HAL Id: hal-01202706https://hal.archives-ouvertes.fr/hal-01202706

Submitted on 21 Sep 2015

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Fatigue Damage Initiation of a PA66/Glass FibersComposite Material

B Esmaeillou, P Fereirra, V Bellenger, A Tcharkhtchi

To cite this version:B Esmaeillou, P Fereirra, V Bellenger, A Tcharkhtchi. Fatigue Damage Initiation of a PA66/GlassFibers Composite Material. Journal of Applied Polymer Science, Wiley, 2012, 125, pp.4007-4014.�10.1002/app.36728�. �hal-01202706�

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.euHandle ID: .http://hdl.handle.net/10985/10083

To cite this version :

B ESMAEILLOU, P FEREIRRA, V BELLENGER, A TCHARKHTCHI - Fatigue Damage Initiationof a PA66/Glass Fibers Composite Material - Journal of Applied Polymer Science - Vol. 125,p.4007-4014. - 2012

Any correspondence concerning this service should be sent to the repository

Administrator : archiveouverte@ensam.eu

Fatigue Damage Initiation of a PA66/Glass FibersComposite Material

B. Esmaeillou, P. Fereirra, V. Bellenger, A. Tcharkhtchi

PIMM, Arts and M�etiers ParisTech, 151, Boulevard de l’Hopital, 75013 Paris, France

ABSTRACT: Fatigue damage initiation of a PA66/glassfiber composite material is studied with interrupted testscarried out with an ‘‘alternative bending device" and asmall applied strain. During the damage initiation period,no change of macroscopic properties, density, cristallinityratio, glass transition temperature, and flexural elasticmodulus is observed. Polysequential tests are carried outwith three rest times differing by their length. These resttimes allow the relaxation of macromolecular chains inthe region of the microdefects and increase the number

of cycles at fracture. The most efficient stop is the onejust before the final fracture. The comparison of thefatigue behavior of the composite and its neat matrixshows that the microdefects relaxed during the breakare identical to those which initiate damage and finalfracture. VC 2012 Wiley Periodicals, Inc. J Appl Polym Sci 125:4007–4014, 2012

Keywords: polyamide 66; composites; damage zone;fatigue tests; relaxation

INTRODUCTION

The fatigue behavior of composite materials withshort fibers is still a key problem despite the pro-gress in material science. For many molded parts,the lifetime prediction is crucial. In the automotiveindustry, pressures and temperatures in engines areinclined to go up, which makes the fatigue proper-ties of these parts critical. A fatigue curve displaysthe variation of the induced stress versus thenumber of cycles for a given strain and frequency.Generally, these curves display three parts: (1) a fastdecay of the induced stress due to the setting upof a thermal regime: an increasing temperatureinvolves a decreasing elastic modulus; (2) a long pla-teau (period 2) during which the induced stressdecreases only slightly without any significant tem-perature change; (3) a sudden stress reductioncorresponding to the coalescence of microdefects fol-lowed by the sample fracture. For PA66/glass fibers,several research works1–3 report that microcavitation(crack) is initiated by the build-up of microdefects atfiber ends. Damage initiation occurs during theperiod 2 of stress stabilization. For unnotched sam-ples of amorphous polymers, no significant changein macroscopic properties4 (macroscopic compliance,cristallinity ratio, and transition temperatures) isobserved during this period. The aim of this work is

to study damage initiation by the way of interruptedfatigue tests and polysequential fatigue tests.Interrupted fatigue tests are carried out and, duringevery rest time, dogbone samples are analyzed bydifferential scanning calorimetry (melting tempera-ture Tm, cristallinity ratio xc), density measurements,and mechanical testing (dynamic mechanical analy-sis, flexural modulus, and ultimate stress). Withpolysequential fatigue tests, the effect of the resttime and stop location on the fatigue lifetime isinvestigated.

MATERIALS AND METHODS

The composite material is a thermoplastic polyamide66 matrix reinforced by short glass fibres. The fibercontent was measured by gravimetry on six samplesafter a 3-h pyrolysis at 500�C. It is equal to 30.7 6

0.05. The length is 203 6 12 lm, and the diameter30 6 1 lm. The fraction crystallinity xc is determinedby DSC experiments using the ratioDHm

DHcm, where DHm

is the experimental value of melting enthalpy andDHc

m the melting enthalpy of the 100% crystallinepolymer. DHc

m ¼ 192 J/g.5 Dogbone specimens,150 � 20 � 4 mm, were injected with an injection-molding machine (DK Codim 175 T) according tothe AFNOR NF T 51-034 standard.The most constrained part of the sample during

the fatigue test was determined by a linear elasticcalculation, xmax. It is located at a distance of 18 mmfrom the embedded part6 (Fig. 1). Sample surface isobserved in this most constrained part by opticalmicroscopy with an Olympus BH2 or by scanning

Correspondence to: V. Bellenger (vbellenger@gmail.com).

electron microscopy with a Hitachi S-4800 apparatusat an accelerating potential of 0.8 kV. Samples arenot metallized, because a morphological analysisdoes not require this surface treatment. A specimensection perpendicular to the injection directionobserved by optical microscopy indicates that mostfibers are oriented in the injection direction. The lossmodulus E00 is measured by viscoelasticimetry on aNETZSCH DMA 242 apparatus. The test is per-formed in three-point bending mode at a 10 Hz fre-quency, with a static force of 4 N and a dynamicforce of 2 N. The heating rate is 5�C/min. The initialmechanical properties are determined at 23�C and50% HR in bending mode with an INSTRON 4502.The strain rate is 2 mm/min, and the bending lengthis 60 mm. The elastic modulus E and the ultimatestress ru are calculated with the following relation-ships, where D is the bending length, h the samplethickness, b the sample width, F/c the initial slopeof the stress–strain curve, and Fmax the maximumvalue of the load.

E ¼ D2

4:h:b3

� �F

c

� �ru ¼ 3FmaxD

2bh2

The fatigue test1 is performed in a flexural alter-nate bending device (Fig. 2), at a frequency of 10 Hzwith an R ¼ emin

emaxvalue equal to �1 at 23�c and 50%

RH. During the fatigue test, a decreasing amplitudeof the induced stress is observed, but the ratio rmin

rmax

does not change. The curves rmin ¼ f (number ofcycles) and rmax ¼ f (number of cycles) are symmet-

rical. The applied amplitude strain is 0.013 for inter-rupted fatigue tests and 0.017 for polysequentialfatigue tests. For these last tests, the first breakoccurs at 30,000 loading cycles, and the rest timevaries between 2 and 30 min. Then, the sameapproach is done for a stop at 45,000 loading cyclesand for a stop at 60,000 loading cycles close to thefinal sample fracture. For some polysequentialfatigue tests, surface sample is polished in the mostconstrained part before optical microscopy observa-tion at the rest time. After observation, the fatiguestrain is again applied on the same sample. The timeof break is at most 10 min. The surface temperaturein the most constrained area is measured during thetest by a RAYTEK infrared camcorder.

RESULTS AND DISCUSSION

Interrupted fatigue tests are performed at 10 Hzwith an applied amplitude strain of 0.013. Thephysicochemical and mechanical properties, density,critallinity ratio, loss modulus, flexural modulus,and ultimate stress are measured after 300,000 and500,000 loading cycles. The values reported in TableI show that these properties are not modified duringthe damage initiation period. It does not mean thatthere is no defect build-up, but their size/concentra-tion is too small [Fig. 3(a)], to induce a variation ofthe macroscopic properties. Before 500,000 cycles,the size of submicron defects in the most constrainedarea is smaller than 0.2 lm. The mean value of theultimate stress seems to increase with the numberof loading cycles, but it is not significant, takinginto account the standard deviations. The loss

Figure 1 Most constrained area of a dogbone sampleindicated by a linear elastic calculation.6

Figure 2 Alternative bending device used for fatiguetests.6

TABLE IPhysicochemical and Mechanical Properties of PA66/GF Versus Number of Loading Cycles

Density (kg/m3) xc (%) E0042�C (MPa) E (MPa) ru (MPa)

0 cycle loading 1409 6 13 35 6 4 343 6680 6 200 190 6 7300,000 cycle loading 1374 6 22 36 6 3 340 6860 6 70 193 6 8500,000 cycle loading 1404 6 8 38 6 5 290 6148 6 612 204 6 19

modulus E00 measured at 42�C (the maximum valuecorresponding to Ta is 55�C) seems to be constantexcept near the fracture (500,000 cycles), whereit decreases when microdefects are generated[Fig. 3(b)]. Between 0 and 500,000 cycles, flexuralproperties are unchanged taking into account the

standard deviation. The standard deviation valuesare high for 500,000 loading cycle samples andreflect the heterogeneity of these samples, heteroge-neity, which is also noticed for density and cristalli-nity ratio measurements.Polysequential fatigue tests are carried out at 10 Hz

frequency and an applied amplitude strain e ¼ 0.017.Figure 4 displays the induced amplitude stress andthe temperature versus the number of cycles. Thecurve r ¼ f (N) displays after 2000 cycles a very noisypattern due to electronics noise. A first decrease of theinduced stress from 85 to 60 MPa is observed between500 and 3000 cycles. It goes with an increasing tem-perature from 27 to 53�C, and then the temperaturekeeps on increasing but more slowly. This tempera-ture increase induced by self-heating is smaller thanthe one reported before for another PA66/glass fibercomposite material.1 The main difference betweenboth composites is the sizing of glass fibers, which isspecific to PA66, whereas in the previous study, thesizing of glass fibers is unknown and probably muchless efficient. The low-mechanical resistance of theinterface favors crack initiation and, as a consequence,an excess of self-heating temperature rise. This beha-vior can be explained with the viscoelastic spectrum(Fig. 5), performed at the same frequency, 10 Hz.Figure 5 displays the variation of the storage modulusE0, the loss modulus E00, and the damping tan d equalto the ratio E00/E0. The increasing temperature from 27to 53�C is due to an increasing value of the loss modu-lus E00. The dissipated heat at each mechanical cycle isproportional to the angular frequency x, the appliedamplitude strain e0, and the loss modulus E00.

Q / � 1

2x e20E

00

These successive breaks do increase the fatiguelifetime (65,000 cycles instead of 60,000 cycles atbreak). Optical microscopy observations are carriedout on the most constrained part, which has beenpolished before the test. Sample polishing can haveopposite effects on the fatigue life. On one hand, it

Figure 3 SEM observation of the most constrained areaof a specimen tested at 10 Hz and e ¼ 0.013 (interruptedtest), (a) after 300,000 loading cycles and (b) after 500,000loading cycles.

Figure 4 Induced stress and temperature versus numberof cycles of a fatigue test at 10 Hz and e ¼ 0.017. Figure 5 Viscoelastic spectrum of PA66/glass fiber at 10

Hz between �10 and 130�C.

can suppress surface microdefects induced by inter-nal stresses, which are built-up during processingcooling and which possibly initiate microcracks andthus increase the fatigue life. On the other hand, itcan modify the sample geometry, and this possiblywill decrease the fatigue life. For instance, in adhe-sively bonded structures, the removal of the adhe-sive fillet has been found to reduce fatigue life.7 Byoptical microscopy, no damage is observed before30,000 cycles [Fig. 6(a)]. From this number of cycles,some pull-out fibers appear near the surface butonly in the most constrained part [Fig. 6(b,c)], andmicrovoids/cracks are only visible just before thebreak Figure 6(d).

Then the study is carried on with an unpolishedsample by using scanning electron microscopy [Fig.7(a)]. Another kind of damage is observed: after1000 cycles, some fiber ends (surrounded by a whitecircle on the figure) are pulling out near the surfaceand initiate the skin fracture [Fig. 7(b,c)]. This da-mage is nanoscale size that is why we call it nanode-fects. Microcracks, � 5 lm, are observed just beforefailure [Fig. 7(d)]. It is very difficult to stop the

fatigue test just before the sample failure, becausethe last phase of crack propagation is very fast.Cracks propagate first in zones where the fiber con-centration is high [Fig. 8(a)] and then in the matrixuntil the final fracture [Fig. 8(b)]. This scenariorecalls the work about the essential work of fractureof polyamide 66 filled with TiO2 nanoparticles.8

Individual nanoparticles act as stress concentrationpoints, which promote tiny cavitations that coalesceinto submicron ones, then rapidly grow into micro-voids, and initiate cracks due to the high-level stressconcentration.For the test conditions used for polysequential fa-

tigue tests, the number of cycles at fracture, Nr, is65,000. The fatigue lifetime increases with the resttime and is the highest when the break occurs closeto the final fracture (Fig. 9), where the crack size isin a micron scale {see, for instance [Fig. 7(d)]}. Itmeans that relaxation of defect build-up at the endof the damage initiation period is more efficient thanthe relaxation of defects build-up earlier. The pointson vertical axis represent fatigue tests, which havebeen carried on until failure. The fatigue life is

Figure 6 Optical microscopy observation of the most constrained part of a specimen tested at 10 Hz and e ¼ 0.017 (poly-sequential test) (a) initial, (b) 30,000 cycles, T ¼ 57�C, (c) 45,000 cycles, T ¼ 60�C, (d) 60,000 cycles, T ¼ 62�C.

shorter than for polysequential tests carried out inthe same test conditions. The effect of the rest timecan be explained in terms of defects relaxation or interms of increasing strength of the specimen inducedby the temperature decrease. The shape of the tem-perature curve during the test, Figure 10 indicatesthat the temperature decrease is quite fast. At thebeginning of the rest time, the temperature is equalto 62�C, then it decreases quickly, in 7.5 minuntil 25�C, and then much more slowly (11.5 min) to22�C. However, the most efficient rest time is20 min, which favors the explanation of increasinglife time in terms of defect relaxation even if a tem-perature effect cannot be kept out. Figure 11 showsthat an increasing rest time up to 20 min doesincrease the fatigue lifetime. Afterward, between20 min and 2 h, the number of cycles at fractureremains constant, about equal to 110,000 cycles. Thatmeans that for an applied amplitude strain, e0 ¼0.017, a rest time of 20 min is enough for microde-fects relaxation. When a rest time of 5 min is appliedclose to the final fracture, the mean stress induced is

equal to 63 MPa, hence a 21 MPa gap with the initialand maximum induced stress. If the rest time isequal to 20 min, the mean amplitude of the inducedstress (82 MPa) is practically equal to the initial andmaximum amplitude of the induced stress (84 MPa).Now, the question is: what is the nature of themicrodefects relaxed during the fatigue test break?Are they conformational changes in the amorphousphase? Robertson9 considered for polystyrene andpolymethyl methacrylate, a molecular model inwhich a shear stress field induced rotational confor-mations of backbone bonds, enough numerous toinduce an increasing concentration of flex bonds,which are in a rotational conformation differentfrom the initial one (Fig. 12). These changes in rota-tional conformations induce an increasing disorderand increasing volume, a polymer structural statesimilar to the rubbery state observed at T > Tg. As aconsequence, these changes break up the rigidityand allow plastic flow. An empirical equation bin-ding the induced maximum strain rate to the testtemperature and to the applied shear stress is used

Figure 7 SEM observation of the most constrained and polished area of a specimen tested at 10 Hz and e ¼ 0.017 (poly-sequential test) (a): initial, (b) 1000 cycles, T ¼ 35�C, (c) 5,000 cycles, T ¼ 52�C, (d) 10,000 cycles, T ¼ 55�C. The whitecircles indicate the fiber ends that are pulled out.

to calculate the plastic strain properties of the poly-styrene and polymethyl methacrylate. The resultsare then confronted with experimental results ofcold drawing experiments, and their good agree-ment validates the model. Could they be packingdensity defects, for instance, at the interphase crys-talline/amorphous phases? An amorphous polymer

can stand for a disorganized arrangement of struc-tural units,10 each unit being located in a cage builtup by the neighbor structural units. A ‘‘quasi-punc-tual defect’’ is a spot made up by a structural unitand her first neighbors (the cage cited above) ofwhich the free enthalpy level is higher than themean value of the whole group of structural units.These ‘‘quasi-punctual defects’’ correspond to localdensity fluctuations existing in the liquid state andwhich are frozen in the glassy state. They are in aphysical nonequilibrium state and thermoreversi-ble.10 The reversible character of these defectsimplies that polysequential tests should be done tostudy the effect of a rest time (relaxation time) onthe fatigue behavior. A decreasing concentration ofthese punctual defects does increase the fatigue life-time. Fatigue craze initiation was investigated inpolycarbonate.11 The disentanglement of polymerchains plays an important role in void-like structuresor ‘‘protocrazes’’ with a size of � 50 nm. Could thesame phenomenon occur in the amorphous phase ofpolyamide 66? Are these defects preplasticizeddomains again in the matrix? The preyield and yieldbehavior of amorphous polymers have been

Figure 8 SEM observation of the most constrained andunpolished area of a specimen tested at 10 Hz, just beforefracture (62,000 cycles). (a) Crack propagation in zoneswith a high fibers concentration and (b) crack propagationin the matrix.

Figure 9 Variation of the cycle number at fracture (Nr)versus the number of loading cycles (N) before the testbreak and for various rest times. e0 ¼ 0.017, r � 90 MPa,x ¼ 10 Hz, and %RH ¼ 50.

Figure 10 Induced stress and temperature variation dur-ing a polysequential test at 10 Hz, e0 ¼ 0.016 with a resttime of 25 min near the fracture.

Figure 11 Variation of the cycle number at fracture, Nr,versus the rest time for three values of the number ofloading cycles before the test break. e0 ¼ 0.017, r � 90 MPa,x ¼ 10 Hz, and %RH ¼ 50.

described with a metallurgical approach.12 Somenew mechanical data were analyzed in terms of dis-location-like defects involved by deformation in themacromolecular chain arrangement. The comparisonof the fatigue behavior of the composite and its neatmatrix gives interesting information. Figure 13shows the evolution of the induced amplitude stressversus the number of loading cycles for the compo-site and the neat matrix strained at 10 Hz frequencywith an applied amplitude strain e0 ¼ 0.019. Theinduced stress is three times higher for the compo-site than for the neat matrix, and, of course, the fa-tigue lifetime is much shorter. The first reason is thedifference in elastic modulus value, but the secondone is the difference in self-heating. The surface tem-perature in the most constrained part of the speci-men is recorded during the test (Fig. 14). The tem-perature rise of the composite is higher than the oneof the neat matrix. Nanocrack build-up at fiber ends,precursors of microcracks, induces an increasingdamping and thus a higher temperature increase.The gap between the number of cycles correspon-ding to these two rises of temperature means thatthe crack build-up at fiber ends plays a key role indamage initiation. Are these defects relaxed during

the test break off, identical to the microcracks thatcoalesce in zones where the fiber concentration ishigh? It is reasonable to answer positively, becauseat the end of a rest time of 20 min just before thefinal fracture, the induced stress when the strain isagain applied is equal to the maximum and initialinduced stress. Before the test interruption, the tem-perature involves a decreasing value of the elasticmodulus. As the applied strain is constant, theinduced stress decreases that favors an increasinglife time. At the beginning of the rest time, the tem-perature decreases (Fig. 10) and, as a consequence,the elastic modulus and the induced stress increaseagain, which would decrease the fatigue lifetime.The opposite effect is noticed, so that the tempera-ture variation during the rest time does not involvean increasing fatigue lifetime. But, at the beginningof the rest time, the temperature is higher than theglass transition temperature. Thus, the matrix is in arubbery state and its molecular mobility is highenough to induce the closing/deletion of micro-cracks. So, it is reasonable to assume that relaxationof macromolecular chains in the region of microde-fects is well responsible for the increasing number ofcycles at break.

CONCLUSIONS

A fatigue curve displays three zones, of which thesecond one is very long, nearly 80% of the fatiguelife, when the applied strain is low; during thisperiod, damage initiation occurs, but the size/concentration of microdefects build-up is too smallto involve a change in macroscopic properties suchas density, cristallinity, and flexural elastic modulus.At the beginning of the third zone, which is shortcompared to the second one, cracks propagation firstoccur in zones where the fiber concentration is high,then in the matrix until the final fracture. Thefatigue lifetime is higher during polysequential tests

Figure 12 Change in rotational conformation * transconformation (low energy state), l cis conformation(flexed state).

Figure 13 Induced stress of the neat matrix (dashed line)and of the composite (dotted line) versus the number ofloading cycles, tested at a 10 Hz frequency, e0 ¼ 0.019,and %RH ¼ 50.

Figure 14 Surface temperature of the neat matrix(dashed line) and of the composite (dotted line) versus thenumber of loading cycles, tested at a 10 Hz frequency, e0¼ 0.019, and %RH ¼ 50.

than during nonstop tests. The most efficient breakis just before fracture. There is an optimum rest timefor microdefects relaxation, which is equal to 20 minfor our material and testing conditions. The natureof these relaxed microdefects was investigatedby comparing the fatigue behavior of the neat matrixand of the composite. For identical testing conditions,the self-heating of the composite is higher than theone of the neat matrix. It means that nanocracksbuild-up at fiber ends, and precursors of microcracksplay the key role. The comparison of the initialinduced stress and the induced stress after a resttime just before the specimen fracture seems to indi-cate that these microcracks are closed during therest period.

Many thanks to Gilles Robert for helpful discussion andRhodia Co for providing PA66/glass fiber compounds.

References

1. Bellenger, V.; Tcharkhtchi, A.; Castaing, Ph. Int J Fatigue 2006,28, 1348.

2. Noda, K. Polymer 2001, 42, 5803.3. Horst, J. J.; Spoomaker, J. L. J Mater Sci 1997, 32, 3641.4. Baltenneck, F.; Trotignon, J.-P.; Verdu, J. Polym Eng Sci 1997,

37, 1740.5. Mark, J. E. Polymer Data Handbook; Oxford University Press:

New York, 1999; p 195.6. Barbouchi S., Bellenger V., Tcharkhtchi A., Castaing Ph, Jolli-

vet T. Effect of water on the fatigue behavior of a PA66/glassfibers composite material. J Mater Sci 2007, 42, 2181.

7. Crocombe, A. D.; Ong, C. I.; Chan, C. M.; Wahab, M. M. A.;Ashcroft, I. A. J Adhes 2002, 78, 745.

8. Yang, J.; Zhang, Z.; Zhang, H. Comp Sci Technol 2005, 65, 2374.9. Robertson, R. E. J Chem Phys 1966, 44, 3950.10. Perez, J. Physique et M�ecanique des Polymeres Amorphes.

Technique et Documentation Lavoisier: Paris; 1992; p 34–35.11. Hristov, H. A.; Yee, A. F.; Gidley, D. W. Polymer 1994, 35, 3604.12. Escaig, B. Polym Eng Sci 1984, 24, 737.