Fretting fatigue of continuous carbon fibre reinforced polymer composites

Fretting fatigue of continuous carbon fibre reinforced polymer composites

Wear, 145 167 (1991) 167-188 Fretting fatigue of continuous carbon fibre reinforced polymer composites 0. Jacobs Polymer and Composites Group, Tec...

2MB Sizes 0 Downloads 55 Views




(1991) 167-188

Fretting fatigue of continuous carbon fibre reinforced polymer composites 0. Jacobs Polymer and Composites Group, Technical University Hamburg-Harburg, Hamburg 90 (F.R.G.)


K. Friedrich Institute for Composite Materials, University of Kaiserslautern, 6750 Kaiserslautern (ER. G.)

K. Schulte Institute for Materials Research, D.L.R., 5000 K&t PO (F.R.G.) (Received August 30, 1990; accepted October 16, 1990)

Abstract Continuous carbon fibre reinforced epoxy resin laminates were exposed to a fatigue loading. Simultaneously, cylindrical metallic pins with flat and polished front surfaces were pressed against the specimen surfaces with an apparent contact pressure between 10 and 23 MPa. The peak-to-peak oscillation width in the contact region amounted to about 700 pm. Fatigue life of the composite could drastically be reduced by this additional fretting component, if load bearing 0” layers were damaged by fretting. This paper studies the mechanisms of damagedevelopmentin carbon fibre reinforced laminates under fretting fatigue conditions and proposes a qu~ti~tive measure for the degree of fretting fatigue damage. The influence of the loading conditions (contact load, counterpart material and slip amplitude) is investigated.

1. Introduction

Relative cyclic sliding motion of two surfaces in intimate contact constitutes a special wear process called fretting. Quite often this type of surface damage is induced by fatigue loading of one or both mating components. The interaction of surface damage due to fretting (e.g. microcracking) and fatigue is able to drastically reduce fatigue life of a structural part. Since fretting fatigue influences a variety of engineering applications such as multilayer leaf springs, bohed joints, fittings, flanges and seals, this problem has attracted numerous research activities f 1, 21. Polymer composites replace metals in an increasing number of practical applications which are subjected to fretting wear or fretting fatigue. The specific advantages of these materials are weight reduction, corrosion resistance, better fatigue performance, high vibration damping or design and manufacturing advantages etc. However, while some studies exist on the fretting wear of polymer composites [3], and in particular of carbon fibre 0043-1048/91/$3.50

0 Elsevier Sequoia/Printed in The Netherlands




% : z 5 2


.P 5 u. & 2 3

150 lo3

IO5 104 Load Cycles to Failure, N



1. Fatigue life vs. maximum load in a fatigue cycle for plain fatigue and fretting Counterpart: aluminium pin [6].


reinforced epoxy resin (CF/EP) laminates [4, 51, the fretting fatigue of this group of materials is still an open field. A pilot study in the framework of this project [6, 71 indicated that an additional fretting surface damage is able to reduce the fatigue life of CFI EP laminates, if load bearing 0” layers (plies with fibres oriented in loading direction) are exposed to fretting. Off-axis plies contribute only little to the load bearing capacity of a laminate so that their damage due to fretting affects the material’s fatigue properties insignificantly (Fig. 1). The current project intends to systematically investigate the influence of several system components on the fretting wear and fretting fatigue performance of CF/EP laminates: sample material, counterpart material and loading parameters. In a previous study, the influence of the loading conditions on the fretting wear of a CF/EP laminate was investigated [5]. On this basis, the fretting fatigue behaviour of the laminates is compared with their plain fatigue performance in order to describe the interaction between surface damage and fatigue of the bulk. Supported by observations of the failure mechanisms, a data reduction method is proposed for the quantitative analysis of fretting fatigue damage development. 2. Experimental


2.1. Materials

The samples consisted of an epoxy matrix (EP) reinforced with aligned continuous carbon fibres (CF). The fibre volume content amounted to about 60%. Table 1 lists the laminates tested. The covering layers possessed 0” orientation. This was required because fretting damage of off-axis plies only insignificantly affects the fatigue performance of CF laminates [ 61. Most of the laminates were manufactured using a peel ply. This is a layer of an open-weave material which separates during curing the laminate from the bleeder cloth, which picks up excessive matrix resin. After curing


TABLE 1 Materials tested under fretting fatigue load Fibre

Stacking sequence

Strength (MPaI 850


BASF R 5212

T 300

[O”z, 90”212s


Ciba Geigy 914 c

T 300

UOOz,+45”, o”2, *45”,



I-I 1 mm Fig. 2. Optical micrograph of laminate Ll. The arrows point at the resin-rich surface layers left by the peel ply.

the peel ply is removed, but it leaves a resin-rich layer with the impression of the fabric on the surface of the sample (Pig. 2). In order to study the effect of this surface structure, one set of specimens of the Ll-type laminate was produced without peel plies, having plain surfaces. Cylindrical metal pins (5 mm in diameter) with flat front surfaces served as fretting pins. The front surfaces were first ground with an abrasive paper (grid up to 1200) and then polished with an alumina powder emulsion. The final surface roughness was about 0.12 pm. The edges of the front surfaces were rounded off in order to avoid cutting of the laminate by sharp edges. Prior to starting the tests, the pins were cleaned with acetone. A low carbon NiCr steel (Vickers hardness HV= 296), an aluminium alloy (HV= 135) and a brass (HV = 160) were chosen as pin materials. 2.2. Test methods Figure 3 schematically illustrates the fretting fatigue test procedure. The laminates were cut to beam-shaped specimens with a length of about




Load Cycles to Failure, N Fig.

3. Principle

of the fretting


test procedure.

Fig. 4. Upper fatigue load vs. lifetime for plain fatigue and fretting fatigue of laminate various contact loads. Aluminium pins are assigned by “Au”, steel pins by “Steel”.

Ll at

280 mm and a width of 6-9 mm. Aluminium end tabs were used for easy load transfer and to protect the specimen surfaces against the action of the clamping grips. Tension fatigue load was applied in a servo-hydraulic testing machine. The loading frequency was 10 SK’ at an R ratio of R= u,~,,/u, = 0.1 (a,, is the upper and Umi”the minimum fatigue stress in each consecutive load cycle). An additional fretting load was applied using a specially designed fretting fatigue test device [7]. The fretting pins were pressed against the specimen surfaces at a fixed position. The contact load was adjusted by a spring mechanism and controlled by a load cell. While the upper part of the specimen was gripped in a fixed cross head, to which the fretting fatigue apparatus was also attached, the lower part of the specimen was gripped to the actuator. Due to the load induced straining of the laminate, relative cyclic motion between fretting pin and specimen occurs. The relative slip between the pins and the laminate can be estimated according to eqn. (1)

where Au is the stress amplitude, UT is the ultimate tensile strength (850 MPa), ETis the strain to failure (1.2%) and Z1is the distance of the fretting pins from the upper, .fixed clamping grip. For a stress amplitude of 700 MPa (upper fatigue stress a,, = 780 MPa), eqn. (1) gives a slip of 650 pm. Measurement with a mechanical displacement transducer (accuracy about 10 pm) resulted in a value of about 700 pm, which is slightly higher than the calculated value because the laminate already


performs a small strain motion within the clamping grips. The slip could be changed by varying the pin position Ii along the laminate beam. Additional fretting wear tests were performed with a special testing system, described in detail elsewhere [ 51. Metallic pins with the same geometry and preparation as for fretting fatigue tests were pressed onto the specimen surfaces with a pre-selected contact load and forced to an oscillatory sliding motion parallel to the specimen surface. The oscillation width and frequency were adjustable. The wear was measured by weighing each specimen before and after the tests. X-ray radiography visualized the development of cracks and delaminations inside the laminate. For this purpose, the edges of the specimens were spread over with a zinc iodide solution as a contrast agent which easily penetrated into cracks and dehnninations.

3. Results

Figures 4 and 5 show Wi%ler diagrams (upper fatigue load apphed vs. resulting lifetime) for fatigue and fretting fatigue of the laminates. In semilogarithmic representation, the pIain fatigue curves can be approached by a linear function [ 81 according to eqn. (2) c+=a&




where aT is the ultimate tensile strength and m is the experimental parameter. In the present case, m= 0.02 for Ll and 0.06 for L2, respectively. Obviously, L2 reacts much more sensitively to a change of the fatigue load than Ll. 700 iii h 8 E tj 4 .;



IA & g

4001 104




pins) of laminate





Load Cycles to Failure, N Fig. 5. Upper fatigue load vs. lifetime


for plain fatigue and fretting

fatigue (against aluminium


Fig. 6. Schematic of the stress distribution (D) of neighbouring 90” plies.

in a 0” layer after cracking (C) and local delamination

Application of an additional fretting load leads to a deviation of the Wijhler curve from the simple logarithmic rule of eqn. (1). This deviation may strongly depend on the particular loading conditions. The hard steel pins as counterparts did not produce any significant fretting fatigue effect up to a contact pressure of 23 MPa (F,=450 N). When the fretting pins consisted of aluminium, the fretting fatigue life of the laminate strongly depended on the contact pressure. A similar effect was found for brass pins. Initially, the shape of the curve will be regarded only in general. Based on studies of the failure processes a quantitative measure for the fretting fatigue damage will be developed. Subsequently, the influence of the contact pressure will be considered in detail. 3.2. Damage mechanisms The development of fretting fatigue damage in homogeneous, isotropic materials can be described by means of fracture mechanics concepts [9, 101. In the area of the fretting contact, local stress peaks due to friction and normal forces initiate cracks. These cracks act as sharp notches and lead to a high concentration of the fatigue stress at their tips. As a result, the cracks propagate into the bulk and, thus, cause premature failure. In continuous fibre reinforced composites, the situation is considerably different. The simple application of fracture mechanics is not possible because cracks do not always propagate perpendicular to the main loading direction but advance preferentially parallel to the fibres [ 111. Neither the initiation and accelerated growth of single cracks, as in homogeneous, isotropic materials, nor morphological changes characterize failure mechanisms in laminated composites under fatigue loading [ 121. Instead, multiple matrix cracking along the fibres causes a reduced load carrying capacity of the offaxis plies and accordingly enhanced stresses in the 0” layers. Delamination between the differently oriented plies develop starting from intersections of matrix cracks and from edges [ 131. Subsequently, the stress in the off-axis plies is reduced while the 0” plies have to carry an increasing part of the applied load. Final failure occurs when the stress in the 0” plies locally exceeds their strength [13] which may additionally be reduced by random cracking of 0” fibres [ 141. Figure 6 schematically illustrates the stress situation in a 0” layer of a cross ply laminate containing transverse cracks in the 90” plies. Laminate L2 contains a large number of 45” plies, which have a higher load carrying capacity than the 90” layers in Ll. Therefore, it is reasonable that the residual strength after a given number of load cycles is more reduced in Laminate L2 than in Ll. The effect of an additional fretting component on the fatigue damage mechanisms of CF/EP laminates is described in detail in a previous publication [ 151. Only some essential features will be reported here and connected to a closed view of fretting fatigue damage development. The X-ray radiographs in Fig. 7 show several damage states of laminate Ll. Figure 7(a) represents a specimen prior loading: the radiograph exhibits homogeneous darkening




Fig. 7. X-ray radiographs of specimens of laminate Ll: (a) prior loading, (b) after exposing to plain fatigue and (c) after fretting fatigue.




El Fig. 8. Optical micrographs of specimens (laminate Ll) after fretting fatigue (a) vs. an aluminium pin (FN = 400 N, 30 000 load cycles) and (b) US. a steel pin (FN= 450 N, 270 000 load cycles).

indicating the absence of any cracks or delaminations. After plain fatigue loading (Fig. 7(b)), the specimen contains transverse matrix cracks in the 90” layers together with some longitudinal cracks in the 0” plies. Figure 7(c) is taken from a specimen after fretting fatigue loading. The white spot at the fretting contact suggests that pressing the pins against the specimen hinders delamination. However, strong longitudinal cracks and delaminations starting from the fretted region grow along the O”-90” interphase. The optical micrograph in Fig. 8(a) shows that this is due to the peeling-off of cracked bundles of 0” fibres. The cracking of 0” layers could be caused by high equivalent stresses in the region of the fretting contact. However, this explanation does not seem to be very probable because the applied apparent contact pressures


(less than 20 MPa) are very small in comparison with the fatigue stress level (greater than 700 MPa). In fact, when a contact pressure of 23 MPa is applied via steel pins instead of aluminium pins, no cracking of the 0” plies was found (Fig. S(b)). Obviously, the effect of the normal load (in the regarded range), with which the pins are pressed against the specimens, on the equivalent stress is not the crucial parameter which controls the development of the fretting fatigue damage. Another assumption considers the initiation and advance of fretting fatigue damage to be controlled by the fretting wear performance of the laminate. This explanation is supported by the fact that the fretting wear of CF/EP vs. aluminium is more severe than against steel counterparts [5] according to the more detrimental effect of the aluminium pins on the fretting fatigue performance. Figure 9 presents a scanning electron micrograph of a CF/EP surface subjected to plain fretting wear VS. aluminium. Besides polishing of fibres and matrix, cracking and removal of broken fibres are visible. From the area denoted by the letter D, a chip of about 100 pm in width is delaminated. This very severe type of wear mechanism occurs only if the loading parameters (pressure, amplitude, frequency) exceed characteristic critical values, which mutually depend on each other. This delamination wear was not observed for steel counterparts up to an apparent contact pressure of about 40 MPa [ 161. Once the fibres or fibre bundles are broken, they can no longer support the load carrying capacity of the laminate although they are actually not yet worn. Therefore, it must be expected that the fretting fatigue damage proceeds considerably faster than the plain fretting wear gravimetrically measured. When the specimen shown in Fig. 9 is simultaneously exposed to a tensile fatigue load, shear stresses arise along the interface between the cracked and the undamaged fibre bundles which are enhanced by the friction force (Fig. 10). This leads to the observed delamination and peelingoff of the cracked 0” layers (Fig. 8(a)) [ 171. This delamination was also

Fig. 9. Scanning electron micrograph of a CF/EP surface subjected aluminium. FN = 400 N @ = 20 MPa), peak-to-peak cycles.


Fig. 10. Schematic illustration of delamination of cracked delamination is forced by crack opening mode I and II.

to fretting wear against width A = 700 pm, 110 000 load



The advance



observed when specimens predamaged by plain fretting were exposed to a quasi-static tensile stress. The following conclusions can be drawn: (1) An additional fretting component is able to drastically reduce the fatigue life of a CF/EP laminate if the fibres exposed to fretting possess a 0” orientation. (2) Fretting fatigue of continuous fibre reinforced laminates cannot be treated in terms of fracture mechanics descriptions of crack initiation and growth. Notch effects in the classical meaning do not occur. (3) In the range of contact pressures considered, an additional fretting component influences the fatigue performance of a laminate primarily not via the enhancement of the equivalent stresses due to the contact stresses but by fretting induced surface damage. (4) Fretting fatigue damage penetrates faster into the laminate than the pure material removal due to plain fretting wear. Fibre bundles which are pre-damaged by fretting tend to break and peel-off under fretting fatigue conditions. 4. Discussion 4.1. ~at~t~caL mod& for damage d~elop~t The conclusions drawn from the observation of the fretting fatigue failure mechanisms suggest: (1) The absence of notch effects causes a relatively uniform distribution of tensile stresses across any cross section of the undamaged 0” plies. Therefore, the change of the load carrying capacity dF should be proportional to the reduction of the cross section dQ of the load bearing 0” layers in the fretting contact region between n and N+W load cycles dF=udQ


where g is the upper fatigue stress at which the specimen fails after N load cycles. (2) Since a notch effect with any accelerated crack growth can be denied, the damage development can be considered to proceed propo~ional to the time:

Actually, the reduction of the cross-section of the 0” layers in the fretted region leads to an enhanced fatigue stress in the remaining 0” fibres. The specimen finalIy fails when this stress exceeds the strength of the 0” plies. However, the actual cross-section is not continuously monitored and, thus, the .+w stress in the laminate between the fretting pins cannot be determined. But the consideration can be turned out as follows: the apparent upper fatigue stress a,, at which the specimen fails after N load cycles, diminishes propo~on~y with the cross section of the load bearing 0” layers. Accordingly,


a,, can be calculated by replacing eqn. (2) by

where Q0 is the initial total cross-section of the 0” plies. The first term in brackets considers the reduction of the tensile strength by reducing Q0 to Q. Inserting eqns. (5) and (4) into eqn. (3) gives

Integration of eqn. (6) from the upper fatigue load at which the specimen would fail after N load cycles without fretting, uF (according to eqn. (2)), to the upper fatigue load at which the specimen would fail after N load cycles under fretting fatigue conditions grr leads to




du,=$$$N+ 0



+alog,$g 0

1+ 0



4&o do

or dehning a parameter Au,,, ~F-%F

Ao;~,= =-

dQ Qo do

where or is the upper fatigue load at which the specimen fails after N load cycles and a, is the upper fretting fatigue load at which the specimen fails after N cycles. The first term indicates that to a first-order approximation the laminate behaves as if its cross-section was reduced prior to fatigue loading. The quadratic term considers the successive nature of the progress of fretting fatigue damage. The last term indicates that the fibres removed at the beginning of the test possessed a somewhat higher load carrying capacity than the fibres which experienced a longer fatigue loading. However, this last term is smaller than 0.003 while the first term is in the range of 0.1. Therefore, the last term will be neglected in the further considerations. The physical reason for this is the high resistance of the carbon fibres against fatigue loading. Conclusively, one can write Aa



It should be emphasized that eqn. (8) does not enable the deduction of the fretting fatigue life of a laminate from simple material and loading parameters. However, Au,,~ gives a quantitative measure for the degree of fretting fatigue damage and it provides a tool for the investigation of fretting fatigue damage development.


Presuming the validity of eqn. (4), eqn. (8) suggests that, to a firstorder approximation, AUK,,increases proportionally with the time. To check this assumption, Fig. 11 presents a plot of AC,,, VS. the fretting fatigue life N for two different laminates. Initially, all curves increase proportionally with time but reach a constant value after a certain number of load cycles. Maximum fretting fatigue damage is expected when the covering 0” layers are totally penetrated by the fretting pins. Further fretting damages only offaxis plies and, thus, influences the fatigue performance of the laminate insignificantly [ 71. Figure 12 presents an optical micrograph of a specimen of laminate L2 which failed after 345900 load cycles, after reaching the plateau region of the Av~,~VS. N curve. The covering 0” layers are totally penetrated and fretting already took place on the 45” plies. Based on these considerations, the maximum value for Au,,~ can be calculated from the maximum reduction of the cross-section of the 0” layers (see Pig. 13)

d kmax


-+ w &t

where d is the diameter of the pm (5 mm), w is the width of the specimen (6.3 for L2 and 8.4 for Ll), iF,_ is the maximum number of fretted 0” plies (4) and it,, is the total number of 0” plies (8 for Ll and 24 for L2). The according maximum fretting fatigue damage can be calculated by using eqn. (8) __ a)





Laminate Al, F,


= 400 N

o Al, F,, = 300 N


200-000 ‘I

Load Cycles to Failure, N


0 @>






Load Cycles to Failure, N

Fig. 11. Relative fatigue strength reduction (AC,,) as a function of fretting fatigue life for laminate Ll (a) and L2 (b) at several contact loads.

Fig. 12. Optical micrograph of a specimen of laminate L2, subjected an aluminium pin. F,=450 N, failure after 345900 load cycles. Fig. 13. Cross-section

of the specimen

in the region

to fretting fatigue against

of fretting contact.

0.25 for Ll = { 0.125 for L2 These values are in good agreement with the experimental results (Fig. 11). However, at lower contact pressures, the maximum fretting fatigue damage is considerably smaller than calculated. This effect will be presented and discussed later in this paper. From the linear part of the Acr,,, US. N curve, the rate of damage development dQ/dN can be derived. For reasons of comparability with the fretting wear tests, this quantity will be transformed to a “specific pseudowear rate” ti,* in analogy to the concept of the specific wear rate ti, rr dAQ 16 FNAN

‘*_ w, _-=- Av W,L

where AV is the volume of removed or cracked 0” fibres, L is the total sliding distance (2AN) and FN is the normal load. The factor l/2 considers that two pins simultaneously rub against the laminate and only one half of the accumulated damage of the laminate can be related to one pin. Resolving eqn. (8) to AQ and inserting this result into eqn. (10) gives ti,*=



[ 1 - (1 - 2 AC~~#~]


zir,* quantitatively measures how deep the fretting fatigue damage penetrated into the laminate and includes material removal due to pure fretting wear as well as cracking of predamaged fibres and fibre bundles. Therefore, ti,* should be greater than ti,.


4.2. Experimental


of the model

In order to avoid confusion, is should be emphasized that the measured points along the Au,,, vs. N curve represent different specimens which were subjected to different fatigue stress amplitudes. The fact that these points follow a straight line primarily means that the fretting fatigue damage accumulated until final failure is proportional to the fretting fatigue life, regardless of the fatigue stress level. However, the presented model states that during the test of one particular specimen the effective cross-section of the load-bearing 0” layers, when exposed to fretting, decreases proportionally with time. Therefore, an additional test was necessary to check this assumption of the model. Several specimens of laminate Ll were exposed to a fretting fatigue loading equivalent to one half of the expected fretting fatigue life. The contact load was set to FN = 300 N, the upper fatigue stress was u,,= 700 MPa. From Fig. 4 a lifetime of 79 000 load cycles can be expected under these conditions. After 45000 load cycles, the test was interrupted according to a theoretical fatigue strength reduction of 4.5% (Fig. 11). Subsequently, the specimens were continued under plain fatigue conditions (a,= 740 MPa) until final failure. Table 2 lists the resulting lifes and the according relative deviation from the u vs. log N curve for plain fatigue of the undamaged laminate (Au,&. The measured mean value of AC,,, approximates to 4.5%. This is in good agreement with the theoretical value which was calculated under the assumption that the fretting fatigue damage proceeds proportional to time during one single test. 4.3. Eflect

of experimental conditions 4.3.1. InJkence of slip amplitude

Aluminium pins were pressed onto laminate Ll at three different positions (11=40, 65, and 90 mm). According to eqn. (l), each pin position correlates with a slip amplitude at a given stress amplitude. Figure 14 plots the calculated slip amplitude A for all three pin positions vs. the upper fatigue stress a,. The applied upper fatigue stress varied between 650 and 700 MPa. The contact pressure was set to FN= 400 N @ = 20 MPa). Table 3 lists the resulting lifetimes of the laminate Ll. It can be seen that under the selected conditions no significant effect of the slip amplitude can be observed. TABLE 2 Residual fatigue life and relative fatigue strength reduction of CF/EP laminates (q = 740 MPa) pre-damaged by 45 000 load cycles under fretting fatigue (FN = 300 N, o, = 700 MPa) conditions No.

Number of fatigue cycles to failure


1 2 3

7320 29970 15630

5.1 3.8 4.5


A~;,,.mean W)











14. Slip amplitude vs. applied upper fatigue load.

TABLE 3 Fretting fatigue life of laminate Ll (number of cycles to failure) for different pin positions 1, and fatigue stress levels q, (counterpart: upturn, F,=400 N)




16600 14500

15000 12000

18000 -

6800 5400

6800 6400

6500 5800







4.3.2. E,ffect of contact pressure and counterpart material Figure 15(a) shows the specific pseudo-wear rate ti,* calculated from the linear part of the AUK,.vs. N curve of laminate LI as a function of the contact pressure p. Obviously, there exists a critical value (boundary value) where the propagation of damage development jumps from an insignificant to a very high level. The specific fretting wear rate io, follows a similar behaviour (Fig. 15(b)). For ahuninium pins, &J,*is smaller than 8 x 10F6 mm3 Nm- ’ at a contact load of 10 MPa (FN= 200 N) and, thus, has a similar magnitude as the specific fretting wear rate (6 X 10W6 mm3 Nm- ‘). At 15 MPa, still below the boundary value of the contact pressure, ti, * is already considerably increased to a value of about 3 X 10e5 mm” Nm-’ while the specific fretting wear rate ti, remains constant below is boundary value. Conclusively, one can say that fretting fatigue damage proceeds as fast as material removal due to plain fretting wear below a critical contact pressure. Above this critical contact pressure, fretting wear and fretting fatigue is accelerated. However, under fretting fatigue conditions, this boundary pressure lies at considerably lower values bcrit= 17 MPa) than under plain fretting (petit= 30 MPa for alum~ium pins). Above the boundary value of the contact pressure, fretting fatigue damage proceeds much faster than the pure material removal due to plain fretting wear ti,*Clr,=MPa)=5x10-4

mm” Nm-’






Fretting Fatigue





Fig. 15. Specific pseudo-wear rate (a) and specific fretting wear rate (b) of laminate function of contact pressure for different counterpart materials.

Ll as a

ti,(p = 45 MPa) = 2.8 x 10e5 mm3 Nm-’ behaviour of the specific pseudo-wear rate as a function of contact pressure can be found for the brass pins. However, the boundary value of the pressure is shifted to lower values, compared with aluminium. This correlates with the fretting wear of CF/EP which was found to be slightly higher for brass than for aluminium counterparts. When steel pins were used as counterparts, no stepwise increase in the specific pseudo-wear rate was observed up to a contact pressure of 23 MPa (FN=450 N). Perhaps, the boundary value of the pressure for steel pins lies at higher values than those which could be achieved with the apparatus employed. A probable reason for the small influence of the hard steel pins on the fatigue life of laminate Ll is their high resistance against abrasion by fibre debris. The surface of the steel pins remains rather smooth during fretting and, in turn, acts only little abrasive to the sample material. Aluminium pins, however, can easily be roughened by the fibre debris but the arising asperities are rather soft and, thus, less abrasive than in the case of brass. A further damage mechanism in the fretting contact could emerge from electrical contact corrosion. However, the experimental results do not support any significant role for electrical contact corrosion in fretting wear of CF/ EP and under the contact conditions used because a similar trend of the effect of counterpart material was observed for fretting wear of glass fibre A similar


(GF/EP) composites. Table 4 compares the specific fretting wear rates of CF/EP and GF/EP worn against steel and aluminium, respectively. Fretted against steel, CF/EP and GF/EP exhibit similar wear rates. Changing to aluminium pins, the wear of GF/EP is much more accelerated than that of CF/EP. This cannot be explained by contact corrosion which is not effective in the case of glass fibres. But the glass fibre particles are more abrasive and, thus, more effective in roughening the counterpart. Another effect of the contact load can be seen in Fig. 1 l(a). At a contact pressure of p = 15 MPa (FN = 300 N), the relative fatigue strength reduction AU,,, becomes constant after about 150 000 load cycles at a value of AU,,, = 14% and does not reach the theoretical value of 25% as it was found for p = 23 MPa (FN = 450 N). Several authors [ 18-201 have reported that in metallic materials the influence of an additional fretting component on material fatigue is also effective only during crack initiation and early stage of crack growth. After a certain number of load cycles, the crack propagation rate under fretting fatigue approaches that under plain fatigue [ 211. The reason is that, while the crack propagates into the interior, the stress concentration at the crack tip diminishes from an initial value, which is enhanced by normal and shear stresses applied by the fretting pin, to that value which is also active under plain fatigue conditions [ 221. This crack propagation behaviour would be in agreement with the fact that AU,,, remains constant after a certain number of load cycles. However, the absence of notch effects in continuous fibre reinforced layers with 0” orientation opposes the application of this model to the CF/ EP laminates investigated here. In fact, it was observed that specimens exposed to fatigue stresses above 640 MPa (contact load FN = 300 N) exhibited a more pronounced surface damage after a given number of cycles than specimens which were fatigued at a lower stress level. Especially, no peelingoff of fibre bundles occurred at fatigue stress levels below 640 MPa even after some hundred thousands load cycles (Fig. 16(a)). These observations suggest that the falling of the fatigue stress level below a critical value for the diminution of the propagation rate (~“,.ri, = 640 MPa) is responsible of fretting fatigue damage rather than the exceeding of a special number of load cycles. In this case, probably, tensile stresses in the fretting region and

TABLE 4 Relative specific fretting wear rates for several sample/counterpart combinations to that of CF/EP us. steel (I$= 300 N, 500 wrn peak-to-peak oscillation width, cycles)




1 1.3

1.8 26.7

normalized 72000 load


shear stresses between cracked and undamaged layers are not high enough to cause amplification of the surface damage due to fretting. In this context, it should be remembered that the several points along the Au,,, curve (Fig. 11) represent differently loaded specimens. The damage progression in the specimens which failed in the plateau region may not have proceeded along the drawn line but by any other course with a smaller slope which ends at the measured points. According to the above considerations on damage development, it can be assumed that the Ao;,, VS. N curve for each individual specimen follows a straight line until final failure. Table 5 represents an experimental check of this concept. Some specimens were exposed to fretting fatigue at an upper fatigue stress of 640 MPa and a contact pressure of p = 15 MPa (FN= 300 N) for 200000 load cycles. According to Fig. 4, the expected fretting fatigue life of these specimens amounts to about 500000 load cycles. Figure 16(a) depicts that these specimens did not exhibit any peeling-off after 200000 load cycles in contrast to specimens subjected to higher fatigue loads. Presuming a linear increase of Au,,~ until final failure after 500000 load cycles at AU,,,= 14%, a relative fatigue strength reduction after 200 000 load cycles of about 5.5% would be expected (see Fig. 11). Subsequent to fretting TABLE 5 Relative fatigue strength reduction due to fretting fatigue pre-damage: specimens were first subjected to 200000 fretting fatigue load cycles (F,=300 N, (r,=640 MPa) and subsequently were exposed to plain fatigue (a,= 750 MPa) until failure


Number of cycles to failure

Au,,, (No>


1 2 3

390 1380 780

6.6 5.4 6.0



Fig. 16. Optical micrographs of laminate Ll after exposure to fretting fatigue. Specimen (a) was exposed to fretting fatigue against aluminium (FN=300 N) at an upper fatigue load of gU= 640 MPa for 200 000 load cycles. Afterwards it was run to failure (b) without fretting at an upper fatigue load of 750 MPa.

fatigue loading, the specimens were run under plain fatigue at a;, = 750 MPa until final failure occurred. During this time, the layers pre-damaged by fretting cracked rapidly and delaminated (Fig. 16(b)). Table 5 lists the number of plain fatigue cycles to failure and the resulting deviation from the cU US. log N curves for the undamaged laminate. The mean value for this difference is AcI,,~= 6%. The coincidence of this result with the expected value confirms the assumptions that (1) fretting fatigue damage proceeds proportional with time in continuous fibre reinforced laminates and (2) at low contact loads (F,~300 N), the fretting fatigue damage propagation rate is controlled by the upper fatigue stress a,,. flowever, for high contact loads (FM&400 N) the propagation rate is independent of the fatigue stress level (see above). This can be explained by assuming the equivalent stress amplitude in the subsurface region instead of the fatigue stress level to be responsible for the amphfication of fretting induced surface damage. 4.3.3. Effect of surface mcwphology All of the experiments described above were performed with a flat-onflat fretting contact geometry. Of course, it is very di~cult to adjust these flat surfaces really parallel and, if the contacting surfaces include a small angle, the initial contact pressure distribution is inhomogeneous and can exhibit peaks, as illustrated in Fig. 17. Therefore, the fretting pin locally penetrates very fast into the laminate at the beginning of the fretting fatigue loading. If the fibres are protected by a resin-rich surface layer due to the use of peel plies (Fig. 2), the real contact area is significantly increased (Fig. 18) and the contact stress peaks are reduced before the fretting pin touches the fibres. Laminates which were produced without using peel-plies do not possess the resin-rich surface layer. The contact pressure peaks due to fretting pin/

Fig. 17. Schematic presentation of a contact stress peak due to small misadjustment of pin and specimen. Fig. 18. Surface of laminate Ll (manufactured with peel ply) after exposure to fretting fatigue vs. aluminium (F,=300 N) for about one hundred load cycles. At the beginning the pin did not touch the surface over the whole contact area. While the pin penetrated into the resin rich surface layer, the real contact area increased.


specimen m&adjustment act directly on the fibres. As a result, very fast wear with subsequent cracking of fibre bundles can be expected at the beginning of the test. Figure 19 shows the resulting Au,,, ZIS.N curve. The curve in principle exhibits a similar shape as for the laminates with peel plies (Fig. 11(a)). However, the intersection with the Au,,~ axis is shifted from 0 to about 6%. Subsequently, the curve again increases linearly with time and reaches a plateau. The maximum value of Acr,,, is also shifted by about 6% to higher values compared with the laminates produced with peel ply so that the span of AU,,~remains at about 14%.

5. Conclusions (1) An additional fretting component may drastically reduce the fatigue life of an CF/EP laminate if the fibres subjected to fretting are under 0” orientation. (2) The occurrence and magnitude of a fretting influence on fatigue performance sensitively depends on the particular loading conditions (counterpart material, contact pressure). The use of hard steel counterparts prevents a fretting fatigue effect, at least up to a contact pressure of 23 MPa under conditions given in the previously described experiments. For softer counterparts (aluminiurn, brass), a fretting fatigue effect occurred above a critical contact pressure. (3) As a quantitative measure for the degree of fretting fatigue damage, the relative fatigue strength reduction

proposed. A fretting fatigue damage propagation rate


c,* = 6 +$A N

[ 1 - (1 - 2Aq,,)1/2]

0 0




Load Cycles to Failure, N Fig. 19. Relative fatigue strength reduction of laminate Ll (without peel ply) vs. fretting fatigue life. Counterpart aluminium, FN= 300 N.


could be introduced in analogy to the concept of the specific wear rate. Based on this model, the fretting fatigue damage was found to penetrate proportional with time into the laminate. No notch effects were observed. (4) The mechanisms of interaction between surface damage due to fretting and fatigue are different for different loading conditions. Figure 20 schematically distinguishes three regimes of fretting fatigue. At low contact pressures (regime I), there exists no synergism between fretting and fatigue. The rate of damage propagation is small and determined by the resistance against plain fretting wear. The specimens live nearly as long as under plain fatigue. At higher pressures (regime II), the fretting fatigue damage already proceeds considerably faster than the pure material removal due to plain fretting wear. Fibre bundles exposed to fretting tend to crack and delaminate. However, if the fatigue stress descends below a certain value, the mutual amplification of fretting and fatigue damage becomes less effective because cracking and peeling-off of fibre bundles decellerates. Above a critical contact pressure (regime III), fretting fatigue damage proceeds about 15 times faster than the pure material removal due to fretting. Fibre bundles which are predamaged by fretting rapidly crack and peel off. The fretting fatigue damage proceeds until the fretting pins reach off-axis plies, which carry only a small part of the applied fatigue stress. The particular value of the critical contact pressure in Fig. 20 depends on the hardness of the counterpart material. For hard steel pins, the boundary pressure is beyond the sensitivity range of the employed fretting fatigue apparatus used, for soft aluminium counterparts it is near p,,,,= 1’7 MPa. The lowest value (pcrit = 12 MPa) for the critical pressure was found for brass pins with the intermediate hardness. The intermediate hardness acts most detrimental because these pins can easily be roughened by fibre debris but the arising asperities are harder and, thus, more abrasive than in case of softer counterparts. (5) Some comparative experiments showed that laminates manufactured without peel plies are more sensitive to fretting fatigue than other laminates. This is especially true for the early stage of fretting fatigue loading. The resin-rich layer, caused by the use of peel plies during curing of the laminate,


I I:‘Regime

Contact Fig. 20. Schematic contact pressure.




of the course

of specific


rate as a function


187 assures a more homogeneous contact pressure distribution until the fretting pins touch the fibres. This may be of practical importance because the cheap and simple step to produce a resin-rich coating, which itself needs not to be very wear resistant [ 51, may significantly improve the fretting fatigue performance of a laminate. Acknowledgments

This work was financed by the German Ministry for Science and Technology (Project 03 M 1022). Thanks are due to the BASF and MBB who supplied the sample materials. Professor G. Marom (The Hebrew University of Jerusalem) and Dr. H. D. Wagner (The Weizmann Institute, Rehovot) contributed to the advance of the work by helpful discussions. References

10 11 12

13 14


R. B. Waterhouse (ed.), Fretting Fatigue, Applied Science, London, 1981. Special issue 012 fretting wear and fretting fatigue, Wear, 125 (1988). R. Heinz and G. Heinke, Die Vorgange beim Schwingungsverschleiss in Abhangigkeit von Beanspruchung und Werkstoff, in Tribology - Reibung, Verschleiss, Schmierung (Documentation of the German Ministry for Science and Technology), Springer, Berlin, 1981. N. Ohmae, K. Kobayashi and T. Tsukizoe, Characteristics of fretting of carbon fibre reinforced plastics, Wear, 29 (1974) 345. 0. Jacobs, K. Friedrich, G. Marom, K. Schulte and H. D. Wagner, Fretting wear performance of glass-, carbon-, and aramid-fibre/epoxy and peek composites, Wear, 135 (1990) 207. K. Schuite, K. Friedrich and S. Kutter, Fretting fatigue studies on carbon fibre/epoxy resin laminates, part II: effects of a fretting component on fatigue life, Comp. Sci. Technol., 30 (1987) 203. K. Friedrich, S. Kutter and K. Schulte, Fretting fatigue studies on carbon fiber/epoxy resin laminates, part I: design of a fretting fatigue test apparatus, Comp. Sci. Technol., 30 (1987) 19. K. L. Reifsnider, K. Schulte and J. C. Duke, Long term fatigue behavior of composite materials, in T. K. O’Brien (ed.), Long Term Behavior of Composites, ASTM STP 813, ASTM, Philadelphia, 1983, p. 136. P. R. Edwards, The application of fracture mechanics to predicting fretting fatigue, in R. B. Waterhouse (ed.), Fretting Fatigue, Applied Science, London, 1981. D. A. Hilis, D. Nowell and J. J. O’Conner, On mechanics of fretting fatigue, Wear, I25 (1988) 129. J. Krey, K. Friedrich and K.-H. Schwalbe, Fracture toughness and fatigue crack propagation of single fibre-bundle reinforced model composites, J. Mater. Sci. Lett., 6 (1987) 851. K. Schulte, Damage development under cyclic loading, in I. Verpoest and M. Wevers (eds.), Proc. Eur. Symp. onDamage Development andFailurel+ocesses in CompositeMaterials, Leuven, Belgium, 4-6 May 1987, p. 39. T. K. O’Brien, M. Rigamonti and C. Zanotti, Tension fatigue analysis and life prediction for composite laminates, Int. J. Fatigue, II (1989) 379. M. Bader, Modelling fiber and composite failure, in I. Verpoest and M. Wevers (eds.), hoc. Eur. Symp. on Damage Development and Failure Processes in Composite Materials, Leuven, Belgium, 4-6 May 1987, p. 8. K. Schulte, K. Friedrich and S. Kutter, Fretting fatigue studies on carbon fibre/epoxy resin laminates, part III: microscopy of fretting fatigue failure mechanisms, Camp. Sci. Technol., 33 (1988) 155.

188 16 0. Jacobs, K. Friedrich and K. Schulte, Schwingverschleiss von kohlenstoff-, aramid und glasfaserverstarkten Epoxidharzund PEEK-Verbundwerkstoffen, in K.-H. zum Gahr (ed.), Reibung und Vwschleiss bei metallischen und nichtwwtallischen Wwkstoflen, Bad Nauheim, 1990, p. 199. 17 M. D. Thouless, H. C. Cao and P. A. Mataga, Delamination from surface cracks in composite materials, J. Mater. Sci., 24 (1989) 1406. 18 K. Endo and H. Goto, Initiation and propagation of fretting fatigue cracks, Wear, 38 (1976) 19 J. A. Alit and A. L. Hawley, On the early growth of fretting fatigue cracks, Wear, 56 (1979) 377. 20 K. Sato, Damage formation during fretting fatigue, Wear, 125 (1988) 163. 21 K. Sato, H. Fqjii and S. Kodama, Crack propagation behavior in fretting fatigue, Wear, 107 (1986) 245. and propagation of fretting fatigue cracks”, 22 D. W. Hoeppner, Comments on “initiation Wear, 43 (1977) 267.