Science Arts & Métiers (SAM)

is an open access repository that collects the work of Arts et Métiers ParisTech
researchers and makes it freely available over the web where possible.

This is an author-deposited version published in: http://sam.ensam.eu

Handle ID: .http://hdl.handle.net/10985/7934

To cite this version :

Muhamad Fatikul ARIF, Nicolas SAINTIER, Fodil MERAGHNI, Joseph FITOUSSI, Yves
CHEMISKY, Gilles ROBERT - Multiscale fatigue damage characterization in short glass fiber
reinforced polyamide-66 - Composites Part B: Engineering - Vol. 61, p.55-65 - 2014

Any correspondence concerning this service should be sent to the repository

Administrator : [email protected]
Multiscale fatigue damage characterization in short glass fiber reinforced
polyamide-66
M.F. Arif a, N. Saintier b,⇑, F. Meraghni a, J. Fitoussi c, Y. Chemisky a, G. Robert d
a
Arts et Métiers ParisTech, LEM3 – UMR CNRS 7239, 4 Rue Augustin Fresnel, 57078 Metz, France
b
Arts et Métiers ParisTech, I2M – UMR CNRS 5295, Esplanade des Arts et Métiers, 33405 Talence, France
c
Arts et Métiers ParisTech, PIMM – UMR CNRS 8006, 151 Boulevard de l’Hôpital, 75013 Paris, France
d
Solvay Engineering Plastics, Avenue Ramboz BP 64, 69192 Saint-Fons, France

a b s t r a c t

This paper aims at studying fatigue damage behavior of injection molded 30 wt% short glass fiber rein-
forced polyamide-66 composite (PA66/GF30). The evolution of dynamic modulus, hysteresis area, cyclic
creep and temperature during fatigue tests were analyzed and discussed. Damage analyses by X-ray
micro-computed tomography (lCT) technique on interrupted fatigue tests at several percentages of total
fatigue life were performed to further understand the damage mechanisms and evolution during fatigue
loading. It can be observed that experimental results related to the evolution of dynamic modulus, strain,
Keywords:
temperature and energy dissipation are important and consistently complement each other for damage
A: Polymer–matrix composites (PMCs)
B: Fatigue
evaluation of PA66/GF30. During fatigue loading, diffuse damage occurs over the entire specimen though
B: Microstructures the damage does not necessarily exhibit the same level between different locations inside the specimen.
E: Injection molding The lCT analysis of voids characteristics demonstrates that the damage continuously increases during
Micro-computed tomography fatigue loading. The damage is developed notably along fiber interface in the form of fiber/matrix inter-
facial debonding.

1. Introduction teresis curve, has been proposed by several authors to evaluate

the damage level of the composites [3–8]. Since the damage pro-
The reduction of vehicle mass is a major concern for automotive cess is thermally activated and, in the same time, a dissipative pro-
industries to comply with the strict pollution regulation, particu- cess by nature, thermography technique by using infra-red camera
larly for the CO2 emission. Short fiber reinforced thermoplastic has become an important tool for fatigue damage evaluation in
materials, among them is the short glass fiber reinforced polyam- composites [9–14]. Finally, due to the complex nature of the micro-
ide-66, are good candidates to provide a compromise between structure, damage is spatially distributed so that X-ray micro-com-
the required lightweight and the expected thermomechanical per- puted tomography (lCT) technique has become a suitable and
formances. However, their structural durability has not yet been efficient tool for fatigue damage characterization in composite
fully investigated. In particular further work to investigate the fa- materials [15–17]. The work of Cosmi and Bernasconi [18] presents
tigue behavior of these composites is necessary to provide physi- the potential and critical aspects of using lCT technique for fatigue
cally based damage mechanisms scenarios for fatigue modeling. damage characterization in short glass fiber reinforced polyamides.
Various techniques have been used to evaluate the damage in The current work is the continuation of the recently published
short fiber reinforced thermoplastics. Early works of Horst and work on damage characterization of PA66/GF30 under quasi-static
Spoormaker [1,2] performed fractography analysis onto the frac- loading by SEM and lCT techniques [19].
ture surface of fatigue loaded short glass fiber reinforced polya- During fatigue loading of the composite, several phenomena
mides by scanning electron microscopy (SEM) and they proposed may develop concurrently, such as damage, cyclic creep and in-
a damage mechanisms scenario which consider that the damage crease of temperature. All of them may contribute to the overall fa-
starts at fiber ends and then propagates along fiber interface. The tigue strength of the material [20,21]. A comprehensive study of
evolution of dynamic modulus, i.e. the slope of stress–strain hys- fatigue damage behavior is necessarily a coupled analysis of all
interrelating phenomena during fatigue loading. The objective of
⇑ Corresponding author. Tel.: +33 556 845 361; fax: +33 556 845 366. this work is to characterize fatigue damage mechanisms of PA66/
E-mail address: [email protected] (N. Saintier). GF30 under uniaxial constant amplitude loading. It is proposed
to use a combined analysis of dynamic modulus, cyclic creep, dis- the initial and final filling stages. Indeed, nonuniform microstruc-
sipated energy and temperature evolution during fatigue testing ture is usually found at these complex flow zones. Moreover, the
together with post-mortem 3D damage analysis by lCT to further selected positions were also used as a guideline for specimens
understand the damage mechanisms of PA66/GF30 during fatigue extraction for the tensile and fatigue tests that will be discussed
testing. To evaluate the process induced anisotropy, two directions, in Section 3. The dimensions of the lCT samples were
longitudinal and transverse to the mold flow direction (MFD), are 2  2  3:2 mm3 , where 3.2 mm corresponds to the sample thick-
examined. In addition, a particular attention is given to the effect ness. The lCT scanning was not carried out through all the sample
of skin–shell–core formation on the damage mechanisms of thickness but it always covered more than half of the thickness so
PA66/GF30. In fact, this type of microstructure is commonly ob- that the skin–shell–core structure can always be captured.
served in the injection molded short glass fiber reinforced polya- The microstructure of PA66/GF30 obtained from the lCT sam-
mides [22–25]. The experimental results allow to analyze the ple extracted from position A of Fig. 1 is shown in Fig. 2. It can
importance of damage and their interrelated phenomena to the be seen from Fig. 2, PA66/GF30 has a specific injection molding
overall behavior of PA66/GF30, as well as to the identification of process induced microstructure characterized by a skin–shell–core
damage mechanisms and evolution. structure. Moreover, transition layers between shell and core are
In summary, the organization of this work is as follows: Section also observed. The microstructure of PA66/GF30 extracted from
2 presents material and process induced microstructure descrip- position B exhibits qualitatively the same trend as the one ex-
tion, followed by the specimens used for mechanical characteriza- tracted from position A. This microscopic condition was achieved
tion in Section 3. In Section 4, the experimental procedures and by an optimal set up of injection molding parameters (barrel and
damage investigation techniques are detailed. In Section 5, the re- mold temperatures, injection speed, etc.) to ensure a high compac-
sults of microscopic and macroscopic damage analysis of PA66/ tion degree of PA66/GF30 and to reduce the core layer thickness.
GF30 are discussed. Afterwards, the damage characteristics and To acquire the real tendency of fiber orientation throughout the
all the contributing factors to the overall fatigue strength, as well samples, the tensor representation of fiber orientation state pro-
as the damage mechanisms chronology of PA66/GF30 are con- posed by Advani and Tucker [27] is used in this work. This method
cluded in Section 6. is widely used to characterize the fiber orientation due to its effi-
ciency, simplicity and relatively short computational time. The
complete review of this method can be found elsewhere [28–30].
2. Material and process induced microstructure description A strategic method of lCT 3D image segmentation of the fibers
via gray level thresholding has been developed to capture effi-
The material studied is a 30 wt% short glass fiber reinforced ciently the Euler orientation angles of the fibers. Avizo and Visilog
polyamide-66 composite (PA66/GF30) provided by Solvay Engi- softwares were used for this purpose. The second-order fiber orien-
neering Plastics-France, under the commercial name of TechnylÒ tation tensors have been computed based on the data of fibers’ Eu-
A218V30. The material was prepared by compounding the polyam- ler angles. The normalized first tensor components (a11 ) of the two
ide-66 pellets and chopped short glass fibers in a twin-screw ex- studied lCT volumes are shown in Fig. 3. The a11 tensor component
truder. Subsequently, PA66/GF30 compound was transferred into represents the degree of fibers to orient longitudinal (parallel) with
an injection molding machine, resulting in 360  100  3:2 mm3 respect to the MFD. Higher value of a11 constitutes higher tendency
of rectangular plate. of fibers to orient to the MFD. It can be seen from Fig. 3 that the
It is commonly established through open literature [22–25] that skin–shell–core layers of the two lCT samples are well-defined.
fiber orientation distribution is widely governed by injection pro- In addition, small transition regions between the shell and core
cess. In fact, for a system where polymer melt flows through a layers are also observed.
small cavity between two parallel walls, shear flow has maximum Based on the qualitative and quantitative analysis by lCT, it can
value near to the mold wall whereas it vanishes at the core zone. be seen that the skin layers, the upper and lower specimen sur-
This leads the fibers to be oriented parallel (and perpendicular) faces, are slightly random in fiber orientation. However, the orien-
to MFD at the shell (and core) layers. Moreover, a thin random skin tation tensor (Fig. 3) shows that the principal fiber orientation in
layer can be formed due to the polymer melt that is in direct con- the skin layer tends to follow the MFD. The shell layers are the
tact with a relatively cold mold wall temperature. most dominant ones with preferential fiber orientation longitudi-
X-ray micro-computed tomography (lCT) technique was em- nal to the MFD. Thin core layer is developed and mostly fibers in
ployed to study the microstructure heterogeneity of PA66/GF30 this layer are oriented transversely to the MFD. It is worth noting
in terms of its fiber orientation. The lCT experiment was carried that instead of purely oriented transversely to the MFD, it has been
out at ID19 beam line of the European Synchrotron Radiation Facil- frequently observed that fibers in the core layer are slightly tilted
ity (ESRF) Grenoble, France [26]. For this microstructure investiga- around 5–20° from the transverse direction, such as the one shown
tion, the experimental setup was conditioned to reach a voxel in Fig. 2d. In addition to the skin, shell and core microstructures,
resolution of 1.4 lm. Two samples for lCT experiments were ex- transition layers between shell and core are also observed, as
tracted from the injection molding plate at two positions along shown in Figs. 2c and 3.
mold flow direction (MFD) axis, as shown in Fig. 1. The selected The normalized orientation tensors of the two studied samples
positions were significantly far from the complex flow zones at show that both positions exhibit almost the same state of fiber ori-
entations. Only small differences on the width of the shell–core
transition zones and the orientation state of the skin layers are ob-
served between the two studied samples. It can be inferred from
the results that the microstructure developed along the MFD axis
of the injection molding plate is generally homogeneous.

3. Specimen

Fig. 1. Locations of lCT sample extraction for injection molding induced micro- Specimens used for mechanical tests were machined from the
structure investigation (A and B). rectangular plate produced by injection molding. To consider the
 Fig. 2. Skin–shell–transition–core microstructure formation of PA66/GF30 observed by lCT technique.

(a)
transition

transition
Normalized first tensor component (a11)

core

shell shell
skin

skin

0.5 Evaluated a11

(b)
(0−2200µm)
A
B

0 500 1000 1500 2000 2500 3000

Thickness (μm)
Fig. 4. Machining locations of (a) longitudinal and (b) transverse specimens, as well
Fig. 3. Normalized first tensor component ða11 Þ of the two studied samples. The a11
as the main specimen dimensions of PA66/GF30.
is evaluated from 0 to 2200 lm thickness. The a11 above 2200 lm is assumed to be
symmetric with another half of the thickness.
4. Experimental procedures and damage investigation
techniques
anisotropy effect induced by the skin–shell–transition–core struc-
ture of the composite, the specimens were machined in longitudi- Tensile and fatigue tests were performed on dry as molded
nal and transverse directions with respect to the MFD, defined PA66/GF30 (<0.2% water content) at room temperature. Quasi-sta-
hereafter as the longitudinal and transverse specimens. The longitu- tic tensile tests were performed upon a servo-hydraulic machine at
dinal and transverse specimens were machined from several posi- a crosshead speed of 1 mm/min, which corresponds to the strain
tions of the injection molding plate as shown in Fig. 4. To ensure rate of 3  104 s1 . Fatigue tests were performed by applying a
microstructure homogeneity between all specimens, the specimen sinusoidal signal, under load controlled mode, at constant ampli-
extraction locations were chosen far from the injection gate and tude and fixed loading ratio. To prevent specimens from buckling,
end-filling zones. Moreover, the specimen gauge-length zones the fatigue tests were conducted under tension–tension mode,
were assured to be along the MFD axis. As discussed earlier, the with a stress ratio of R ¼ rmin =rmax ¼ 0:1. The frequency of 3 Hz
microstructure developed along the MFD axis of the injection was chosen in order to avoid a global heat of the sample reaching
molding plate is generally homogeneous. The machining positions the glass transition of dry PA66/GF30 (65–70 °C) during fatigue
of the longitudinal and transverse specimens, as well as the spec- loading. Fatigue strength of the material was evaluated within
imen main dimensions are presented in Fig. 4. During mechanical the range of 103 —106 cycles. Continuous temperature monitoring
tests, the longitudinal and transverse specimens were selected ran- of the active zone of the specimen was assured by the CEDIP Jade
domly from these locations.
III MWR infrared camera with a spectral range between 3.9 and thickness of the specimens are predominantly occupied by the
4.5 lm. shell layer where the fibers are longitudinally oriented to the
Damage mechanisms and its evolution were investigated by the MFD. This yields the shell layer to dominantly govern the tensile
use of lCT analyses performed on longitudinal and transverse behavior of PA66/GF30. Moreover, Fig. 5 illustrates different data
specimens after being subjected to fatigue loading up to several scatter between longitudinal and transverse specimens. Transverse
percentages of total fatigue life. For this damage investigation, specimens demonstrate a much larger data dispersion than that in
the lCT experimental setup was conditioned to reach voxel resolu- longitudinal specimens. This is probably due to the fiber orienta-
tions of 0.7 and 1.4 lm. The longitudinal specimens were consis- tion variability in the core layer. As previously discussed in Section
tently carried out at 0.7 lm resolution, whereas the transverse 2, instead of exclusively oriented transversely to the MFD, tilted fi-
specimens were employed at 1.4 lm resolution. Both resolutions bers around 5° to 20° from the transverse direction are frequently
demonstrated consistent results though the ones with 0.7 lm pro- observed. Since the core layer in the transverse specimens is
vided stronger physical interpretations. The lCT sample dimen- responsible for the load transfer mechanism due to its parallel ori-
sions for this damage investigation were the same as the ones for entation to the applied load direction, the core layer orientation
the microstructure analysis previously discussed in Section 2. variability may lead to a data scatter in the mechanical properties
The lCT samples were machined from the central gauge-length of the transverse specimens.
zone of the fatigue loaded specimens. For the specimens that have
been fatigue loaded up to failure, the extraction locations of the
5.1.2. Strain rate evolution during fatigue test
lCT samples were far from the fracture surface in order to avoid
The strain rate during fatigue loading can be estimated using
fast crack propagation effect due to the final failure. The lCT inves-
the derivative of the strain equation. If one considers that the strain
tigation on virgin sample was also carried out as a reference for the
response of the composite follows a sinusoidal wave equation
damage analysis of fatigue loaded specimens. By segmentation
eðtÞ ¼ De sinðxtÞ þ e, the strain rate will be e_ ðtÞ ¼ Dex cosðxtÞ,
technique with proper and consistent selection of threshold value,
where e and e_ are respectively the strain and strain rate (s1 ), De
the voids inside the analyzed lCT volume element can be isolated,
and e are respectively the strain amplitude and mean strain,
thereby the void features such as volume, Euler orientation angles
x ¼ 2p=T is the angular frequency (rad/s), T is the period (s) and
and aspect ratio can be quantified. The threshold level was defined
t is the time (s). The description of these variables is given in Fig. 6.
and optimized based on visual observation of the voids at fatigue
According to the above-mentioned strain rate equation, the
loaded specimens. All the specimens, including the virgin, were
composite experiences a continuous change of strain rate during
treated using similar and constant threshold level in order to allow
cyclic loading. The maximum strain rate during a cycle is given
a direct comparison between them. In this work, lCT images are
by e_ max ¼ Dex. Evaluating only the maximum value of strain rate,
presented to demonstrate the fatigue damage mechanisms of the
one can see from Fig. 7 that the strain rates in longitudinal and
composite. The void volume, aspect ratio and Euler orientation an-
transverse specimens are ranging from 7  102 to 2  101 s1
gles are also presented to confirm the damage mechanisms and
even though the maximum stress levels of both specimens are very
evolution of PA66/GF30.
different. The strain rates in all specimens are nearly constant dur-
ing fatigue loading except the one of the transverse specimen
5. Experimental results and discussion loaded at the highest maximum stress level, which could be due
to the high heat generation and thus higher viscous properties of
5.1. Overall material characteristics PA66/GF30. It has to be noted that the range of the strain rate in
the fatigue test of PA66/GF30 is significantly higher than that in
5.1.1. Tensile properties the quasi-static tensile test. Tensile tests performed at a strain rate
The tensile properties of PA66/GF30 are described in Fig. 5. This of about 101 s1 led to a fracture before reaching the ultimate
figure illustrates the strong anisotropy effect induced by the injec- stress commonly achieved at quasi-static strain rates. For these
tion molding process. The Young modulus and ultimate stress of reasons, fatigue load levels will be given by referring to the ulti-
the longitudinal specimens are twice than the ones of the trans- mate stress obtained from the quasi-static tensile tests.
verse specimens. Nevertheless, strain to failure characterizing the
material ductility of the longitudinal specimens is half than that 5.1.3. Fatigue S–N curve
of the transverse specimens. The fiber orientations through the Fatigue life (S–N) curve of PA66/GF30 is reported in Fig. 8. In
line with the results of quasi-static tensile tests, the fatigue
strength of longitudinal specimens are about twice higher than
that of transverse specimens. It is also shown that the S–N curves
1
Longitudinal
0.8
T
Normalized Stress

0.6 Δε
Transverse

0.4

0.2

0 0
0 1 2 3 4 5 6 t (s)
Strain (%)
Fig. 6. Schematic of sinusoidal wave form of strain response of material under
Fig. 5. Tensile properties of longitudinal and transverse specimens of PA66/GF30. fatigue loading.
 endurance limit has been reached at those stress levels and that no
damage occurs below the corresponding stress levels. According to
the published works on fatigue behavior of short glass fiber rein-
forced polyamides, one cannot reach a distinctive endurance limit
−1
stress up to 106 cycles [1,20,31], or even up to 107 cycles [32,33].
Max strain rate (s−1)

10

5.2. Macroscopic damage evaluation

σ = 65% σ 5.2.1. Thermo-elasto-visco-damage coupling

max u
σmax= 60% σu Fig. 9 illustrates the evolution of the monitored parameters, i.e.
normalized dynamic modulus (EN =E0 ), maximum strain (emax ) and
σmax= 55% σu
mean temperature (T mean  T room ) for the longitudinal and trans-

−2
(a) σ
max
= 50% σ
u verse specimens. During cyclic loading, energy dissipation can be
10 1 2 3 4 5 6
associated to different phenomena such as damage development
10 10 10 10 10 10
and intrinsic dissipation (viscous behavior). Part of the mechanical
Number of cycles
strain energy due to the damage development and viscous effect of
the material is turned into heat so that thermo-elasto-visco-dam-
age coupling can occur during fatigue loading.
The dynamic modulus reduction can be used as a damage indi-
cator when considering a classical continuum damage mechanics
framework [6,34]. For all loading cases encountered in this study,
Max strain rate (s )

−1
−1

10
the normalized dynamic modulus evolution demonstrates a stable
value for the first 103 cycles and then decreases at a rate that de-
run−outs pends on the loading level. In all cases, the intensity of the decrease
of the normalized dynamic modulus is directly related to the fati-
σmax= 70% σu gue life of the specimen.
σmax= 60% σu The evolution of mean temperature exhibits two regimes
σmax= 50% σu (Fig. 9). The first one corresponds to a stable normalized modulus,
where heat dissipation seems to be mostly related to the intrinsic
−2
(b) σmax= 40% σu
energy dissipation associated to the viscous nature of the compos-
10
10
1
10
2
10
3
10
4
10
5
10
6 ite. The second regime is associated to the onset of the decrease of
Number of cycles the normalized dynamic modulus and it corresponds to an inflec-
tion point on the mean temperature evolution. This regime change
Fig. 7. Evolution of maximum strain rate of (a) longitudinal and (b) transverse
can be associated to the fact that the strain energy is not only dis-
specimens during fatigue loading of PA66/GF30. (ru represents the ultimate tensile
strength of its respective orientation angles). sipated into heat due to the viscous nature of the composite but
also dissipated into a damage development and accumulation.
As shown in Fig. 9, the maximum strain continuously increases
1 during fatigue life. The minimum strain exhibits the same trend
0.9
0.8 (Fig. 10) so that the strain amplitude is almost constant during fa-
0.7 tigue loading. It can then be stated that the observed damage is
0.6 associated to a combined effect of creep and cyclic loading. How-
Normalized maximum stress

0.5 ever, since the mean strain evolution is maintained in reasonable

0.4 ranges for most of the fatigue tests (lower than 1%), and that the
dynamic modulus reduction is not associated to a change in the
0.3 mean strain evolution, which could indicate an acceleration of
the creep damage, it is believed that most the observed damage
0.2 is related to the cyclic component of the loading and that the creep
tests stopped damage is limited for these dry as molded specimens. The situation
Longitudinal is different for the transverse specimen loaded at the highest max-
Transverse imum stress level. At such high stress level, thermal softening is
0.1 2 3 4 5 6
considered to arise as revealed by the significant increase of the
10 10 10 10 10 maximum strain at the end of the fatigue life. It is also found a
Number of cycles to failure
small evolution of dissipated strain energy during the fatigue tests
Fig. 8. The S–N curves of longitudinal and transverse specimens. Maximum stresses and the decrease in the dynamic modulus is not associated to a sig-
are normalized with the ultimate tensile strength of longitudinal specimen. nificant modification in the intrinsic dissipation characteristics of
the material (Fig. 11). Therefore, it can be inferred that the dynamic
modulus reduction is related to a high extent to the damage evolu-
are linear in terms of maximum stress vs. number of cycles in a tion, except for the highest loading level of the transverse speci-
log–log representation. The current result is consistent with the men ðrmax ¼ 70%ru Þ where thermal softening may occur and
previously published work in short glass fiber reinforced polya- would lead to the final failure.
mides [22,31,32]. It is worth noting that for the low stress levels The longitudinal and transverse specimens exhibit different
of transverse specimens, the fatigue tests were stopped if the spec- behavior considering that the thermo-elasto-visco-damage cou-
imens did not reach final failure at 5  105 cycles. Stopping the fa- pling in longitudinal specimens is higher than that in transverse
tigue tests at 5  105 cycles was a compromise that allow the tests specimens. With the stress levels significantly lower than those
to be performed in a reasonable time with significant evolution of in longitudinal specimens, the transverse specimens generate
the observable variables. However it does not mean that a fatigue higher changes in dynamic modulus, strain and temperature. This
 (a) 1 (b) 1
0.95
0
E /E 0.98
0.9
N
0.96 run−outs

0
0.85

E /E
0.94 σmax= 65% σu
σ = 70% σ

N
0.8
σ = 60% σ max u
max u
0.75 σ = 60% σ
σmax= 55% σu max u
0.7 σ = 50% σ
4 σmax= 50% σu max u
0.65 σmax= 40% σu
3.5
3 4
(%)

2.5

(%)
max

3
2
ε

max
1.5 2

ε
1
1

30
10
25

(°C)
(°C)

7.5 20

room
room

15
5
−T
−T

10
mean
2.5
mean

5
T
T

0 0

1 2 3 4 5 6 1 2 3 4 5 6
10 10 10 10 10 10 10 10 10 10 10 10
Number of cycles (N) Number of cycles (N)

Fig. 9. Evolution of normalized dynamic modulus (EN =E0 ), maximum strain ðemax Þ and mean temperature ðT mean  T room Þ of (a) longitudinal and (b) transverse specimens
during fatigue loading of PA66/GF30. (ru represents the ultimate tensile strength of its respective orientation angles).

4 lCT are necessary to further understand the fatigue damage in

ε PA66/GF30.
max
3.5
εmin
3 5.2.2. Spatial distribution of damage
In a material discontinuity such as damage, heat transfer is hin-
2.5
dered since there is only partial or no contact between the internal
Strain (%)

2 surfaces of the damage zone. This could result in a local increase of

temperature in the damage zone. If one records the mean and max-
1.5 imum temperature on the active zone area of the specimen, it is
expected that the temporal increase of the maximum temperature
1
will be higher than that of the mean temperature when a localized
0.5 damage is detected. However, the nature of this technique suggests
that the level of localized damage should be high enough so that
0 1 2 3 4 5 6 the heat generated in the damage zones can reach the surface
10 10 10 10 10 10
Number of cycles and thus can be captured by the infrared thermography camera.
This experimental approach has been examined in PA66/GF30
Fig. 10. Evolution of maximum and minimum strain of longitudinal specimen specimens. As described in Fig. 12, the maximum temperature evo-
fatigue loaded at maximum stress of 65% ru . All loading cases encountered in this lution follows that of the mean temperature, with approximately a
study exhibit similar trend as emin follows emax at approximately a constant offset.
constant offset. This result indicates that during fatigue loading,
there is no macroscopic damage induced strain localization and
is due to the fact that in transverse specimens, the polyamide-66 the damage is preferably diffused over the entire specimen. An in-
matrix has a more important role than that of fibers during the fa- crease of maximum temperature is only observed when the num-
tigue loading. ber of cycles reaches 99% of fatigue life. This is due to the sudden
It can be summarized that the information of dynamic modulus crack propagation leading to the final fracture of the sample.
is important though it becomes more difficult to completely com-
prehend the fatigue damage behavior without the information of 5.3. Microscopic damage evaluation
strain, temperature and dissipated energy evolution. The evolution
of dynamic modulus can be considered as a damage indicator. The lCT is a powerful 3D non-destructive investigation method
However for high stress levels, it may overpredict the damage evo- to analyze microscopic characteristics of a medium. By applying 3D
lution due to the high viscous effect contribution of the composite, reconstruction, fibers in a composite can be observed and further
such as the one shown in the highest loading level of the transverse post-processed to obtain their features such as volume, aspect ratio
specimen. Therefore, the spatial distribution analysis of damage and orientation angles. Also, voids that are the characteristic of
and the microstructural investigation of damage mechanisms by damage can be observed and quantified. Indeed, in fatigue loaded
 (a) 2.2 σmax= 65% σu (a) 14 T
max
−T
room
2 σ = 60% σ Tmean−Troom
max u 12
σmax= 55% σu
Normalized energy dissipation

1.8 10
σ = 50% σ
max u σmax=60% σu
1.6 σmax=65% σu

ΔT (°C)
8
1.4
6
1.2 σ =55% σ
max u
4
1
2
0.8 σ =50% σ
max u
0
0.6
1 2 3 4 5 6
1 2 3 4 5 6 10 10 10 10 10 10
10 10 10 10 10 10 Number of cycles
Number of cycles

(b) 40 T
max
−T
room
(b) 2.2 σ
max
= 70% σ
u
35 T
mean
−T
room
2 σ = 60% σ 30
max u σmax=70% σu run−outs
σ = 50% σ
Normalized energy dissipation

1.8 max u 25
σ = 40% σ

ΔT (°C)
max u 20
1.6
run−outs 15
1.4
10 σmax=60% σu
1.2
σ =50% σ
5 max u
1
0
σmax=40% σu
0.8
1 2 3 4 5 6
0.6 10 10 10 10 10 10
Number of cycles
1 2 3 4 5 6
10 10 10 10 10 10
Number of cycles Fig. 12. Evolution of mean and maximum temperature of (a) longitudinal and (b)
transverse specimens during fatigue loading of PA66/GF30.
Fig. 11. Evolution of normalized energy dissipation of (a) longitudinal and (b)
transverse specimens during fatigue loading of PA66/GF30. The energy dissipation
was obtained by evaluating the hysteresis area at each loading cycle normalized by
the initial value.

specimens of PA66/GF30, the appearing of zones darker than the

matrix is the characteristic of voids. By comparison with virgin
specimen, the presence of these dark zones can be associated to fa-
tigue damage.
In virgin specimen, almost no visible damage is observed inside
the sample, as shown in Fig. 13. Few zones darker than the matrix
are observed in the virgin sample, which could be due to the real
initial damage or due to the image artifact. In fatigue loaded spec-
imens, these dark zones are frequently observed, notably along fi-
ber interface. These zones correspond to the damage in the form of
fiber/matrix interfacial debonding, as shown in Fig. 14. However,
one cannot asses whether the local damage is adhesive or cohesive
from this lCT investigation. Indeed, fiber/matrix interfacial deb-
onding depends on several parameters, among them are relative
humidity and fiber orientation. Voids at fiber ends are also ob-
served in the fatigue loaded specimens, though it is not necessarily
involved in the fiber/matrix interfacial debonding. Few fiber break-
ages are also observed but it seems this damage mechanism is not
dominant. In a particular case, matrix microcracks can be devel-
oped, such as the one observed in the core layer of the transverse
specimen (Fig. 15). The relatively thin core layer is believed to bear
Fig. 13. The lCT image of virgin sample.
higher stress level due to its longitudinal (parallel) orientation to
the applied load direction. Matrix microcracks, with preferential
direction transversely to the applied load direction, as well as to tation process of lCT volumes has been carried out to allow a joint
the fibers direction, are found to develop favorably in this region. representation of the damage and fibers, each of them having dif-
Since the presence of the damage can be in the form of nano- ferent colors, and without the matrix on the final images. Fig. 16
scopic voids, visual demonstration by the original lCT radiographs shows an example of magnified image in a zone with highly deb-
is sometimes not sufficient. A further treatment after the segmen- onded fibers. This fiber/matrix interfacial debonding is frequently
Fig. 14. Damage mechanisms observed in the shell zone of longitudinal specimen that has been fatigue loaded up to failure at maximum stress of 60% ru ; (a) fiber/matrix
interfacial debonding, (b) void at fiber ends, and (c) fiber breakage.

the entire specimen, the damage level between different locations

inside the specimen is not necessarily the same (Fig. 17).
To propose a first attempt to quantify the damage and build a
more consistent analysis on damage mechanisms of PA66/GF30,
the void aspect ratio vs. void orientation angles are analyzed. To re-
duce the number of analyzed voids and to establish a consistent
comparison between the void properties in the shell and core lay-
ers, the evaluation of void aspect ratio has been carried out in a
50 lm volume slice of the shell and core layers of the lCT volume,
as described in Fig. 18. The lCT result of virgin specimen is given in
Fig. 19 in the form of polar graph. For the virgin specimen, voids
with low aspect ratio are observed in the shell layer. The trend in
the core layer is the same as the one noticed in the shell layer.
These small voids could be partly due to the real initial damage
and partly due to the gray level fluctuation-induced inherent arti-
facts as a result of the thresholding technique. It should be noted
that the thresholding technique was defined and optimized based
on visual observation of voids in the fatigue loaded specimens. The
Fig. 15. Matrix microcracks observed at the core zone of transverse specimen that chosen threshold level is unique and applied to all specimens,
has been fatigue loaded up to failure at maximum stress of 60% ru . including the virgin sample. By this method, one can ensure that

Fig. 16. A rendered lCT image of a zone with highly debonded fibers in the shell
layer of longitudinal specimen that has been fatigue loaded up to failure at
maximum stress of 60% ru . The blue and green colors represent the fiber and
damaged zone, respectively. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)
Fig. 17. A rendered lCT image which shows heterogeneous damage level at
microscopic scale (the highlighted zone shows higher damage). The image was
taken from longitudinal specimen that has been fatigue loaded up to 75% of fatigue
observed at the fatigue loaded specimens and can be considered as
life at maximum stress of 60% ru . The blue and magenta colors represent the fiber
the main fatigue damage mechanism for PA66/GF30. In terms of and damaged zone, respectively. (For interpretation of the references to color in this
spatial distribution of damage, though the damage is diffused over figure legend, the reader is referred to the web version of this article.)
 of the composite. The random skin layers in the longitudinal and
transverse specimens exhibit the same behavior as those in the
shell layers. Even though the thin random skin layers are devel-
oped, the degree of random orientation is not high and the fibers
in this layer tend to orient according to MFD, which is the same ori-
entation as the ones in the shell layer (Fig. 3). For the transverse
specimen, less fiber/matrix interfacial debonding is observed. Thi is
probably due to the high development of matrix microcracks in
addition to the fiber/matrix interfacial debonding as observed in
the 2D images (Figs. 14 and 15). These figures illustrate clearly
the anisotropic nature of the damage developed at the local scale.
In addition to the final stage damage investigation, similar lCT
analyses were performed on longitudinal specimens that have
Fig. 18. Description of the analyzed lCT volumes in the shell and core regions of
the composite.
been fatigue loaded up to several percentages of fatigue life.
Fig. 22 illustrates the evolution of void volume as a function of void
orientation and number of cycles at the shell layer of longitudinal
90 30 specimens that have been fatigue loaded up to several percentages
120 60 of fatigue life. It clearly shows that the void volumes increase
20 throughout the fatigue life. It is worth noting that voids with low
150 30 volumes are also observed in the virgin sample, which could be
10 due to the real initial damage or fluctuation of the gray level. The
majority of voids with high volume in the shell layers are oriented
180 0 at 0°, longitudinal to the fiber orientation as well as to the applied
load direction. This signifies that the voids, notably with high vol-
umes, are located along fiber interface in the form of fiber/matrix
210 330 interfacial debonding. These results are consistent with the aspect
ratio analyses performed previously. In terms of evolution, the
240 300 damage increase shows little evolution in the first 0–50%. However
270 it evolves more significantly in the second half of the total fatigue
life. It is important to notice that this evolution corresponds to the
Fig. 19. Void aspect ratio vs. void orientation angle in the shell zone of virgin
specimen.
dynamic modulus evolution given in Fig. 9. If one considers the
change in the total void volume DV with respect to the initial void
volume, one can obtain that DV 50%Nf < 0:5%; DV 75%Nf ¼ 1:3% and
any deviations of the subsequent void aspect ratio and volume DV 100%Nf ¼ 2:5%. This evolution agrees very well with the dynamic
analyses for the fatigue loaded specimens are not due to the thres- modulus evolution in the final stage. However the lCT analyses
holding technique but mostly related to the damage development. were not able to capture the damage evolution at the early stage
Figs. 20 and 21 illustrate the void aspect ratio vs. angular posi- of fatigue life. This is most probably due to the fact that opening
tion of longitudinal and transverse specimens, respectively, that of the interface above the resolution limit is needed to make it pos-
have been fatigue loaded up to failure at maximum stress level sible to be detected by the lCT analyses.
of 60% ru . Significant damage is observed when comparing these In terms of spatial distribution of the damage, based on the
results to the one of the virgin material (Fig. 19). In longitudinal macroscopic characterization technique discussed in Section
specimen, the voids in the shell and core layers are mainly oriented 5.2.2, the damage is preferentially diffused over the entire speci-
at 0° and 90°, respectively (Fig. 20). It is worth noticing that these men. However, based on the microscopic observation by lCT tech-
orientations are the same as the principal fiber orientations in the nique, though the diffuse damage is observed, the damage level
shell and core layers. It means the voids are mainly located along between different locations inside the specimen is not necessarily
fiber interface in the form of fiber/matrix interfacial debonding. the same. Quantification of the local fiber density and its effects on
For the transverse specimen, the voids in the shell and core layers the fiber/matrix interfacial debonding process is underway and
are both oriented at 90°, as shown in Fig. 21. It shows that fiber/ will be the subject of a forthcoming paper.
matrix interfacial debondings occur in the shell layer, while matrix Finally, it can be summarized that the observed voids can be
microcracks with preferential direction transversely to the applied associated in a large extent to the fiber/matrix interfacial debond-
load as well as to the fibers direction are dominant in the core layer ing, complete or partial depending on the local configuration of the

(a) (b) (c)

Fig. 20. Void aspect ratio vs. void orientation angle in (a) shell and (b) core regions of fatigue loaded longitudinal specimen. Laminate representation analogy of longitudinal
specimen is presented in (c) for an interpretation guideline of void orientation angle.
 (a) (b) (c)

Fig. 21. Void aspect ratio vs. void orientation angle in (a) shell and (b) core regions of fatigue loaded transverse specimen. Laminate representation analogy of transverse
specimen is presented in (c) for an interpretation guideline of void orientation angle.

8000 8000
0% N 50% N
f f
Void volume (µm )

Void volume (µm )

3

3
6000 6000

4000 4000

2000 2000

0 0
−90 −45 0 45 90 −90 −45 0 45 90
Void orientation angle Void orientation angle

8000 8000
75% Nf 100% Nf
Void volume (µm )

Void volume (µm )

3

6000 6000

4000 4000

2000 2000

0 0
−90 −45 0 45 90 −90 −45 0 45 90
Void orientation angle Void orientation angle

Fig. 22. Void orientation angle vs. void volume graphs of the upper shell layer of longitudinal specimens that have been fatigue loaded up to several percentages of fatigue life
(N f ) at maximum stress level of 60% ru .

fibers. However, it is difficult to quantify the degree of interfacial microcrack with preferential direction transversely to the applied
debonding as the fiber length is not unique. Indeed, fiber length load is dominant due to the locally high stress concentration.
variability is developed due to the injection process. Matrix micro- The lCT results also demonstrate that the damage continuously
cracks at high stress level have been observed but do not seems to increases during fatigue loading. However, the damage evolution
be dominant in the damage process for dry as molded PA66/GF30. occurs more significantly in the second half of the fatigue life. De-
spite high resolutions used in this work, detection of fiber/matrix
interfacial debonding in the early stages of fatigue life was difficult.
6. Concluding remarks Observation with higher resolution is then required to further
investigate these early stages.
The macroscopic and microscopic fatigue damage behavior of The observed damage mechanisms and evolution are currently
PA66/GF30 have been studied. The experimental results related integrated into a micromechanical modeling in order to propose a
to the evolution of dynamic modulus, strain, temperature and en- physically based, microstructure dependent fatigue model. Further
ergy dissipation are important to evaluate the damage evolution. work on spatial damage distribution by lCT analysis is underway
The dynamic modulus can be used as a damage indicator, though and will be presented in a forthcoming paper.
in high stress level it may overpredict the damage evolution due
to the high viscous effect contribution of the composite.
Based on thermography analysis on PA66/GF30 surface, diffuse
damage occurs over the entire specimen. However, the micro- References
scopic observation by lCT technique states that the diffuse dam-
[1] Horst JJ, Spoormaker JL. Mechanisms of fatigue in short glass fiber reinforced
age does not necessarily exhibit the same level between different polyamide 6. Polym Eng Sci 1996;36(22):2718–26.
locations inside the specimen. [2] Horst JJ, Spoormaker JL. Fatigue fracture mechanisms and fractography of
The lCT analysis shows that the damage is mainly developed short-glassfibre-reinforced polyamide 6. J Mater Sci 1997;32(14):3641–51.
[3] Ben Cheikh Larbi A, Sai K, Sidhom H, Baptiste D. Constitutive model of
along fiber interface in the form of fiber/matrix interfacial debond- micromechanical damage to predict reduction in stiffness of a fatigued SMC
ing, except at the core layer of transverse specimen where matrix composite. J Mater Eng Perform 2006;15(5):575–80.
 [4] Toubal L, Karama M, Lorrain B. Damage evolution and infrared thermography [19] Arif MF, Meraghni F, Chemisky Y, Despringre N, Robert G. In situ damage
in woven composite laminates under fatigue loading. Int J Fatigue mechanisms investigation of PA66/GF30 composite: effect of relative
2006;28(12):1867–72. humidity. Compos Part B: Eng 2013(0).
[5] Mourglia-Seignobos E. Compréhension des mécanismes physiques de fatigue [20] Mallick PK, Zhou Y. Effect of mean stress on the stress-controlled fatigue of a
dans le polyamide vierge et renforcé de fibres de verre. Ph.D. thesis; INSA Lyon; short E-glass fiber reinforced polyamide-6,6. Int J Fatigue 2004;26(9):941–6.
2009. [21] Bernasconi A, Kulin RM. Effect of frequency upon fatigue strength of a short
[6] Nouri H, Meraghni F, Lory P. Fatigue damage model for injection-molded short glass fiber reinforced polyamide 6: a superposition method based on cyclic
glass fibre reinforced thermoplastics. Int J Fatigue 2009;31(5):934–42. creep parameters. Polym Compos 2009;30(2):154–61.
[7] Meraghni F, Nouri H, Bourgeois N, Czarnota C, Lory P. Parameters identification [22] Bernasconi A, Davoli P, Basile A, Filippi A. Effect of fibre orientation on the
of fatigue damage model for short glass fiber reinforced polyamide (PA6-GF30) fatigue behaviour of a short glass fibre reinforced polyamide-6. Int J Fatigue
using digital image correlation. Proc Eng 2011;10:2110–6. 2007;29(2):199–208.
[8] Nouri H, Czarnota C, Meraghni F. Experimental parameters identification of [23] Nouri H. Modélisation et identification de lois de comportement avec
fatigue damage model for short glass fiber reinforced thermoplastics GFRP. In: endommagement en fatigue polycyclique de matériaux composite a matrice
Haddar M, Romdhane L, Louati J, Ben Amara A, editors. Design and modeling of thermoplastique. Ph.D. thesis; Arts et Métiers ParisTech; 2009.
mechanical systems. Lecture notes in mechanical engineering. Berlin, [24] De Monte M, Moosbrugger E, Quaresimin M. Influence of temperature and
Heidelberg: Springer Berlin Heidelberg; 2013. p. 523–30. thickness on the off-axis behaviour of short glass fibre reinforced polyamide
[9] Steinberger R, Valadas Leitao TI, Ladstatter E, Pinter G, Billinger W, Lang RW. 6.6 Quasi-static loading. Compos Part A: Appl Sci Manuf 2010;41(7):859–71.
Infrared thermographic techniques for non-destructive damage [25] Klimkeit B, Castagnet S, Nadot Y, El Habib A, Benoit G, Bergamo S, et al. Fatigue
characterization of carbon fibre reinforced polymers during tensile fatigue damage mechanisms in short fiber reinforced PBT+PET GF30. Mater Sci Eng: A
testing. Int J Fatigue 2006;28(10):1340–7. 2011;528(3):1577–88.
[10] Ghorbel A, Saintier N, Dhiab A. Investigation of damage evolution in short glass [26] ID19 ESRF. <http://www.esrf.eu/UsersAndScience/Experiments/Imaging/
fibers reinforced polyamide 6,6 under tensile loading using infrared ID19>.
thermography. Proc Eng 2011;10:2123–8. [27] Advani SG, Tucker CL. The use of tensors to describe and predict fiber
[11] Esmaeillou B, Fitoussi J, Lucas A, Tcharkhtchi A. Multi-scale experimental orientation in short fiber composites. J Rheol 1987;31.
analysis of the tension–tension fatigue behavior of a short glass fiber [28] Mlekusch B. Fibre orientation in short-fibre-reinforced thermoplastics II.
reinforced polyamide composite. Proc Eng 2011;10:2117–22. Quantitative measurements by image analysis. Compos Sci Technol
[12] Esmaeillou B, Ferreira P, Bellenger V, Tcharkhtchi A. Fatigue behavior of 1999;59:547–60.
polyamide 66/glass fiber under various kinds of applied load. Polym Compos [29] Eberhardt CN, Clarke A. Fibre-orientation measurements in short-glass-fibre
2012;33(4):540–7. composites. Part I: Automated, high-angular-resolution measurement by
[13] Esmaeillou B, Fereirra P, Bellenger V, Tcharkhtchi A. Fatigue damage initiation confocal microscopy. Compos Sci Technol 2001;61(10):1389–400.
of a PA66/glass fibers composite material. J Appl Polym Sci [30] Lee Y, Lee S, Youn J, Chung K, Kang T. Characterization of fiber orientation in
2012;125(5):4007–14. short fiber reinforced composites with an image processing technique. Mater
[14] Montesano J, Fawaz Z, Bougherara H. Use of infrared thermography to Res Innovat 2002;6(2):65–72.
investigate the fatigue behavior of a carbon fiber reinforced polymer [31] Zhou Y, Mallick P. Fatigue performance of an injection-molded short E-glass
composite. Compos Struct 2013;97:76–83. fiber-reinforced polyamide 6,6. I. Effects of orientation, holes, and weld line.
[15] Bayraktar E, Antolonovich S, Bathias C. Multiscale study of fatigue behaviour of Polym Compos 2006;27(2):230–7.
composite materials by v-rays computed tomography. Int J Fatigue [32] De Monte M, Moosbrugger E, Quaresimin M. Influence of temperature and
2006;28(10):1322–33. thickness on the off-axis behaviour of short glass fibre reinforced polyamide
[16] Lambert J, Chambers A, Sinclair I, Spearing S. 3D damage characterisation and 6.6 cyclic loading. Compos Part A: Appl Sci Manuf 2010;41(10):1368–79.
the role of voids in the fatigue of wind turbine blade materials. Compos Sci [33] Sonsino C, Moosbrugger E. Fatigue design of highly loaded short-glass-fibre
Technol 2012;72(2):337–43. reinforced polyamide parts in engine compartments. Int J Fatigue
[17] Withers PJ, Preuss M. Fatigue and damage in structural materials studied by X- 2008;30(7):1279–88.
ray tomography. Annu Rev Mater Res 2012;42(1):81–103. [34] Ladeveze P, Le Dantec E. Damage modelling of the elementary ply for
[18] Cosmi F, Bernasconi A. Micro-CT investigation on fatigue damage evolution in laminated composites. Compos Sci Technol 1992;43(3):257–67.
short fibre reinforced polymers. Compos Sci Technol 2013;79:70–6.