You are on page 1of 10

International

Journalof
International Journal of Fatigue 29 (2007) 199–208
Fatigue
www.elsevier.com/locate/ijfatigue

Effect of fibre orientation on the fatigue behaviour of a short glass


fibre reinforced polyamide-6
a,*
A. Bernasconi , P. Davoli a, A. Basile b, A. Filippi b

a
Dipartimento di Meccanica, Politecnico di Milano, Milano, Italy
b
Radici Plastics, Villa d’Ogna(BG), Italy

Received 25 October 2005; received in revised form 23 March 2006; accepted 12 April 2006
Available online 15 June 2006

Abstract

The effect of fibre orientation on the fatigue strength of a short glass fibre reinforced polyamide-6 has been investigated. Tension–
tension axial fatigue tests were conducted with specimens extracted from injection moulded plates. Specimens were cut out of plates with
different orientations with respect to the longitudinal axis of plates, and therefore displayed different orientations of the reinforcing fibres.
Results are presented in the form of S–N curves, showing the variation of the fatigue strength as a function of the specimen orientation.
The experimental data, both tensile and fatigue tests, have been compared with values predicted by a failure criterion derived from the
Tsai–Hill formula. The influence on the fatigue lives of the fibre orientation distribution in the thickness of the specimen is discussed.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Short glass fibre reinforced polyamide-6; Fatigue strength; Fibre orientation; Tsai–Hill criterion

1. Introduction strength and failure mechanism of SGFR polyamides has


been studied.
The use of short glass fibre reinforced (SGFR) poly- Like tensile strength, the fatigue strength of SGFR ther-
amides for load bearing applications is increasing, particu- moplastics also depends on the orientation of fibres and is
larly in the automotive, the electronic and the electric influenced by the skin, shell and core layered structure typ-
equipment industry. The substitution of metals by injection ically observed in injection moulded parts [7]. In the injec-
moulded SGFR thermoplastics in the production of highly tion moulding process, the fibre orientation distribution is
stressed components allows for both the reduction of influenced not only by the shape of the part and the relative
weights and the increase of production rates and is partic- position and type of injection gates, but also by the thick-
ularly appreciated for the ease of obtaining complex ness of the component.
shapes. However, this choice calls for design rules different For simple mould geometry, like a thin plate with longi-
than those valid for metals and for appropriate assessment tudinal injection through a film gate, mould walls affect the
methods for the design of these parts against fatigue. orientation of fibres, so that close to them, the melt poly-
Relatively, few data have been published so far on the mer flowing into the mould cavity is prone to shear stresses,
fatigue behaviour of SGFR polyamides. Results of experi- which tend to orientate the fibres parallel to the mould flow
ments conducted using standard specimens have been col- direction (MFD), whereas in the central (core) zone, where
lected and presented in classical references [1,2]. More shear stresses vanish and extensional flow prevails, fibres
recently, the effect of temperature and/or load frequency tend to orientate perpendicular to the MFD, as shown in
[3,4], moisture content [5] and mean stress [6] on the fatigue Fig. 1. A third, very thin skin layer exists at the surface,
with random fibre orientation, where the melt polymer
*
Corresponding author. Tel.: +39 2 2399 8222; fax: +39 2 2399 8202. touches the colder mould walls and freezes without any ori-
E-mail address: andrea.bernasconi@polimi.it (A. Bernasconi). entation effect on fibres.

0142-1123/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2006.04.001
200 A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208

length distribution is usually described by the weight aver-


age fibre length, which is expressed by
PN
ni l2i
LW ¼ Pi¼1N ð1Þ
i¼1 ni li

where the N is the number of measured fibres and ni the


number of fibres having length li. The fibre length distribu-
tion was obtained by image analysis of pictures of fibre
samples extracted from specimens by hydrolysis of the ma-
trix and observed at optical microscope. Image processing
Fig. 1. The different orientations of the reinforcing fibres in the skin, shell followed the method described in Ref. [11]. The weight
and core layers of an injection moulded plate.
average fibre length was 275 lm.
Rectangular plates, 120 mm · 180 mm and 3.2 mm
thickness, were injection moulded through a film gate
The fibre orientation in these plates, and more likely in on the shorter edge. Non-standard specimens were
real parts, can be very different from that obtained in stan- extracted from the plates by water jet cutting with differ-
dard specimens usually employed for tensile tests, like ent orientations with respect to the plate axis, indicated
those defined by the ASTM [8] and ISO [9] standards. In by the angle h in Fig. 2. The specimen shape and dimen-
fact, standard specimens have regular geometry, are sions are reported in Fig. 3. Specimens were cut with ori-
injected through an edge gate at one end and filled along entations of 0, 30, 60 and 90 (being the plates 120 mm
their axis. Thus, fibres tend to be well aligned with the spec- wide, specimens cut at 90 had a shorter total length of
imen axis. On the other hand, real parts are very complex 116 mm, obtained by reducing the length of the grips
in shape and therefore have complex and irregular flow only, without modifying the gauge length and the fillet
and orientation patterns. This makes transferring results radii). The shape of the specimens was carefully designed,
from these tests to the fatigue assessment of real compo- with large fillet radii to avoid stress concentration in
nents difficult without knowledge of the effects of fibre proximity of the grips. The gauge length was designed
orientation. for the application of an extensometer of 25 mm base
The progress in computational simulations allows for length. The position of the specimen gauge with respect
the prediction of the fibre orientation in injection moulded to the plate centrelines for specimens having different ori-
parts made of SGFR thermoplastics with increasing accu- entations was not exactly the same because of the neces-
racy. Commercial software packages exist, which allow sity of simultaneously displacing and rotating the
for the analysis of the moulding process of very complex specimens in order to cut them with different orientations.
parts. This is useful for the prediction of the effects of injec- However, these variations were minimized by placing the
tion parameters on the final shape of the parts, which are gauge of the specimens as close as possible to the plate
often prone to warping and dimensional changes. Subse- centreline.
quent structural analyses by the finite elements method The plate geometry and the choice of a film gate ensured
are possible and research efforts are currently made to a uniform melt polymer flow during the injection process,
improve results by taking into account the local stiffness which is the basis for the assumption of a regular mean ori-
and strength, which results from the orientation distribu- entation of fibres in the centre of the plate. In such condi-
tion of fibres predicted by the software simulating the injec- tions fibres in the shell layer are expected to be oriented
tion moulding process [10]. parallel to the MFD, whereas fibres in the core are more
Although the numerical methods for predicting fibre ori- likely to be oriented perpendicular to the MFD. Therefore,
entation still require improvements, in order for these tools the specimen orientation with respect to the plate axis was
to be used in the design of components against fatigue, a supposed to coincide with the mean fibre orientation, at
deeper knowledge of the influence of fibre orientation on least in the shell layers of specimens.
the fatigue behaviour of SGFR polyamides is believed to The width of the core layer depends on the total width
be even more crucial. and the moulding conditions (amongst others: flow speed,
viscosity, temperature of the melt polymer and of the
2. Experimental mould walls). The injection moulding variables were opti-
mized in order to minimize the width of the core layer.
The material investigated is a short glass fibre reinforced However, the existence of a well-defined shell–core struc-
polyamide-6, produced by Radici Plastics and commer- ture and the presence in the core layer of fibres oriented
cially available under the trade name of Radilon, with a perpendicular to the MFD is still visible, as it can be
E-glass fibre content of 30% by weight (PA6 GF 30). Fibres observed in Fig. 4. This picture was obtained by magnify-
have a nominal diameter of 10 lm and after injection ing at the optical microscope a transverse section of the
moulding fibre length values were comprised in the range central part of a plate, cut perpendicular to its main axis.
25–550 lm, following a Weibull distribution. This fibre It is worth mentioning that the cutting and the polishing
A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208 201

Fig. 2. Dimensions of the injection moulded plates and definition of the orientation angle h of specimens.

Fig. 3. Specimen type and dimensions.

Fig. 4. Optical microscope observation of a polished section perpendicular to the plate axis, central position.

of the section are likely to have removed some of the fibres However, although the precise extension of the core layer
lying on the cutting plane. Moreover, the probability to cannot be estimated with high accuracy, the width of the
display fibres which are aligned in the cutting plane is much core may be assumed between 0.2 mm and 0.3 mm over a
lower than for fibres which are aligned perpendicular to it. total specimen width of 3.2 mm.
202 A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208

After water jet cutting, the lateral surface of the speci- 3. Results
mens was polished with emery papers of decreasing grain
size in order to reduce the initial roughness of specimen 3.1. Tensile tests
edges cut by the water jet to a mean value of Ra = 0.4 lm.
To take into account the variation from the nominal geom- The results of tensile tests are reported in Table 2. It
etry resulting from machining and polishing, the reference clearly appears that the ultimate tensile stress (UTS)
section value was defined as the mean value over the areas decreases as the orientation angle h increases, with a max-
measured at three different locations along the specimen imum at 0 and a minimum at 90. The same trend is
length. In order to test the material close to real service observed for the elastic modulus (E), whereas the maxi-
conditions, it was subjected by accelerated conditioning mum strain increases. The values reported in Table 2 are
(moisture absorption) until equilibrium was reached with evaluated as a mean value over five samples. The stress–
an ambient of 23 C and 50% relative humidity. Plates strain curves of the specimens are shown in Fig. 5. Com-
had been conditioned by submerging them into water for parison of the tensile properties of specimens oriented at
four days at the temperature of 23 C and then kept in 0 reported in Table 2 with those of Table 1 referring to
sealed bags for several weeks to allow for water diffusing ISO 527-2 specimens is not straightforward because equal
to the core of the specimen. The weight of the plates was crosshead speeds results in different strain rate for speci-
monitored during this accelerated conditioning to ensure mens of different gauge length and thickness. However,
that equilibrium was reached before tests began. The water specimens oriented at 0 displayed lower elastic modulus
content at equilibrium was 2.2% by weight. and tensile strength values than standard specimens
Uniaxial tensile tests and fatigue tests were conducted (respectively, 21% and 15%).
using an MTS 810 servo-hydraulic test system, with a This can be explained in terms of the fibre orientation
capacity of 100 kN, equipped with an additional load cell distribution and of the presence of the core layer, although
of 10 kN, as required to increase the accuracy of load mea- minimized by choosing appropriate moulding variables,
surements. All tests were conducted at a room temperature enforced by the plate geometry. The core layer in standard
of 23 C. The tensile tests were carried out at the crosshead specimens tends to be thinner; moreover the lateral walls
speed of 5 mm/min, with a corresponding strain rate (as have the same orienting effect as the top and bottom sur-
measured by the extensometer over the reference length faces. This effect is clearly absent in specimens cut from
of 25 mm) of 0.072 min1. plates, and the presence of the core causes a decrease of
Load controlled tension–tension fatigue tests were con-
ducted in the range of cycles to failure from 103 to 106.
Sinusoidal load cycles were applied keeping the load ratio Table 2
R = (min. load)/(max. load) = 0.1. The cyclic frequency Mean values (±standard deviation) of the tensile properties of PA6 GF 30
was set to 4 Hz. The failure criterion for fatigue tests was specimens extracted from plates at different orientations
specimen separation. Otherwise, tests were interrupted Specimen orientation UTS (MPa) E (MPa) Strain at break (%)
when the number of cycles reached 106 (run-outs). The 0 89.5 (±3.6) 4607 (±114) 6.47 (±0.73)
applied stress was evaluated by dividing the applied load 30 75.6 (±2.2) 3229 (±141) 9.82 (±0.52)
by the reference section area. During the load controlled 60 58.3 (±2.9) 2468 (±70) 9.97 (±1.85)
90 53.2 (±4.2) 2352 (±198) 10.98 (±0.44)
fatigue tests, the axial strain was measured: maximum
and minimum strain values were recorded for each load
cycle, while full stress strain cycles were recorded at fixed
time intervals, to allow for the observation of the cyclic
creep. Tensile and fatigue tests were also conducted with
standard ISO 527-2 Type 1A specimens under the same test
conditions (i.e. temperature, water content, crosshead
speed for tensile tests and cyclic frequency for fatigue tests),
to allow for comparison of results obtained with non-stan-
dard specimens. The results of these preliminary tests on
standard specimens are reported in Table 1.

Table 1
Mean values (±standard deviation) of the tensile properties and values of
the fatigue strength coefficient rf and the fatigue strength exponent b of the
PA6 GF 30, as obtained on ISO 527 standard specimens
Tensile properties Fatigue properties
UTS (MPa) E (MPa) Strain at break (%) rf (MPa) b
104.8 (±0.6) 5846 (±139) 6.78 (±0.15) 101.9 0.048
Fig. 5. Tensile stress–strain curves for different specimen orientations.
A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208 203

both stiffness and UTS. Moreover, for the same mean fibre tensile strength values of Table 2) and the fatigue strength
orientation angle, the scatter of the FOD in plates can be exponents b showed comparable values, i.e. the S–N curves
wider than in standard specimens and the fibre distribution at different orientations were nearly parallel to each other.
tend to be planar, whereas standard specimens display a Test results included run-outs, i.e. tests stopped after 106
three dimensional orientation pattern. Finally, part of the cycles (bright symbols in Fig. 6). Therefore, as suggested
differences in tensile properties may be due to the cutting in the ASTM standard E739 [12], the corresponding values
and polishing process, which is likely to have weakened of rf and b were evaluated by applying the maximum log-
the specimens extracted from plates. likelihood method [13], assuming a constant standard devi-
ation of the number of cycles to failure for all load levels.
3.2. Fatigue tests The scatter width of the fatigue test results, TN (i.e. the ra-
tio between the values of the specimen life at the extremes
The fatigue results of the specimens having different ori- of the 10–90% scatter band for a given load level), has been
entations are reported in Fig. 6, where the log (maximum indicated for each S–N curve in Fig. 6.
applied fatigue stress), or log rmax, is plotted vs. log (cycles During fatigue tests, cyclic creep was observed. The rela-
to failure) or log Nf. It appears that the fatigue strength tionship between applied stress and measured strain is non-
decreased with increasing angle h. The stress-life curves linear even at fairly moderate stress levels, and the cyclic
(also known as S–N curves or Woehler curves), interpolat- response is characterized by hysteresis loops typical to
ing the experimental results are also shown. The following composite materials with visco-elastic matrix like polyam-
equation was used to describe the relationship between ide, as shown in Fig. 7. These loops move along the strain
maximum stress and cycles to failure: axis at a constant speed throughout most of the life of the
specimen, as shown in Fig. 8. The shape and size of these
rmax ¼ rf N b ð2Þ
loops, as well as the rate of increase of the maximum strain
The values of fatigue strength exponent b and fatigue per cycle, depends on the applied stress, and is influenced
strength coefficients rf for different angles h are listed in by frequency of loads and environment temperature.
Table 3. Values of rf decreased with increasing angle h (it During fatigue testing, a general degradation was
can be observed that the rf values are proportional to the observed in the specimens, instead of the nucleation and
propagation of a single dominant crack, as indicated by a
growing number of white lines that became visible at the
surface of the specimen, which can be interpreted as zones
of plastically deformed matrix during diffused cohesive
cracking. These observations and the aspect of fracture sur-
faces were consistent with the fatigue damage mechanism
proposed by Horst [7] for conditioned SGFR polyamide,
i.e. for a material having reached hygro-thermal equilib-
rium with an ambient of 23 C and 50% relative humidity.
Horst proposed a model consisting of initial debonding
of matrix from fibres because of progressive failure of the

Fig. 6. S–N curves for different specimen orientations; bright symbols


indicate test interrupted at 106 cycles (run-outs).

Table 3
Values of the fatigue strength coefficient rf and the fatigue strength
exponent b of PA6 GF 30 specimens extracted from plates at different
orientations
Specimen orientation rf (MPa) b
0 88.5 0.057
30 66.5 0.043 Fig. 7. Cyclic creep: stress–strain cycles; percent numbers indicate the
60 56.0 0.050 stage of specimen life at which the full cycles have been recorded, with
90 50.5 0.045 reference to the number of cycles to failure.
204 A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208

ous bridged cracks propagate until one reaches the critical


depth, and the specimen fails during the last load cycle in a
brittle manner.
The fracture surface of one specimen cut at 0 is shown
in Fig. 9. The micro-ductile area, corresponding to the
upper part of the section, is characterized by high matrix
drawing and fibre pullouts and corresponds to the area,
where fatigue crack nucleated by the above mentioned
mechanism of cohesive cracking. The observed high matrix
deformation is explained by the increase of ductility due to
water absorption prior to testing. The micro-brittle area,
visible in the central and lower part of the picture, corre-
sponds to the failure of the specimen during the last load
cycle and is characterized by failure of the matrix in brittle
mode.
Another consequence of conditioning of specimens is
Fig. 8. Cyclic creep: maximum and minimum strain variation with the decreasing of the glass transition temperature of the
number of cycles (specimen orientation h = 0; test conditions
rmax = 40 MPa, Nf > 106).
matrix [14]; for a water content of 2%, it falls below ambi-
ent (23 C) temperature [15]. Thus, even if hysteresis caused
self heating of the specimen, the material was always tested
over its glass transition temperature and no change in the
fibre–matrix interface, starting at fibre ends, where shear failure mechanism at different stress levels was observed.
stresses reach maximum values. Then the micro-voids at During the test the temperature at specimen surface
fibre ends propagate along the fibres, causing a transfer increased at a low but constant rate; the rate of tempera-
of load from fibres to matrix. Consequently, the matrix ture increase is stress and frequency dependent. The choice
close to these fibres deforms plastically. Due to the plasti- of the optimal test frequency is determined by the volume
cizing effect of water in the conditioned material, the to surface ratio of the specimen, which governs the heat
matrix can sustain high deformation. Voids tend to grow dissipation (e.g. Horst [7] found an optimal frequency of
into cracks, but cracking is made cohesive by bridges of 1 Hz for specimens 5.75 mm thick). The chosen test fre-
drawn matrix filaments joining the crack faces. This mech- quency of 4 Hz was a trade-off between avoiding thermal
anism is similar to craze in unreinforced polymers. Numer- failures and running fast tests.

Fig. 9. SEM micrographs of the fracture surface of a specimen oriented at 0 showing the presence of micro-brittle and micro-ductile areas.
A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208 205

4. Discussion represents a simplification of the real mechanical behaviour


of the specimens under investigation. Actually, only the
4.1. Tensile tests mean fibre orientation in the dominant shell layers can be
supposed to coincide with the reference angle h.
The tensile properties of short fibre reinforced polymers The formula derived from the Tsai–Hill criterion relates
(SFRP) result from the combination of numerous factors, the value of the normal stresses applied in the 1-direction
mainly matrix and fibre properties, fibre length distribution r1 and 2-direction r2 and the shear stress s12 acting on
(FLD) and fibre orientation distribution (FOD). Modelling the 1–2 plane, to the respective experimental UTS values
of the tensile behaviour of the composite material on the r1,u, r2,u and s12,u
basis of these factors is complex and has given rise to  2  2  2
r1 r2 r1 r2 s12
numerous approaches. Bowyer and Bader [16] proposed a þ  2 þ ¼1 ð3Þ
model to derive the ultimate tensile properties of short fibre r1;u r2;u r1;u s12;u
reinforced composites on the basis of the FLD and a cer- The value of r1,u and r2,u are the experimental value of
tain number of macro- and micro-mechanical parameters UTS for the specimens taken at 0 and 90, respectively;
that can be easily derived from stress–stain curves obtained the value of s12,u can be derived from the UTS values of
with tensile tests [17]. The role of the FOD was simply rep- the tests conducted on the specimens extracted at 30 or
resented by one of these parameters. Fu and Lauke pro- 60, by applying the following relationship, derived from
posed a model with the additional feature of including Eq. (3) for the case of uniaxial loading in the h direction
the FOD explicitly [18]. These models were employed for " #1=2
the prediction of the ultimate tensile properties of SFRP cos2 ðhÞðcos2 ðhÞ  sin2 ðhÞÞ sin4 ðhÞ cos2 ðhÞ sin2 ðhÞ
ru ðhÞ ¼ þ 2 þ
composites obtained with standard specimens and allowed r21;u r2;u s212;u
to capture the influence of differences in FLD and FOD ð4Þ
upon tensile strength. However, for these models to be
applied, knowledge of the complete FOD is required. The value of s12,u obtained by applying Eq. (4) to the case
Moreover, the influence of the layered structure frequently of h = 30 was used to predict the tensile strength for
encountered in SFRP is not included. h = 60. The agreement of the predicted (57.9 MPa) with
Other models have been proposed, based on the lami- the experimental strength value for h = 60 (58.4 MPa)
nate analogy approach [19–21], which consists of assuming confirmed that, for this material and this type of specimen,
the misaligned SRFP composite as a stacked sequence of the Tsai–Hill criterion could be used to predict the varia-
unidirectional laminae, each of them being characterized tion of the tensile strength versus the mean fibre orientation
by a certain fibre length and a certain fibre orientation with reference to the load direction.
angle. Then, the stiffness matrix of each ply with reference The values of r1,u, r2,u and s12,u are reported in Table 4
to the loading direction can be determined and the stiffness (being the values of s12,u obtained for h = 30 and for
of the composite is finally derived by summing the contri- h = 60 very close to each other, the reported s12,u value
bution of each lamina. These approaches are particularly was chosen by minimizing the deviations of experimental
suitable for modelling the shell–core–shell structure fre- UTS values for h = 30 and for h = 60 and those obtained
quently encountered in injection moulded parts. These applying Eq. (4)). By introducing these values in Eq. (4),
methods allow to predict correctly the stiffness of SFRP the reference curve of Fig. 10 was drawn, obtaining a good
composites for different loading directions. However, being correlation of predicted with experimental strength values.
based on elastic constants, they are less suitable for predict-
ing the strength of SGFR polymers, particularly for very 4.2. Fatigue tests
ductile matrices, like that of the material used in our tests,
which displayed deviations from the linear behaviour even The observed proportionality of the rf values to the ulti-
at fairly moderate strain values. mate tensile strength suggested the use of a relationship
Considering that the complete FOD is not usually avail- between fatigue strength values and specimen orientation
able to the designer, and that injection moulding software derived from the Tsai–Hill criterion, like that proposed
provide mean fibre orientation values only, reference to by Jen and Lee [23] for the fatigue behaviour of unidirec-
the mean fibre orientation was preferred and a quadratic tional fibre reinforced composites. The relationship was
formula derived by the Tsai–Hill criterion [22] was used obtained by modifying Eq. (3) on the basis of fatigue
to interpolate the tensile strength values plotted versus strength values at different number of cycles to failure
the assumed mean fibre orientation angle. Thus, the mate-
rial was implicitly treated as if it were unidirectional with
Table 4
reference to the longitudinal axis of the plates, which is
Constants of the Tsai–Hill formula interpolating the UTS values at
assumed to be the 1-direction (the 2-direction is perpendic- different orientations
ular to it and lays in the mid-plane of the plate, and the 3-
r1,u (MPa) r2,u (MPa) s12,u (MPa)
direction is perpendicular to the plate surface). The
89.5 53.2 41.9
assumption of a material with unidirectional reinforcement
206 A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208

Fig. 10. Interpolation of the tensile strength values by the Tsai–Hill


criterion. Fig. 11. Interpolation of the fatigue strength values by a Tsai–Hill type
criterion.

and for varying specimen orientation expressed by Eq. (2),


with the values of the fatigue strength coefficient rf and the However, the relationship expressed by Eq. (5) cannot
fatigue strength exponent b reported in Table 3. For a be employed for the design of parts in a straightforward
given number of cycles to failure Nf, it was then possible manner, because the behaviour of specimens is strongly
to interpolate fatigue strength values by the formula dependent upon the fibre orientation distribution and the
 2  2  2 relative thickness of the core layer. Although this layered
r1;max r2;max r1;max r2;max s12;max structure is typical to injection moulded parts, the possibil-
þ  2 þ ¼1
r1;fat ðN Þ r2;fat ðN Þ r1;fat ðN Þ s12;fat ðN Þ ity of using directly for the design of parts the results pre-
ð5Þ sented herein is restricted to the case of parts having the
same relative thickness of the core layer and the same fibre
where the maximum values during one load cycle of the nor- orientation distribution in each layer as the specimens used
mal stresses r1,max and r2,max and of the shear stress s12,max for this experimental work.
are related to the respective experimental fatigue strength From the design perspective, the presence of a core layer
values r1,fat(N), r2,fat(N) and s12,fat(N) for a specimen life can be taken into account by means of master curves, like
of N cycles. As for tensile tests, the value of r1,fat(N) and the one reported in Fig. 12, where S–N data of the speci-
r2,fat(N) are determined by means of the S–N curves mens cut at different orientations are reported on a normal-
reported in Fig. 6 for the specimens taken at 0 and 90, ized scale, i.e. maximum stress values are divided by the
respectively, whereas the value of s12,fat(N) can be derived respective UTS. The S–N curve of the standard specimen
by minimizing the deviation of predictions by Eq. (5) from
fatigue strength values obtained with the tests conducted on
the specimens extracted at 30 and/or at 60.
Values of r1,fat(N), r2,fat(N) and s12,fat(N) as a function
of the number of cycles to failure Nf in the range from
104 to 106 are reported in Table 5. By introducing these val-
ues in Eq. (5), the reference curves of Fig. 11 were drawn.
The differences between experimental results and predicted
values are greater than in the case of tensile tests, mainly
because of the greater scatter of fatigue test results. Never-
theless, the predictive capabilities of the simple model
derived from the Tsai–Hill criterion are still good.

Table 5
Constants of the Tsai–Hill formula interpolating the fatigue strength
values at different orientations and at different fatigue lives
Nf cycles to failure r1,fat (MPa) r2,fat (MPa) s12,fat (MPa)
4
10 52.4 33.4 24.5
105 45.9 30.1 22.2
Fig. 12. Normalized S–N data with superimposed S–N curve of ISO 527
106 40.3 27.1 20.2
standard specimens.
A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208 207

is superimposed, showing a good correlation of results. from the Tsai–Hill criterions for the fatigue assessment of
Apart from the usual scatter of fatigue data, deviations a SGFR polyamide, combined with the use of normalized
from the normalized S–N curve of standard specimens S–N curves, has been proposed and has given satisfactory
can also be explained in terms of different specimen heat results when compared with the results of the experiments
build-up because of different specimen geometry. Note that presented in this study.
master curves can be used, provided that tensile tests are
run at the same strain rate and fatigue tests at the same cyc-
lic frequency. References
This methodology is applicable only if UTS values are
[1] Mandell JF. Fatigue behavior of short fiber composite materials. In:
available for the particular case investigated and a more Reifsnider KL, editor. The fatigue behavior of composite materials.
general method would be preferred. This could consist of Amsterdam: Elsevier; 1991 [chapter 7].
simulating the injection moulding of the part to be [2] Hertzberg RW, Manson JA. Fatigue of engineering plastics. New
designed against fatigue, evaluating the extension of the York: Academic Press; 1980.
[3] Jia N, Kagan VA. Effects of time and temperature on the tension–
core layer and performing subsequent structural analysis
tension fatigue behavior of short fiber reinforced polyamides. Polym
by assuming a layered structure. As mentioned above, this Compos 1997;19:408–14.
approach has been validated for the evaluation of stiffness [4] Handa K, Kato A, Narisawa I. Fatigue characteristics of a glass-
but requires more experimental investigation to be applied fiber-reinforced polyamide. J Appl Polym Sci 1999;72:1783–93.
to the assessment of the fatigue strength, where the non-lin- [5] Horst JJ, Spoormaker JL. Fatigue fracture mechanism and fractog-
raphy of short-glassfibre-reinforced polyamide 6. J Mater Sci
earity of the cyclic stress–strain relationship and its evolu-
1997;32:3641–51.
tion during the life has necessarily to be taken into account. [6] Mallick PK, Yuanxin Zhou. Effect of mean stress on stress-controlled
From this point of view, the effect of fibre orientation on fatigue of a short E-glass fiber reinforced polyamide-6,6. Int J Fatigue
the fatigue behaviour of SGFR polymers should be studied 2004;26:941–6.
with varying core layer widths or by removing the core [7] Horst JJ, Spoormaker JL. Mechanism of fatigue in short glass fiber
layer completely, e.g. by using different moulding methods, reinforced polyamide 6. Polym Eng Sci 1996;36:2718–26.
[8] ASTM D 638–00. Standard test method for tensile properties of
like compression moulding of aligned extrusion rods [24]. plastics. American Society For Testing Materials; 2000.
[9] ISO 527-2. Plastics – determination of tensile properties – part 2: test
5. Conclusions conditions for moulding and extrusion plastics. International Orga-
nization For Standards; 1993.
[10] Whiteside BR, Coates PD, Hine PJ, Duckett RA. Glass fibre
Tensile and fatigue tests were conducted on specimens of
orientation within injection moulded automotive pedal. Simulation
short glass fibre reinforced polyamide (PA6 with 30 wt% and experimental studies. Plast Rubber Compos 2000;29:38–45.
GF) extracted from injection moulded plates. The orienta- [11] Fu S-Y, Mai Y-W, Ching EC-Y, Li RKY. Correction of the
tion of specimens with reference to the longitudinal axis of measurement of fiber length of short fiber reinforced thermoplastics.
the plates was varied. Consequently, the mean fibre orien- Compos: Part A 2002;33:1549–55.
[12] ASTM E 739-91. Standard practice for statistical analysis of linear or
tation angle with respect to the applied stresses was
linearized stress-life (S–N) and strain-life (e–N) fatigue data. Amer-
expected to follow the same relationship. ican Society For Testing Materials; 1991.
Results showed decreasing values of elastic modulus, ulti- [13] Nelson W. Accelerated testing: statistical models, test plans, and data
mate tensile stress and fatigue strength for increasing values analyses. New York: Wiley-Interscience; 1990.
of the orientation angle of specimens. The Tsai–Hill crite- [14] Batzer H, Kreibich UT. Influence of water content on thermal
rion allowed to predict the dependence of the UTS values transitions in natural polymers and synthetic polyamides. Polym Bull
1981;5:585–90.
upon the orientation angle with good accuracy. A similar [15] Ishisaka A, Kawagoe M. Examination of the time–water content
relationship held true also for the fatigue strength values superposition on the dynamic viscoelasticity of moistened polyamide
in the range from 103 to 106 cycles to failure. However, these 6 and epoxy. J Appl Polym Sci 2004;93:560–7.
relationships are empirical and difficulties arise when trying [16] Bowyer WH, Bader MG. On the re-inforcement of thermoplastics by
imperfectly aligned discontinuous fibres. J Mater Sci 1972;7:1315–21.
to transfer these results to the design of components.
[17] Thomason JL. Micromechanical parameters from macromechanical
In fact, the observation of fracture surface and polished measurements on glass reinforced polyamide 6,6. Compos Sci
sections evidenced a layered structure, typical to injection Technol 2001;61:2007–16.
moulded parts, characterized by the presence of shell lay- [18] Fu S, Lauke B. Effects of fiber length and fiber orientation
ers, close to the mould surfaces, having most of the fibres distributions on the tensile strength of short-fiber-reinforced poly-
mers. Compos Sci Technol 1996;56:1179–90.
oriented parallel to the mould flow direction, and a central,
[19] Fu S, Lauke B. The elastic modulus of misaligned short-fiber-
though thin, core layer, with many fibres oriented perpen- reinforced polymers. Compos Sci Technol 1998;58:389–400.
dicular to the MFD. [20] Saito M, Kukula S, Kataoka Y. Practical use of the statistically
The thickness of the core layer in real parts, and also the modified laminate model for injection moldings. Part 1: method and
fibre orientation distribution inside each layer, may be dif- verification. Polym Compos 1998;19:497–505.
[21] Skourlis TP, Pochiraju K, Chassapis C, Manoochehri S. Structure-
ferent than that found in the specimens used for the present
modulus relationships for injection-molded long fiber-reinforced
study. For this reason, transfer of experimental results to polyphthalamides. Compos Part B 1998;29B:309–20.
the design of real parts against fatigue is not straightfor- [22] Azzi VD, Tsai SW. Anisotropic strength of composites. Exp Mech
ward. However, the possibility of using a formula derived 1965:283–8.
208 A. Bernasconi et al. / International Journal of Fatigue 29 (2007) 199–208

[23] Jen M-HR, Lee C-H. Strength and life in thermoplastic composite [24] Lumini F, Pavan A. Short fibre composite materials: relationship
laminates under static and fatigue loads. Part II: Formulation. Int J between fibre orientation and fracture toughness. Plast Rubber
Fatigue 1998;20:617–29. Compos Process Appl 1998;27:240–6.

You might also like