus
共degraded by tensile load兲 in the direction parallel to the compres-
sive load and how it may still cause damage evolution 共cracks
extension兲 in the transverse direction. These two directions corre-
spond to the two principal damage directions in this load case. We
denote by 1 and 2 the directions that are parallel and transverse to
the loading direction, respectively.
Figure 11 illustrates the stress versus strain curve during a
tensile–compressive loading test. The specimen is first loaded in
tension up to 60 MPa, which induces damage and causes Young’s
modulus E 1 to decrease. The specimen is then, unloaded and sub-

Fig. 8 „a… Comparison of predicted Young’s modulus evolu-

tion vs stress with the corresponding experimental data „ ␪
Ä0 deg…. „b… Comparison of predicted Young’s modulus
evolution vs stress with the corresponding experimental data
„ ␪ Ä45 deg…. „c… Comparison of predicted Young’s modulus
evolution vs stress with the corresponding experimental data
„ ␪ Ä90 deg….

jected to a compressive stress. The change of slope indicates that

the apparent modulus is restored when the damaged material is in
Fig. 7 Comparison of cyclic shear stress–strain curves with compression. Figures 12共a兲 and 12共b兲 present the evolution of the
the corresponding experimental data applied stress and the damage variables during the tension–

Journal of Applied Mechanics MARCH 2005, Vol. 72 Õ 255

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 10 Comparison of predicted Shear modulus evolution vs
stress with the corresponding experimental data

stress has reached a certain level (⫺78 MPa), which depends on

the tensile load level applied previously, the degree of damage in
the 2-direction (D 2 ) increases again, as shown in Fig. 12共b兲. The
compressive load may transmit damage and induce its evolution
in the transverse direction. The cracks nucleated initially due to
previous tensile load are not always perpendicular to the loading
direction, but also may be along individual fiber directions and
thus be inclined or parallel to the loading direction. This set of
cracks becomes active under compressive load applied in the pre-
vious tensile load direction.

5 Conclusion
In this paper, a damage model for random short glass fiber
reinforced composites based on the fundamental principle of ther-
modynamics of irreversible process was presented. The model
takes into account the unilateral effect 共crack closure effect兲 and
the permanent strains after unloading 共residual effect兲. Using tests
results and an appropriate identification procedure, all unknown
parameters have been identified. After implementation of the
model in a finite-element code, simulations are obtained and are
found to be in good agreement with experimental results. It can
thus be concluded that the proposed model is appropriate to simu-

Fig. 9 „a… Comparison of predicted Poisson ratio evolution vs

stress with the corresponding experimental data „ ␪ Ä0 deg….
„b… Comparison of predicted Poisson’s ratio evolution vs stress
with the corresponding experimental data „ ␪ Ä45 deg…. „c…
Comparison of predicted Poisson’s ratio evolution vs stress
with the corresponding experimental data „ ␪ Ä90 deg….

compressive test simulation, respectively. First, up to 60 MPa, the

tensile load causes material damage, and consequently, increases
principal damage variables D 1 and D 2 , as shown in Fig. 12共b兲.
During unloading and until starting the compressive load, the
damage variables stop increasing. As demonstrated by Eq. 共20兲
and shown in Fig. 11, the compressive load restores the apparent Fig. 11 Stress–strain curve in tensile–compressive load
elastic modulus in the 1-direction. However, once the compressive „Model…

256 Õ Vol. 72, MARCH 2005 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 ERR – EVO ⫽ damage relative error in the 1-direction
G ⫽ shear modulus of the undamaged material
H(x) ⫽ Heaviside function
I2 ⫽ second-order unit tensor
Iᠪ4 ⫽ fourth-order unit tensor
Mᠪ (D) ⫽ fourth-order damage operator
Mᠪ̆ (D) ⫽ fourth-order tensor to take into account the uni-
lateral effect
Pᠪ ( ␴,D)⫹ ⫽ fourth-order tensor corresponding to a positive
projection operator
Pᠪ ( ␴ ,D)⫺ ⫽ fourth-order tensor corresponding to a negative
projection operator
Tol ⫽ tolerance on the evolution law of damage
U e0 ⫽ complementary elastic energy
Up ⫽ energy blocked by the microcracks inducing per-
manent strain
Yi ⫽ thermodynamic force associated with D i
Ȳ i ⫽ weighed thermodynamic force associated to D i
Y0 ⫽ initial damage threshold
a ⫽ material parameter
b ⫽ material constant bounded between 0 and 1
ni ⫽ normal vector to the crack
␣, ␤ ⫽ unknown material parameters of the fourth-order
tensor A ᠪ
␦ ik ⫽ Kronecker delta symbol
␧ ⫽ second-order strain tensor
␧e ⫽ second-order elastic strain tensor
␧p ⫽ second-order permanent strain tensor
␧mp ⫽ measured second-order permanent strain tensor
␧tp ⫽ predicted second-order permanent strain tensor
␧ 11
p p
,␧ 22 ⫽ permanent strain in the 1- and 2-directions
␯ ⫽ Poisson’s ratio of the undamaged material
␯ 12 ⫽ Poisson’s ratio of damaged material
␴ ⫽ second-order Cauchy stress tensor
˜␴ ii⫹ ⫽ tensile stress 共positive兲 tensor normal to the
plane of the microcrack system
˜␴ ii⫺ ⫽ compressive stress 共negative兲 tensor normal to
the plane of the microcrack system
丢 ⫽ dyadic or tensor product
Fig. 12 Tensile–compressive load simulation: „a… Incremental 储•储 ⫽ Euclidean norm
axial stress level and „b… predicted damage variables evolution
„Model…

late damage evolution in structures made of random short glass

References
fiber reinforced composites in a proportional loading case. The 关1兴 Pensée, V., and Kondo, D., 2001, ‘‘Une Analyse Micromécanique 3-D de
l’Endommagement par Mésofissuration,’’ C. R. Acad. Sci., Ser. IIb: Mec.,
nonproportional loading case will be treated in a future paper. Phys., Chim., Astron., 329, pp. 271–276.
关2兴 Lu, T. J., and Chow, C. L., 1990, ‘‘On Constitutive Equations of Inelastic
Solids with Anisotropic Damage,’’ Theor. Appl. Fract. Mech., 14, pp. 187–
Acknowledgment 218.
The financial support provided for this research by the Natural 关3兴 Lemaı̂tre, J., Desmorat, R., and Sauzay, M., 1999, ‘‘Loi d’Évolution de
l’Endommagement Anisotrope,’’ C. R. Acad. Sci., Ser. IIb: Mec., Phys., Chim.,
Sciences and Engineering Research Council of Canada, ADS Astron., 327, pp. 1231–1236.
Groupe Composite, Bombardier and Prevost Car, is gratefully 关4兴 Halm, D., 1997, ‘‘Contribution à la Modélisation du Comportement Unilatéral
acknowledged. et du Frottement Dans les Matériaux Mésofissurés,’’ PhD thesis, Ècole Natio-
nale Supérieure de Mécanique et d’Aérotechnique et Faculté des Sciences
Fondamentales et Appliquées, France.
Nomenclature 关5兴 Kachanov, M., 1992, ‘‘Effective Elastic Properties of Cracked Solids: Critical
Review of Some Basic Concepts,’’ ASME Appl. Mech. Rev., 45, 共8兲, pp.
ᠪ ⫽ symmetric fourth-order tensor taking into ac-
A 304 –335.
count permanent strain 关6兴 Dano, M. L., Gendron, G., and Mir, H., 2002, ‘‘Mechanics of Damage and
គ̃ ⫺1 ⫽ fourth-order elastic compliance tensor of the
C Degradation in Random Short Glass Fiber Reinforced Composites,’’ Journal of
Thermoplastic Composite Materials, 15, 共2兲, pp. 169–177.
damaged material 关7兴 Maillette, F., 2002, ‘‘Caractérisation Expérimentale d’un Matériau Composite
D ⫽ second order damage tensor à Fibres Courtes et Orientés Aléatoirement,’’ Master’s thesis, Dept of Civil
D i ⫽ damage in the principal direction i Eng., Laval University, Quebec, Canada.
关8兴 Dano, M.-L., Maillette, F., Gendron, G., and Bissonnette, B., 2001, ‘‘Damage
Dm1 ⫽ measured damage variable in the 1-direction Modelling of Random Short Glass Fibre Reinforced Composites,’’ Proceeding
D t2 ⫽ predicated damage variable in the 1-direction of the Third Canadian Conference on Composites, Montreal, Canada, pp. 263–
E ⫽ Young’s modulus of the undamaged material 270.
关9兴 Mir, H., 2003, ‘‘Contribution à la Modélisation de l’Endommagement des
E 1 ⫽ secant damage Young’s modulus Matériaux Composites avec Fibres de Verre Courtes: Anisotropie Induite, Ef-
ERR – STR ⫽ permanent strain relative error fets Unilatéral et Résiduel,’’ PhD thesis, Department of Civil Engineering,

Journal of Applied Mechanics MARCH 2005, Vol. 72 Õ 257

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Laval University, Quebec, Canada. 关11兴 Chen, X. F., and Chow, C. L., 1995, ‘‘On Damage Strain Energy Release Rate
关10兴 Sidoroff, F., 1981, ‘‘Description of Anisotropic Damage Application to Elas- Y,’’ Int. J. Damage Mech., 4, pp. 251–263.
ticity,’’ IUTAM Colloquium, Physical Nonlinearities in Structural Analysis, J. 关12兴 Ciarlet, P. G., 1982, Introduction à l’Analyse Numérique Matricielle et à
Hult, ed., pp. 237–244. l’Optimization, Masson, Paris.

258 Õ Vol. 72, MARCH 2005 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use