Вы находитесь на странице: 1из 22

04-R-08 edit 6/3/04

4:27 PM

Page 1

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER:

A LITERATURE SURVEY

W. V. M ARS*

ADVANCED RESEARCH ENGINEER COOPER TIRE & RUBBER COMPANY, 701 LIMA AVE., FINDLAY, OHIO, 45840

A. FATEMI

UNIVERSITY OF TOLEDO, TOLEDO, OHIO, 43606

ABSTRACT

Many factors are known to influence the mechanical fatigue life of rubber components. Four major categories of factors are reviewed here: the effects of mechanical loading history, environmental effects, effects of rubber formulation, and effects due to dissipative aspects of the constitutive response of rubber. For each category, primary factors are described, and existing literature is presented and reviewed. Rubber’s fatigue behavior is extremely sensitive to both the maximum and minimum cyclic load limits. Other aspects of the mechanical load history are also discussed, including the effects of static loaded periods (“annealing”), load sequence, multiaxiality, frequency, and loading waveform. Environmental factors can affect both the short and long term fatigue behavior of rubber. The effects of temperature, oxy- gen, ozone, and static electrical charges are reviewed. A great range of behavior is available by proper manipulation of formulation and processing variables. Effects of elastomer type, filler type and volume fraction, antidegradants, curatives, and vulcanization are discussed. The role of dissipative constitutive behavior in the improvement of fatigue properties of rubber is also reviewed. Four distinct dissipative mechanisms are identified, and their effects on fatigue behavior are described.

CONTENTS

I. Introduction

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.392

II. Mechanical Load History

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.393

A. Alternating Load and Maximum Load

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.394

B. Minimum Load, Mean Load and R-Ratio

.

.

.

.

.

.

.

.

.

.

.

.

.

.396

C. Statically Strained Rest Period

 

.397

D. Load Sequence

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.398

E. Multiaxiality

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.398

F. Loading Frequency and Waveform

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.401

III. Environmental Conditions

 

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.402

A. Temperature

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.402

B. Ozone

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.402

C. Oxygen

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.403

D. Electrical Charges

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.404

IV. Rubber Formulation

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.404

A. Polymer Type

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.404

B. Filler

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.406

C. Antidegradants

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.406

Vulcanization

D. .

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.406

V. Constitutive Behavior

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.407

A. Mullins Effect

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.408

B. Strain-Crystallization

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.408

C. Rate Independent Hysteresis

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.408

D. Viscoelasticity

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.408

* Corresponding author. Ph: 419-423-1321; email: wvmars@coopertire.com

391

04-R-08 edit 6/3/04

4:27 PM

Page 2

392

RUBBER CHEMISTRY AND TECHNOLOGY

 

VOL. 76

 

VI. Conclusions

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.409

VII. References

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.

.410

I. INTRODUCTION

Understanding the factors that affect the fatigue life of rubber is helpful in at least three ways. First, at least a qualitative understanding of these factors is required in order to develop durable rubber compounds, rubber components and systems that rely upon rubber components. Second, the analysis and prediction of fatigue failure in rubber components depends on a quan- titative understanding of these factors. Finally, it should be acknowledged that many aspects of the fatigue failure process remain incompletely understood. Future progress in understanding the fatigue behavior of rubber requires the identification of that which is already known and, to the extent possible, that which remains unknown. Prior reviews 1-8 identify many factors that affect the fatigue crack nucleation and growth process in rubber. This article updates existing reviews to reflect recent and previously unnoticed developments, offers new interpretations of existing studies and theories, and identifies areas where additional research is needed. The present review will not address approaches for fatigue analysis and testing in detail, except as required to discuss specific effects. A recent review dis- cussing crack nucleation and crack growth analysis approaches for rubber is available else- where. 9 Factors that influence the fatigue life of rubber are broadly classified in this review as being related to the mechanical loading history, to environmental conditions, to formulation of the rub- ber compound, and to certain aspects of the stress-strain constitutive behavior. Primary consid- erations with regards to the mechanical loading history include both the maximum and minimum oscillation limits. Other aspects of the mechanical load history are also discussed, including the effects of static loaded periods (“annealing”), load sequence, multiaxiality, frequency, and load- ing waveform. Environmental factors can affect both the short and long term fatigue behavior of rubber. The effects of temperature, oxygen, ozone, and static electrical charges are reviewed. A great range of behavior is available by proper manipulation of formulation and processing vari- ables. Effects of elastomer type, filler type and volume fraction, antidegradants, curatives, and vulcanization are discussed. The role of dissipative constitutive behavior in the improvement of fatigue properties of rubber is also reviewed. Four distinct dissipative mechanisms are identified and their effects on fatigue behavior are described.

II. MECHANICAL LOAD HISTORY

By definition, mechanical fatigue involves crack nucleation and growth due to fluctuating loads. In the following discussions, the general term “load” will be used to denote the mechani- cal severity, or “driving force”, as quantified by any of various parameters associated with spe- cific analysis approaches: strain, stress, strain energy density, energy release rate, etc. Simple mechanical histories applied to lab specimens are usually characterized by parameters such as maximum, alternating, minimum and mean loading, and/or the R-ratio. These parameters are illustrated in Figure 1. For example, one scheme for specifying the mechanical load history P(t) in terms of the alternating load P a and the mean load P m , for some parameter P, is:

( ) =

Pt

P

a

sin(ω ) +

t

P

m

(1)

Equivalently, one could use schemes given in terms of the maximum load P max and minimum load P min ,

04-R-08 edit 6/3/04

4:27 PM

Page 3

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

( ) =

Pt

P

max

1

+ sin(

2

t

ω

)

+ P

min

1

sin(

2

t

ω

)

or in terms of P max and the R-ratio, R = P min / P max ,

( )

P t

P

=+[

max

2

1

sin(

ω

tR)

+

(

1

sin(

t)

ω

)]

(2)

(3)

393

max 2 1 sin( ω tR ) + ( 1 − sin( t ) ω )

FIG. 1.— Commonly used parameters for the description of constant-amplitude mechanical loading history.

It can be seen that the parameters P a , P m , P max , P min , and R are interdependent, because fix- ing the value of any two of them uniquely determines the values of the remaining parameters. Two following subsections are therefore devoted to discussing the effects of these parameters. The discussion first addresses parameters that are traditionally presented as the primary variable against which crack growth rate or fatigue life data is plotted (i.e. alternating or maximum load). Next, the discussion addresses parameters that modify the influence of the aforementioned parameters (minimum load, mean load, or R-ratio). Complex histories that arise in applications usually cannot be characterized by unique val- ues of these simple parameters. Techniques for decomposing complex histories into a sequence of simpler histories (sometimes called a fatigue spectrum) are needed, but have not been dis- cussed extensively in the rubber literature. It should be mentioned that one instance was found in which the Rainflow procedure 10 was applied to fatigue analysis of a rubber mount. 11 The authors are aware of more than one industrial development center where such procedures are applied in the analysis of rubber vibration isolation components. As a full discussion of analysis approaches for variable amplitude fatigue is beyond the scope of the present work, and as very little discussion of this topic is made in the rubber literature, no further discussion is offered here. Clearly, research is required in this area. Other aspects of the mechanical load history that may influence the fatigue process include prior load history (the sequence of high and low severity events, prolonged periods of rest), load-

04-R-08 edit 6/3/04

4:27 PM

Page 4

3944 RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

ing rate, and waveform shape. Each of these factors is discussed below.

A. ALTERNATING LOAD AND MAXIMUM LOAD

In the earliest rubber fatigue studies, fatigue life was tabulated against alternating strain, fol- lowing approaches used originally in metals. 12,13 Tabulation against alternating load acknowl- edges the important difference between a static load (which does not cause fatigue failure, even at high, sub-fracture levels), and a fluctuating load (which can cause fatigue failure, even at low levels). Although large static loads do not cause fatigue failures (by definition), they can cause failures due to steady, time-dependent crack growth in the case of elastomers that do not exhib- it strain crystallization, or in the presence of environmental attack. In later development of the fracture mechanics approach for fatigue crack growth in rubber, the fatigue crack growth rate was tabulated against the maximum energy release rate attained during the cycle. 14 The applied history in that work was fully unloading cyclic tension (R = 0). Although the maximum load does not quantify, by itself, the magnitude of the fluctuation driv- ing the fatigue process, it seems likely that fatigue data were tabulated against maximum load for the following reasons. First, the maximum load is defined for a single instant, and clearly corre- sponds to a specific material configuration (the same may be said of the minimum loading). The amplitude, however, is not uniquely related to any single instant, or to a particular material con- figuration. In rubber, it is important to track the instantaneous material configuration, to deter- mine the state of embedded cracks (are they open or closed?), and to determine whether strain- crystallization will occur. Note also that the energy release rate of a crack (or any other positive definite fatigue correlation parameters, such as strain energy density or maximum principal strain) cannot be negative. This is in contrast to the implication of certain naive combinations of energy release rate amplitude and mean. Another reason for tabulating fatigue data against the maximum load is that the cyclically stabilized stress-strain response of filled elastomers depends heavily on the maximum load, through the Mullins effect. Other than the considerations given above, the choice of whether to tabulate fatigue data against alternating or maximum load appears to be somewhat arbitrary, so long as the minimum loading (or mean or R ratio) is also stated. Either parameter (maximum or alternating load) then completely specifies the cycle limits. After the fracture mechanics approach for fatigue crack growth in rubber was established, both fatigue crack growth and fatigue crack nucleation studies have been traditionally conducted under conditions of fully unloading cyclic tension (R = 0). The results are often tabulated against maximum loading. Note that under such conditions, the ampli- tude of the loading is one half of the maximum loading. Plotted as a function of maximum energy release rate, R = 0 fatigue crack growth behavior in rubber follows a roughly sigmoidal relationship, as shown in Figure 2. In this case, four regimes of behavior may be identified: a sub-threshold regime, a transition regime, a power-law regime, and a failure regime in which unstable crack growth occurs.

04-R-08 edit 6/3/04

4:27 PM

Page 5

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

395

Page 5 FA CTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER 395 F IG . 2.

FIG. 2. — Lindley’s 19 fatigue crack growth results for unfilled natural rubber, showing typical regimes of behavior, and the effects of R > 0 loading.

In the sub-threshold regime, when the maximum (for R = 0 conditions) load remains below the mechanical threshold, cracks or flaws do not grow due to mechanical fatigue, suggesting the possibility of infinite life. 15 Under such conditions, however, other factors may still eventually cause mechanical failure. These factors are generally environmental in nature, and include the growth of cracks under steady load (due to ozone or chemical attack), changes to the operating load (due to embrittlement of the elastomer), or changes to the fatigue crack growth threshold. When the loading is above the threshold required for mechanical crack growth, but below the critical value that results in sudden fracture, a power law relationship is commonly observed. A power law relationship is often observed both in fatigue crack nucleation tests (load vs. life) and in fatigue crack growth tests (load vs. crack growth rate). An analytical relationship between fatigue crack growth tests and crack nucleation tests in this regime has been derived. 16 Above a critical load, the fatigue crack growth rate accelerates rapidly to unstable fracture. For small cracks in crystallizing rubbers at high strains, the bulk material may strain crystallize

04-R-08 edit 6/3/04

4:27 PM

Page 6

396

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

before the point of unstable crack growth is reached. When this occurs, the strength can be increased by more than a factor of two. A detailed discussion of this phenomenon is given by Lake. 5

B. MINIMUM LOAD, MEAN LOAD AND R-RATIO

The effect of minimum or mean loading on fatigue life depends greatly on polymer and filler types. In rubbers that exhibit strain-crystallization, increasing the minimum strain has a signifi- cant beneficial effect, to a point. This effect has great practical significance, since rubber com- ponents often support a large static load, upon which smaller dynamic loads are superimposed. 13 In such situations, the load is rarely fully relieved, in contrast to typical lab studies. The benefit of increased minimum strain was studied first in natural rubber by Cadwell et al. 13 Several years later, Fielding 17 studied two newly developed synthetic rubbers, hoping to find a similar minimum strain effect. He found that Butyl-B exhibited the benefit, but that GR-S (SBR) rubber did not. In rubbers that do not exhibit strain-crystallization, increasing the mini- mum strain can have a deleterious effect, 17 but not necessarily. 18 Fielding identified strain crys- tallization as the mechanism responsible for the life improvement in Natural Rubber and in Butyl-B. The lack of strain-crystallization in the synthetic GR-S rubber presented a critical engi- neering challenge at the time, because World War II had severely reduced the availability of nat- ural rubber. The effects of non-zero minimum loading on fatigue crack growth were briefly mentioned in the first application of fracture mechanics to rubber fatigue. 14 Comprehensive studies were later reported by Lindley for fatigue crack growth in crystallizing 19 and in non-crystallizing 18 rubbers. An important difference between the crack growth behavior of crystallizing and non- crystallizing rubbers is that under a static load, non-crystallizing rubbers can exhibit steady crack growth, while crystallizing rubbers typically exhibit no crack growth. For non-crystallizing rub- bers, the crack growth rate under static load is a function of the energy release rate. Under cyclic load, the crack growth rate in non-crystallizing rubbers can be computed as the sum of steady and cyclic contributions to the crack growth rate. 18 Lindley’s study 19 characterized the fatigue crack growth rate of gum Natural Rubber as a function of the R-ratio. Lindley computed R based on the history of the energy release rate, ie R = T min /T max . Lindley’s study appears to be the first to use the R-ratio in rubber. Bathias et al. 20 have also studied R-ratio effects directly. The fatigue crack growth rate drops significantly for strain-crystallizing compounds under cycles which do not fully unload (R > 0). For example, a decrease of two orders of magnitude in fatigue crack growth rate was reported for R = 0.06 load- ing, relative to R = 0 loading. As noted earlier, when R is computed based on the energy release rate, it is always the case that R 0, since the energy release rate is a positive-definite quantity. Busfield et al. 21 studied the applicability of existing fracture mechanics approaches for pre- dicting the direction of crack growth in an inclined pure shear test specimen under mixed mode, R = 0 (mode I and mode II) loading. The ability to predict direction of growth with existing tech- niques appears to be limited. The growth direction was successfully predicted for R = 0 condi- tions. Under R > 0 conditions, the observed crack path deviates significantly from the path asso- ciated with the maximum energy release rate. This result has been attributed to anisotropy in the fatigue crack growth properties, induced by the action of finite strains on rubber’s molecular net- work. The tendency of the crack to deviate as described above also depends on polymer type, fillers, temperature, and strain state. More deviation is favored by polymers that exhibit strain crystallization, higher filler levels, lower temperature, 22 and strain states that favor crystallization (for example, simple tension is more favorable to strain crystallization than equibiaxial ten- sion). 23 The effects described above have been observed both in crack nucleation tests, and in crack

04-R-08 edit 6/3/04

4:27 PM

Page 7

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

397

growth tests. 5 A practical consequence is that tests in which the load does not fully relax (R > 0) tend to be more difficult to control, because of large sensitivity to the minimum and because of imperfect elasticity. Imperfect elasticity gives rise to small residual strains that affect the value of the minimum load.

C. STATICALLY STRAINED REST PERIOD

Roland 24 studied the effect of prolonged static load periods (“annealing”) on the strength and fatigue life of polyisoprene (a crystallizing rubber). He found that “annealing of an elastomer can result in significant changes in its failure properties, the nature of which depends on the deformation of the rubber during the annealing.” Annealing was found to have parallel effects on both the fatigue life and the ultimate tensile strength of the rubber. An annealing strain of approx- imately 15% resulted in minimum life and strength. Over a range of annealing strains from 1% to 400%, the ratio of maximum to minimum breaking strength is about 2, and the ratio of max- imum to minimum fatigue life is about 2.5. Fatigue life was evaluated under R = 0 displacement control to a maximum strain of 124%. Annealing periods in the range from 0 to 3 days were investigated. The effect of annealing strain level on fatigue life and strength are illustrated in Figure 3.

on fatigue life and strength are illustrated in Figure 3. F IG . 3. — Roland’s

FIG. 3. — Roland’s results 24 for unfilled polyisoprene, showing the effect of an annealing period (top = 24 hours, bottom = 72 hours) under a static strain on fatigue life and tensile strength.

04-R-08 edit 6/3/04

4:27 PM

Page 8

398

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

D. LOAD SEQUENCE

Sun et al. 25 studied the effects of applying step-up and step-down sequences of strains on residual ultimate tensile strengths in typical filled tire compounds. The purpose of the study was to evaluate the suitability of Miner’s linear damage rule for the development of accelerated tire durability tests. Miner’s rule predicts no dependence on loading sequence. The study found that “in all cases, a series of increasing strains… reduced the (residual ultimate) strength to a greater degree than the same strains applied in decreasing order.” This effect is attributed to a strain- dependent Paris-law exponent, and to the effect of stress-softening (the Mullins effect). Given that a constant Paris-law exponent is observed to hold over much of the fatigue crack growth curve (with deviations near the threshold and critical energy release rates), stress-softening may be the most relevant explanation of the sequence effect. Note that, in the case of displacement controlled tests, stress-softening would reduce the load, presumably reducing the damaging effect of subsequent cycles. Also, note that if stress-softening is the mechanism of the observed load sequence effect, then an opposite load sequence effect would be exhibited under load con- trol (i.e. a series of decreasing loads would reduce the residual ultimate strength to a greater degree than a series of increasing loads).

E. MULTIAXIALITY

While early researchers recognized the need for a better understanding of multiaxial effects on rubber fatigue, to date, the effects of multiaxial loading are not fully understood. Rubber com- ponents in service rarely operate under the simple strain states evaluated in fatigue studies using lab specimens. Although recent advances in simulation technology make it possible to predict realistic multiaxial loading histories, the ability to predict fatigue life from computed loading his- tories has been lacking. The need for progress in this area is great. The early work of Cadwell et al. 13 used a number of different test specimen types, in both axial and shear modes. That study recognized differences due to multiaxial conditions, but did not attempt to rationalize them quantitatively. The test specimen used for investigation of com- bined axial and shear loading is illustrated in Figure 4, as well as the results.

loading is illustrated in Figure 4, as well as the results. F IG . 4. —

FIG. 4. — Test specimen used by Cadwell et al. 13 to investigate combined effects of axial and shear loading at various levels of maximum and minimum strain, with results.

04-R-08 edit 6/3/04

4:27 PM

Page 9

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

399

Page 9 FA CTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER 399 F IG . 4

FIG. 4. Test specimen used by Cadwell et al. 13 to investigate combined effects of axial and shear loading at various levels of maximum and minimum strain, with results.

Gehman, Rohall, and Livingston 26 developed a test machine capable of subjecting rubber specimens to a complicated biaxial loading cycle; in order to measure the fatigue life. Their machine was developed specifically to simulate loading cycles typical of a rolling tire, for use in developing new rubber formulations. One drawback to this empirical approach is that the strain cycle is a function of the elastic properties, so the loading history experienced by one rubber for- mulation is different from the history experienced by another. While no other instances of this approach were found in the literature, it again shows that early researchers recognized a need to account for multiaxial effects. Roberts and Benzies 27 studied differences in crack nucleation life between cyclic simple ten- sion and equibiaxial tension loading. The relative severity of the two cases depends on what parameter is used to make the comparison. If the comparison is made based on strain energy den- sity, equibiaxial tension is less severe. If the comparison is made based on maximum principal strain, simple tension is less severe. Other researchers have confirmed these conclusions. 23 For comparisons of crack nucleation life between strain states other than simple or equibiaxial ten- sion, no existing literature has been identified. There have been several investigations of fatigue crack growth under multiaxial loading con- ditions. The focus of such studies differs somewhat from crack nucleation studies. In multiaxial crack nucleation studies, the objective is to ascertain the number of cycles to the appearance of a crack, its location, and possibly its orientation. In multiaxial crack growth studies, the objec- tive is to ascertain the growth rate and path of a given initial crack, or at least the growth rate and propagation direction of a given crack. Given the growth rate history, the number of cycles to a specified final crack length may be computed. The earliest study in which fracture mechanics was applied to studying fatigue crack growth in rubber 14 included two specimen types: the trousers specimen, and the planar tension (also known as the pure shear) specimen. Extension of the trouser specimen legs results nominally in mode III crack loading. It was found that crack growth rates from the mode I planar tension spec- imen matched quite accurately with crack growth rates from the trouser specimen, when plotted

04-R-08 edit 6/3/04

4:27 PM

Page 10

400

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

based on energy release rate. Roach 23 investigated crack growth in “strained pure shear” specimens. In these experiments,

a planar tension specimen was stretched along the long axis, then clamped in the stretched state.

Fatigue crack growth was then measured under the action of a cyclic strain applied transverse to the long axis. Roach showed that increasing the static strain increases the total strain energy den-

sity, but that crack growth is controlled only by the portion that is actually available to be released by crack growth. Aboutorabi et al. 28 and De and Gent 29 have recently developed a short, cylindrical test spec- imen for fatigue crack growth testing in rubber. The specimen is capable of sustaining combined axial-torsion loading. To date, only torsional (mode III) results have been published for that spec- imen. Radial-inward crack growth starts from a circumferential pre-cut. Crack growth in this specimen may not remain self-similar, in which case estimation of the energy release rate becomes complicated. 30 Stevenson et al. 31,32 studied fatigue crack growth under combinations of compression and shear loading. Using mode I fatigue crack growth tests run in simple tension, Stevenson report- ed reasonable predictions of the fatigue life of off-shore structural bearings under combined com- pression and shear loading. The predictions were based on analytical estimates of the individual energy release rates of small cracks in axial compression T c , and in shear T s , combined to obtain

a total energy release rate T according to the following rule, which was reported to be conservative.

T

=

(

TT

+

c

s

) 2

(4)

Busfield et al. 21 used energy release rates computed via finite element analysis to success- fully predict both the rate and direction of crack growth in an inclined pure shear test specimen, under mixed mode (mode I and mode II) R = 0 loading. Mixed mode crack growth rates for R = 0 loading were successfully predicted based on mode I R = 0 fatigue crack growth data. Under multiaxial loading, the crack path tends to grow in such a way that mode I loading is approached. This result is not unique to rubber. 33 Taken together, the apparent successes reported in these studies suggest that, if direction of propagation is properly accounted for, the fatigue crack growth characteristic curve of rubber may be regarded as independent of whether the loading is mode I, mode II, or mode III. A com- prehensive study confirming this conclusion was not found. Indeed, it should be noted that most of the cited studies were conducted with relatively thin lab specimens, and that a significant thickness effect is known to exist. 34-36 To the extent that the characteristic behavior is actually independent of the mode of loading, this result may be partially the consequence of the large strains at the crack tip. Whether the crack tip strain field arises due to opening or shearing motions of the crack faces, it may be the case that tensile strains always dominate at the crack tip, and that their average magnitude is uniquely related to the energy release rate. 37 Although the effects of multiaxial loading are not completely understood, it is clear that the following factors are important: 38 i) The relationship between applied loads and the loads expe- rienced locally by cracks. When applying a crack nucleation approach, it is important to consid- er loading on the failure plane. For a crack growth approach, the location and size of the initial crack are considered; ii) Whether the applied loads cause closure of critical cracks or material planes; iii) The dependence of material properties on the type and magnitude of imposed defor- mation. For example, strain crystallization can occur more readily under simple tension than under equibiaxial tension, because of differences in the degree of alignment of polymer chains induced by the deformation.

04-R-08 edit 6/3/04

4:27 PM

Page 11

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

401

F. LOADING FREQUENCY AND WAVEFORM

The effect of loading frequency depends on polymer type. For rubbers that exhibit strain crystallization, frequency is not observed to have much effect on fatigue life, under isothermal conditions, over a range of at least 10 -3 to 50 Hz 3 . For amorphous rubbers, frequency is observed to have a larger effect. The effect has been attributed to time-dependent steady crack growth, associated with viscoelasticity. 18,39,40 Lake and Lindley found that crack growth rates due to time-dependent (steady crack growth), and cycle-dependent (mechanical fatigue) contributions are additive. The time-dependent component of the crack growth rate depends on the mean value of the energy release rate. At sufficiently high strains and frequencies, an additional failure mechanism becomes important: thermal runaway. Thermal runaway occurs when the rate of internal energy dissipa- tion becomes greater than the rate of heat transfer to the surroundings. 8 When this occurs, the temperature increases rapidly, causing severe degradation of the rubber. Numerical simulations of this heating have been made. 41 Thick components exhibit a larger temperature rise than thin components. In fatigue testing, care must be taken to measure and compare actual material tem- peratures, since these may differ depending on material properties. Young 36,42 found that the fatigue crack growth rate can depend on strain rate (using a pulsed triangle waveform), particularly for non-strain-crystallizing elastomers, and for crystallizing elastomers at high energy release rates (close to fracture). Results for CIIR and NR compounds are shown in Figure 5. As a rule of thumb, it has been suggested that time-dependent crack growth for non-strain-crystallizing rubbers is most critical at frequencies below 0.2 Hz. 3

rubbers is most critical at frequencies below 0.2 Hz. 3 F IG . 5. — Y

FIG. 5. — Young’s results 42 showing the effect of strain rate (at equal cycle frequency) for unfilled CIIR (left) and NR (right) rubber compounds. T is the Strain Energy Release Rate.

Waveform shape has sometimes been found to influence fatigue behavior, particularly for amorphous polymers. A pulse waveform, considered more realistic to the strain history in a tire, has been compared to a sinusoidal waveform by several researchers. 36,42-44 The effect depends not only on polymer type, but also on other compound ingredients. Hardy et al. 43 found that, for amorphous bromobutyl (BIIR) rubber, certain filler types resulted in as much as an order of mag- nitude difference between crack growth rates under pulse and sine waveforms. Other filler types

04-R-08 edit 6/3/04

4:27 PM

Page 12

402

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

showed no dependence on waveform. For a given peak energy release rate, the pulse waveform was reported to result in higher crack growth rates. This result does not appear to be consistent with Lindley’s assertion 18 that steady and cyclic components of the crack growth rate are addi- tive. For the cyclic component, since the cyclic limits are the same, the cyclic contribution to the total crack growth rate must be the same for both pulse and sinusoidal loading. For the static component, the time average value of the R = 0, sinusoidal waveform is greater than the time average value of the R = 0, pulse waveform, and should therefore correspond to a higher steady crack growth rate, in conflict with the observed results. The resolution to this apparent conflict may lie in accounting for the effect of loading rate on the cyclic crack growth rate. Note that the loading rate in the pulse waveform is much greater than the loading rate in the sinusoidal wave- form.

III. ENVIRONMENTAL CONDITIONS

Environmental effects play a crucial role in the fatigue process, particularly at long life. Both the stress-strain and fatigue properties of rubber evolve in a way that is highly dependent upon the temperature of the material, and upon the presence and concentration of various chemical reactants in the surroundings of a rubber component.

A. TEMPERATURE

Elevated temperature has a deleterious effect on rubber, both on crack nucleation life, and on fatigue crack growth rate. The effect is greatest in amorphous rubbers. For gum SBR in dis- placement controlled tests, fatigue life drops by a factor of 10 4 as the temperature increases from 0 °C to 100 °C. 40 For gum NR in displacement controlled tests, fatigue life drops by a factor of only 4, for the same temperature range. The incorporation of fillers may reduce the temperature dependence slightly. 4 A parallel effect is observed in fatigue crack growth tests. 40,42,45 The aforementioned temperature effects occur independently of any chemical changes that may occur due to aging or continued vulcanization. Temperature has a large effect on the rate of these chemical processes, which can result in additional degradation of fatigue life at elevated temperatures, or over long periods. The extent to which temperature actually influences longer- term behavior is dependent on compound formulation: curatives, antioxidants, etc. These factors will be discussed subsequently.

B. OZONE

Exposure to ozone during a long-term fatigue test significantly increases crack growth rate and shortens life. 15,46-53 Due to the stress concentration, elastomer network chains at a crack tip are energetically favorable for reaction with ozone. Ozone reacts with carbon-carbon double bonds in the main polymer chain, causing scission of the chain. Crack growth due to ozone attack occurs whenever the instantaneous energy release rate exceeds a small threshold, often denoted G z . G z is typically much smaller than the mechanical fatigue threshold T o . The value of G z depends strongly on compound formulation, particularly the presence of antioxidants and antiozonants. For unprotected rubber, G z 0.1 J/m 2 . The pres- ence of antiozonants can increase G z by a factor of 10 or more. 4 For comparison, the mechani- cal fatigue threshold is approximately T o 50 J/m 2 . Ozone does not appear to affect the value of the mechanical fatigue threshold. Other chemical agents can attack rubber in a manner similar to ozone. 54 Gent and McGrath 55 studied the effect of temperature on ozone crack growth rate over a large range. Two mechanisms were found to control the crack growth rate da/dt. At temperatures

04-R-08 edit 6/3/04

4:27 PM

Page 13

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

403

near the glass transition T g , the crack growth rate is proportional to the temperature Θ, inde- pendent of ozone concentration. At sufficiently high temperatures (Θ-T g > 100°C), the crack growth rate depends totally on ozone concentration and is independent of temperature. The total rate of crack growth was given approximately by the following equations.

da

dt

=

1

() c

R

1

O

3

+

R

1

=

(

0 8

.

×

10

R

2

=

  0 12 × 10

.

12

3

1

R (Θ −

2

T )

g

cm

sec mg

L

)

c

O

3

1

cm

sec

  e

(

52

40 (

o C

Θ − T

)( +

)

g

Θ −

g )

T

(5)

(6)

(7)

It has been proposed that R 1 reflects the rate at which ozone molecules impact elastomer net- work chains at the crack tip. It is believed that R 2 is related to viscoelastic behavior, through the finite time required to redistribute crack tip strains, after scission of a network chain.

C. OXYGEN

Oxygen influences mechanical fatigue behavior in at least two ways. First, exposure to oxy- gen decreases the mechanical fatigue crack growth threshold from its value in vacuum. Second, oxygen dissolved or diffused in the rubber may induce chemical changes over time to the bulk elastomer network structure; this process is commonly called oxidative aging. Oxidative aging causes embrittlement and reduced resistance to fatigue crack growth. Even for new rubber specimens with no prior exposure, the presence of oxygen increases the fatigue crack growth rate, at constant energy release rate. The effect is conveniently described in terms of the energy release rate required to maintain a constant crack growth rate. At low ener- gy release rates, for unfilled natural rubber, the energy release rate in air is approximately half the energy release rate in vacuum, for the same fatigue crack growth rate. 4 The difference becomes less significant at higher energy release rates, but depends on test frequency. For lower frequencies, the difference persists to higher energy release rates. The magnitude of the effect depends on polymer type. Gent and Hindi 56 compared crack growth rates under both static and cyclic loadings, in air and in vacuum, for NR, SBR and BR. They observed only a factor of two change in fatigue crack growth rate for NR, due to the presence of oxygen. For SBR and BR, a factor of 8 difference between air and vacuum was observed. They demonstrated that the oxygen effect on fatigue crack growth rate is reversible, in an experiment in which oxygen atmosphere was applied, then removed, then applied again. In the experiment, the rate of crack growth was observed first to increase, then decrease, then increase again to the first level, in correspondence with the presence or absence of oxygen. Oxidative aging irreversibly changes the mechanical properties of the rubber, including the fatigue and fracture properties. In general, aging causes embrittlement of the rubber, faster fatigue crack growth, 57 reduced threshold and critical fracture energies, 58 and shorter fatigue life. Aging has been observed to affect both the slope and intercept of the power-law relationship for crack nucleation, 59 with the aged elastomer showing higher sensitivity to strain. Blackman and McCall 60 studied the effect of various periods of aging on the fatigue life of natural rubber. They

04-R-08 edit 6/3/04

4:27 PM

Page 14

404

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

note that absorption of oxygen in the amount of 1% of the specimen weight results in a factor of 2 drop in fatigue life. The aging effect can depend strongly on formulation and processing vari- ables, which are discussed below.

D. ELECTRICAL CHARGES

An unusual study from the former USSR suggests that the fatigue life of filled rubber com- pounds is sensitive to electric charges built up during cyclic deformation. Dogadkin and co- workers 61 ran identical fatigue tests in which the specimen grips were grounded and then insu- lated. The life for the grounded condition was reported to be longer by a factor that depended on the level of filler loading. They found that the larger the filler loading, the larger the effect. No explanation for the effect is offered and no follow-up studies were found.

IV. RUBBER FORMULATION

A unique aspect of engineering with rubber is that a wide range of mechanical properties can be achieved by varying the compound formulation and the manufacturing process. The main compounding factors that are known to influence fatigue behavior are polymer type, filler type and volume fraction, and the type and quantity of antioxidants, antiozonants and curatives. Manufacturing processes also influence fatigue behavior through their influence on factors such as ingredient dispersion, state of cure and the presence and nature of precursor sites for crack ini- tiation.

A. POLYMER TYPE

The polymer type impacts several aspects of the fatigue behavior. Whether or not the poly- mer exhibits strain crystallization is a primary consideration. 62 As discussed previously, strain crystallization has been shown to have beneficial consequences on fatigue performance at mod- erate or high strain levels, and appears to be associated with reduced sensitivity to environmen- tal effects. 56 In polymers where strain-crystallization is not present, continuous crack growth under a static load can be significant. Examples of polymers that exhibit strain crystallization include natural rubber (NR), isoprene rubber (IR) and polychloroprene (CR). Polymers exhibit- ing little or no crystallization include styrene-butadiene rubber (SBR), polybutadiene (BR), butyl rubber (IIR), acrylonitrile-butadiene rubber (NBR) and ethylene-propylene rubber (EPDM). A review of elastomeric polymer types and their applications is given by Hamed. 63 Due to the wide range of properties that are possible in engineered rubbers, it is difficult to give results representative of a given polymer type in specific applications, except for highly sim- plified or standardized formulations. Lake and Lindley’s results for fatigue crack growth behav- ior of 8 polymer types, 50 both filled and unfilled, are shown in Figure 6. Power-law fatigue crack growth constants due to Young, 36 for four filled elastomers and their blends are given in Table 1. From these data, it can be seen that rubbers that exhibit strain crystallization (NR and CR — compounds A and D of Figure 6) have a lower fatigue crack growth exponent F, and are superi- or at high energy release rates, relative to other elastomers. At low strains, CR is superior to NBR, EPDM and NR 3 . Figure 6 shows that BR (compound G) also exhibits superior perform- ance at low energy release rates. At high energy release rates, it can be seen that differences in elastomer type can account for as much as two orders of magnitude in crack growth rate. At low energy release rates, the differences appear to be smaller. Young’s results show that the power- law exponent on energy release rate can be controlled somewhat by blending of different poly- mer types.

04-R-08 edit 6/3/04

4:28 PM

Page 15

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

405

TABLE I REGIME 3 FATIGUE CRACK GROWTH PROPERTIES OF VARIOUS FILLED ELASTOMERS, da/dN = B T F , FROM REFERENCE 36

Elastomer

B, (m/cycle)/( J/ m 2 ) F

F

50/50 NR/BR Blend

1.38x10 -14

2.12

45/45/10 CIIR/NR/EPDM Blend

3.39x10 -14

1.91

60/35/5 CIIR/NR/EPDM Blend

0.95x10 -14

1.96

NR

4.46x10 -12

1.35

BR

12.0x10 -18

3.44

CIIR

0.85x10 -18

3.31

BIIR

0.12x10 -18

3.42

- 1 8 3.31 BIIR 0.12x10 - 1 8 3.42 F IG . 6. — R

FIG. 6. — R = 0 fatigue crack growth behavior for eight elastomer types. Results for unfilled, and two filler types are shown, from Lake and Lindley. 50 T is the Strain Energy Release Rate. Note that 1 kg/cm 981 J/m 2 .

04-R-08 edit 6/3/04

4:28 PM

Page 16

406

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

B. FILLER

The addition of carbon black to rubber compounds can have a pronounced strengthening effect, 64 depending on both the filler type and the volume fraction used. 65 At fixed volume frac- tion, low-structure blacks are apparently superior to high-structure blacks, and higher filler sur- face area is beneficial. 66 The effect of filler on fatigue properties has been attributed to multiple mechanisms. These include: i) pronounced changes induced by fillers on stiffness and hysteresis properties; ii) crack tip blunting, deviation and branching, induced by nonhomogeneity of the rubber-filler composite, at the crack tip; iii) agglomeration of filler particles, resulting in increased effective initial flaw sizes. When compared based on equal energy release rate, an opti- mal filler loading is observed that minimizes crack growth rate (maximizes fatigue life). 67 Below the optimal volume fraction, improved reinforcement and dissipation result in improved fatigue life with increasing volume fraction. Above the optimal volume fraction, increases in effective initial flaw size apparently offset other gains, resulting in overall reduced fatigue life. Based on the data of Auer et al., 67 the optimal filler loading for HAF black in SBR is between 30 and 60 parts carbon black, per hundred parts rubber, by weight. Lake and Lindley’s results for eight polymer types with three different filler systems 50 (none, 50 phr N900 (MT black), and 50 phr N300 (HAF black)) are shown in Figure 6. It can be seen that the addition of carbon black generally reduces crack growth rates relative to the gum elas- tomer, by an amount that is strongly dependent on the type of carbon black. Finer grades of car- bon black are seen to yield greater improvement, by as much as an order of magnitude. Corresponding effects (i.e. the filled elastomer is superior to the gum elastomer, finer grades of carbon black are superior to coarser grades) are observed on fatigue life. 65 Because of the large effect of volume fraction on compound stiffness, it is important to con- sider whether an application operates under displacement control, load control, or somewhere in- between. In displacement control, minimizing compound stiffness (by using less filler, for exam- ple) will minimize the energy release rate, resulting in maximum fatigue life. In load control, however, minimizing compound stiffness can increase the energy release rate, resulting in poor fatigue life. Effective comparisons of compound differences must therefore account not only for differences in the fatigue properties (at constant energy release rate), but also for differences in the operating energy release rate, due to differences in compound stiffness.

C. ANTIDEGRADANTS

Antidegradants can be added to rubber compounds to partially avoid the deleterious effects

of oxygen and ozone. In some cases, these agents can be almost as effective as operating in vac-

uum. 4 The various agents rely upon several mechanisms to prevent degrading chemical reac- tions. 63 Waxes in a rubber compound migrate to the surface, providing a physical barrier between the elastomer and degradants in the environment. Other antidegradants function by chemically

reacting directly with degradants, or with other species required to complete degrading reactions.

A full discussion of available antidegradants, their use and effects on material properties requires

a detailed understanding of polymer chemistry and is beyond the scope of this review. For the

interested reader, Scott’s 68 review should serve as a helpful starting point.

D. VULCANIZATION

Vulcanization is used in thermosetting elastomers to create covalent bonds, or crosslinks, between adjacent polymer chains. To a large degree, the crosslink density determines the physi- cal properties of rubber. 69 Higher crosslink density is associated with increased stiffness and reduced hysteresis. There is an optimum crosslink density, which results in the longest fatigue life. The peak in fatigue life appears to be due to the competing effects of strengthening due to

04-R-08 edit 6/3/04

4:28 PM

Page 17

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

407

an increased number of crosslinks, and the decreasing ability of the elastomer to dissipate ener- gy through hysteresis. 70 The type of crosslink formed during vulcanization has an effect on fatigue properties. 3,71,72 Polysulfidic crosslinks appear to be superior to monosulfidic or direct carbon-carbon (peroxide cure) crosslinks. The superiority of polysulfidic crosslinks is attributed to the ability of the crosslink to break before primary elastomer network chain carbon-carbon bonds, and subse- quently to reform. This mechanism leads to increased crack tip blunting and to increased dissi- pation due to hysteresis. Again, effective comparisons of compound differences must account not only for differences in fatigue properties, but also for differences in the operating energy release rate. Compound stiffness has a direct effect on energy release rate, depending on whether the application is sub- ject to load control or displacement control. Certain environmental factors have an effect on fatigue life, through their effect on state of vulcanization. Irradiation is an example. 73

V. CONSTITUTIVE BEHAVIOR

As originally conceived, Griffith’s hypothesis was applicable only to perfectly elastic mate- rials. 74 In that work, application of the theory was limited to linear elastic materials, and it was assumed that mechanical energy lost from the system was balanced by corresponding increases in the surface free energy of the fracture surfaces. Rivlin and Thomas 75 showed that Griffith’s approach applies to the prediction of monotonic fracture of unfilled elastomers, which exhibit highly nonlinear elastic constitutive behavior. In making this extension, it was recognized that the decrease in mechanical energy of the system was balanced, not only by increased surface free energy, but also by work done through dissipative processes occurring in material surrounding the crack tip. In the case of the unfilled elastomer, for example, slippage or severing of network chains is irreversible. Strain crystallization is another source of dissipation in some elastomers. Of practical importance, filled rubbers exhibit additional dissipative mechanical responses, not only at high strains (network chain breakage, strain crystallization), but also at small and moderate strains, particularly under cyclic loading. The precise effect of these dissipative processes on crack tip fields is not well understood because multiple mechanisms are involved, and because constitutive models capable of describing the various mechanisms are complex and have not been applied. It seems clear that dissipative processes near the crack tip result in crack tip blunting and in reduced crack tip stress/strain levels. Also, that dissipation in the bulk of a rubber specimen reduces the energy available to be released by crack growth. Experience sug- gests however, that despite the uncertainties, a fracture mechanics approach may be usefully applied to fatigue crack growth in filled rubbers. 76 Relationships between hysteresis and fatigue properties have been observed by many researchers. 77-83 Lake and Thomas 84 have suggested that if a material exhibits no hysteresis what- soever, only sudden fracture is possible. Further, hysteresis controls the sensitivity of the crack growth rate with respect to tear energy: the higher the hysteresis, the lower the slope (power-law exponent) of the energy release rate-crack growth relationship. In filled rubber, several mechanisms result in hysteresis. Each of these has its own depend- ence on strain level, strain rate and temperature. Hysteresis effects that are manifest primarily at high strain levels are especially beneficial. When hysteresis can be kept small at low strains, but high-strain hysteresis is maximized, the material retains its fracture toughness and fatigue resist- ance, without appreciable mechanical energy loss and the associated temperature rise in the bulk material. High strain hysteresis mechanisms include the Mullins effect and strain-crystallization. Low strain hysteresis mechanisms include filler-induced, rate independent hysteresis and vis- coelasticity of the gum elastomer. A brief description of each mechanism follows.

04-R-08 edit 6/3/04

4:28 PM

Page 18

408

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

A. MULLINS EFFECT

The Mullins effect 62,85,86 is an initial transient softening of the stress-strain curve in rubber. The first loading of the virgin rubber gives a much stiffer response than subsequent loadings. Subsequent loadings of equal or lesser magnitude rapidly tend towards a steady state, nonlinear elastic response. If the rubber is loaded beyond the previous maximum loading, an additional increment of softening will be observed, followed by the attainment of a new equilibrium stress- strain curve. The Mullins effect is exhibited primarily by filled rubbers. The improvement to fatigue properties of filled, non-strain-crystallizing elastomers relative to unfilled, non-strain crystallizing elastomers has been attributed to this dissipative mechanism. 3

B. STRAIN-CRYSTALLIZATION

Certain elastomers exhibit another high-strain hysteresis mechanism, crystallization due to an applied strain. 62-80 At large strains, the polymer network chains become sufficiently aligned such that a highly ordered state exists. 87 This causes a time-dependent phase change in the elas- tomer from an amorphous rubbery state to a crystalline state. When the strain is reduced suffi- ciently, the crystallized elastomer returns to the rubbery state. In contrast to Mullins effect, in which the hysteresis and stiffness permanently decrease after initial application of a load, the added stiffness and hysteresis due to strain crystallization do not diminish with additional cycles. While in the crystalline state, rubber exhibits increased resistance to crack growth. This causes a significant improvement of fatigue properties with increasing minimum strain, as discussed pre- viously. In rubbers which do not strain crystallize, steady crack growth under a static load is com- monly observed. In strain crystallizing rubbers, steady crack growth under a static load is usual- ly not observed. 13,16,17,40,88

C. RATE INDEPENDENT HYSTERESIS

In typical filled rubbers, at low and moderate strains, an approximately rate-independent hysteresis loop is often observed. The rate-independent behavior is not captured accurately by linear viscoelasticity. Triboelastic models have recently been used to model this behavior. 89,90 While this loss mechanism is still active at high strains, its magnitude is small compared to loss- es associated with Mullins effect and strain crystallization. The hysteresis due to this mechanism is dependent on the square of the strain amplitude, and decreases with increasing temperature, and increasing mean tensile strain. 91

D. VISCOELASTICITY

The vulcanized, unfilled elastomer exhibits time-dependent stress-strain behavior that in cer- tain situations can be reasonably modeled as purely viscoelastic. Viscous behavior is considered to arise from the time- and temperature- dependent mobility of molecular chains within the elas- tomer network. The amount of dissipation associated with viscous behavior of the unfilled elas- tomer, however, is usually negligible in comparison with the total inelastic behavior of the filled elastomer. 92 Still, steady crack growth under a static load, observed in amorphous rubbers, appears to be associated with viscoelasticity of the elastomer. This was clearly demonstrated by Mullins, 93 in experiments on unfilled vulcanizates of blends of butadiene-styrene and butadiene-acrylonitrile elastomers. He showed that the relationship between the energy release rate, the rate of steady tearing, and the temperature obeyed the same type of time-temperature dependence as other properties characteristic of viscoelastic materials. He also found a simple proportionality between damping measurements (loss modulus) and the energy release rate at a given crack

04-R-08 edit 6/3/04

4:28 PM

Page 19

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

409

growth rate, over a wide range of crack growth rates. Generally, it was shown that higher damp- ing is associated with higher required energy release rates to drive a crack at constant velocity.

VI. CONCLUSIONS

Many factors influence the fatigue behavior of rubber. Four general classes of factors have been discussed: those related to the mechanical load history, to environmental conditions, to for- mulation of the rubber compound, and to dissipative aspects of the stress-strain constitutive behavior. A primary consideration relating to the mechanical load history is that for fatigue failure to occur, a fluctuating load must be present. It is traditional, however, in the rubber literature, to present fatigue life and fatigue crack growth data against the maximum loading, which, by itself is ambiguous about the presence of a fluctuating load. Practical and theoretical reasons that this tradition has persisted were presented. Rubber’s fatigue behavior is extremely sensitive to both the maximum and minimum cyclic load limits. For R = 0 cycles, the relationship between max- imum load and fatigue crack growth rate (or fatigue life) exhibits a threshold, below which crack growth and fatigue failure do not occur. There is also a critical load above which unstable (time- dependent) fracture occurs. Between these extremes, rubber often exhibits a substantially power- law relationship between the load and the crack growth rate or fatigue life. Depending on poly- mer type and the presence of fillers, the effect of minimum or mean loading may be either ben- eficial or harmful. In strain crystallizing rubbers, increased minimum strain is beneficial. In non- strain-crystallizing rubbers, increased time-average strain can be harmful, due to steady crack growth. Indeed, in such rubbers, failure under a static load due to steady crack growth must be considered. Other aspects of the mechanical load history were also discussed, including the effects of static, loaded periods (“annealing”), load sequence, multiaxiality, frequency, and load- ing waveform.

Environmental factors can affect both the short and long term fatigue behavior of rubber. In the short term, elevated temperatures and the presence of oxygen or ozone are detrimental. The rate of crack growth may vary reversibly with these variables, in the short term. At long life, irre- versible changes to the elastomer due to elevated temperature and oxidative aging become impor- tant considerations. These changes include evolution of both the elastic and fatigue properties. Rubber is unique in that a great range of behavior is available by proper manipulation of for- mulation and processing variables. Proper selection of elastomer type, filler type and volume fraction, antidegradants, curatives, and vulcanization should ensure maximum fatigue life. The effects of these variables were discussed. The role of dissipative constitutive behavior in the improvement of fatigue properties of rub- ber was discussed. Four distinct dissipative mechanisms were identified, as well as their respec- tive effects on fatigue behavior. Generally, dissipation has been found to improve the fatigue properties of rubber. While a large number of factors have been studied individually, the ability to integrate such diverse results into accurate fatigue life predictions remains an ambitious and elusive goal. These results do provide a great deal of qualitative insight that can be used to advantage in designing durable components and systems. They also point out a number of areas in which future progress can be made.

1. The ability to predict fatigue behavior under complex mechanical loading histories. Multiaxial effects on crack nucleation life are not well understood. Multiaxial effects on crack growth are understood in principle, but have not been studied comprehensively. Robust tools for the analysis of crack growth are not widely available. The effects of variable amplitude loading have not been studied, to the authors’ knowledge. In partic- ular, there is a need to identify and apply a suitable cycle-counting algorithm capable of

04-R-08 edit 6/3/04

4:28 PM

Page 20

410

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

parsing complex loading histories into simpler histories that may be compared with fatigue data based on material testing.

2. The ability to integrate environmental factors into fatigue life predictions. Models for the effects of temperature, oxygen and ozone on fatigue life must be integrated with pre- dictions based on mechanical load history. As elastomer aging affects both fatigue prop- erties and elastic properties, quantitative models of the effects of aging on constitutive behavior must be developed.

3. Proper selection of elastomer type, filler type and volume fraction, antidegradants, cura- tives, and vulcanization. A thorough understanding of the relationship between materi- al properties and component performance is usually required to achieve maximum fatigue life.

4. The role of dissipative constitutive behavior in the improvement of fatigue properties of rubber is understood qualitatively, but its effect on crack tip fields is not understood quantitatively. Further investigation in this area may yield additional insights. A pre- requisite to progress on this front is the development and application of constitutive models capable of capturing the highly nonlinear, dissipative behavior typical of filled rubbers.

1

2

3

4

5

6

7

8

9

VII. REFERENCES

G. J. Lake, RUBBER CHEM. TECHNOL. 68, 435 (1995).

A. G. Thomas, RUBBER CHEM. TECHNOL. 67, G50 (1994).

M. D. Ellul, “Engineering with Rubber, How to Design Rubber Components”, A. Gent, Ed., Carl Hanser Verlag, Munich (1992), ch. 6.

G. J. Lake, A. G. Thomas, “Engineering with Rubber, How to Design Rubber Components”, A. Gent, Ed., Carl Hanser Ve rlag, Munich (1992), ch. 5.

G. J. Lake, Prog. Rubber Technol. 45, 89 (1983).

A. N. Gent, “Science and Technology of Rubber,” F. R. Eirich, Ed., Academic Press, (1978) p.419.

G. J. Lake, RUBBER CHEM. TECHNOL. 45, 309 (1972).

J. R. Beatty, RUBBER CHEM. TECHNOL. 37, 1341 (1964).

W. V. Mars, A. Fatemi, Int. J. Fatigue (2002), in press.

10 S. D. Downing, D. F. Socie, Int. J. Fatigue 4, 31 (1982).

11 D. Klenke, A. Beste, Kautsch. Gummi Kunstst. 40, 1067 (1987).

12 A. Wöhler, Engineering 2, 160 (1867).

13 S. M. Cadwell, R. A. Merrill, C. M. Sloman, F. L. Yost, Ind. Eng. Chem., Anal. Ed., 12, 19 (1940).

14 A. G. Thomas, J. Polym. Sci. 31, 467 (1958).

15 G. J. Lake, P. B. Lindley, J. Appl. Polym. Sci. 10 , 343 (1966).

16 A. N. Gent, P. B. Lindley, A. G. Thomas, J. Appl. Polym. Sci. 8 , 455 (1964).

17 J. H. Fielding, Ind. Eng. Chem. 35, 1259 (1943).

18 P. B. Lindley, R UBBER C HEM . T ECHNOL . 47 , 1253 (1974).

19 P. B. Lindley, Int. J. Fracture 9 , 449 (1973).

20 C. Bathias, K. LeGorju, C. Lu, L. Menabeuf, “Fatigue and Fracture Mechanics: 27 th Volume,” ASTM STP 1296,

R. S. Piascik, J.C. Newman, N. E. Dowling, Eds., (1997) p. 505.

21 J. J. C. Busfield, C. H. H. Ratsimba, and A. G. Thomas, “Finite Element Analysis of Elastomers,” D. Boast and

V. A. Coveny, Eds., IMechE, United Kingdom, (1999) p. 235.

22 H. J. Kim, G. R. Hamed, RUBBER CHEM. TECHNOL. 73, 743 (2000).

23 J. F. Roach, Ph.D. Dissertation, University of Akron (1982).

04-R-08 edit 6/3/04

4:28 PM

Page 21

FACTORS THAT AFFECT THE FATIGUE LIFE OF RUBBER

411

24 C. M. Roland, J. W. Sobiesky, RUBBER C HEM. T ECHNOL. 62, 683 (1989). 25 C. Sun, A. N. Gent, P. Marteny, Ti re Sci. Technol. 28 , 196 (2000). 26 S. D. Gehman, P. Rohall, D. I. Livingston, R UBBER C HEM . T ECHNOL . 34 , 506 (1961). 27 B. J. Roberts, J. B. Benzies, Proc. of Rubbercon ’77 2.1, 2.1 (1977). 28 H. Aboutorabi, T. G. Ebbott, A. N. Gent, O. H. Yeoh, RUBBER CHEM. TECHNOL. 71, 76 (1998). 29 D. K. De, A. N. Gent, RUBBER CHEM. TECHNOL. 71, 84 (1998). 30 T. S. Fleischman, V. Kerchman, T. G. Ebbott, Tire Sci. Technol. 29, 91 (2001). 31 A. Stevenson, J. R. Hawkes, J. A. Harris, P. Hansen, SAE Paper No. 983210 (1998). 32 A. Stevenson, RUBBER CHEM. TECHNOL. 59, 208 (1986). 33 Fracture Mechanics – Fundamentals and Applications,” T. L. Anderson, CRC Press, (1995). 34 A. Kadir, A. G. Thomas, RUBBER CHEM. TECHNOL. 54, 15 (1981). 35 K. A. Mazich, K. N. Morman, F. G. Oblinger, T. Y. Pan, P. C. Killgoar, R UBBER C HEM . T ECHNOL . 62 , 850 (1989). 36 D. G. Young, RUBBER CHEM. TECHNOL. 58, 785 (1985). 37 A. G. Thomas, J. Polym. Sci. 18, 177 (1955). 38 W. V. Mars, RUBBER C HEM. T ECHNOL. 75, 1 (2002). 39 H. W. Greensmith, J. Polym. Sci. 8, 1113 (1964). 40 G. J. Lake, P. B. Lindley, J Appl. Polym. Sci. 8 , 455 (1964). 41 A. Becker, V. Dorsch, M. Kaliske, H. Rothert, Tire Sci. Technol. 26, 132 (1998). 42 D. G. Young, RUBBER CHEM. TECHNOL. 59, 809 (1986). 43 D. Hardy, H. Moneypenny, M. Holderied, J. Harris, R. Campion, G. Morgan, Paper No. 2 presented at RubberChem’99, Belgium, (1999). 44 A. J. M. Sumner, S. A. Kelbch, U. G. Eisele, Rubber World, 212 (2), 38 (1995). 45 D. G. Young, J. A. Danik, RUBBER CHEM. TECHNOL. 67, 137 (1994). 46 M. Braden, A. N. Gent, J. of Appl. Polym. Sci. 3 (7), 100 (1960). 47 M. Braden, A. N. Gent, J. Appl. Polym. Sci. 3 (7), 90 (1960). 48 M. Braden, A. N. Gent, J. Appl. Polym. Sci. 6, 449 (1962). 49 G. J. Lake, P. B. Lindley, Rubber J. 146 (10) , 24 (1964). 50 G. J. Lake, P. B. Lindley, Rubber J. 146 (11) , 30 (1964). 51 G. J. Lake, Rubber Age, 30 (1972). 52 G. J. Lake and P. B. Lindley, J. Appl. Polym. Sci. 9 , 2031 (1965). 53 A. N. Gent, H. Hirakawa, J. Polym. Sci.: Part A-2 5, 157 (1967). 54 A. N. Gent, H. Hirakawa, J. Polym. Sci.: Part A-2 6, 1481 (1968). 55 A. N. Gent, J. E. McGrath, J. Polym. Sci., Part A 3, 1473 (1965). 56 A. N. Gent, M. Hindi, RUBBER CHEM. TECHNOL. 63, 123 (1990). 57 S. G. Kim, S. H. Lee, RUBBER CHEM. TECHNOL. 67, 649 (1994). 58 A. S. Deuri, A. K. Bhowmick, J. Mater. Sci. 22, 4299 (1987). 59 M. K. Khromov, G. A. Niazashvili, N. L. Sakhnovskii, T. A. Koroleva, Int. Polym. Sci. Tech. 23 (5), T/82 (1996). 60 E. J. Blackman, E. B. McCall, RUBBER CHEM. TECHNOL. 43, 651 (1970). 61 B. A. Dogadkin, V. E. Gul, N. A. Morozova, RUBBER CHEM. TECHNOL. 33, (1960). 62 C. M. Roland, RUBBER CHEM. TECHNOL. 62, 863 (1989). 63 “Engineering with Rubber, How to Design Rubber Components”, A. Gent, Ed., Carl Hanser Verlag, Munich (1992), ch. 2. 64 A. I. Medalia, RUBBER CHEM. TECHNOL. 60, 45 (1987). 65 E. S. Dizon, A. E. Hicks, V. E. Chirico, RUBBER CHEM. TECHNOL. 47, 231 (1974). 66 J. Zhao, G. Ghebremeskel, Rubber Plast. News November 27, 2000, p. 14.

04-R-08 edit 6/3/04

4:28 PM

Page 22

412

RUBBER CHEMISTRY AND TECHNOLOGY

VOL. 76

67 E. E. Auer, K. W. Doak, I. J. Schaffner, RUBBER C HEM. T ECHNOL. 31, 185 (1958). 68 G. Scott, RUBBER CHEM. TECHNOL. 58, 269 (1985). 69 G. R. Hamed, RUBBER CHEM. TECHNOL. 56, 244 (1983). 70 A. Y. Coran, “Science and Technology of Rubber,” F. R. Eirich, Ed., Academic Press, (1978), p. 291. 71 G. J. Lake, P. B. Lindley, J. Appl. Polym. Sci. 9 , 1233 (1965). 72 H. Chun, A. N. Gent, RUBBER CHEM. TECHNOL. 69, 577 (1996). 73 A. N. Gent, J. Appl. Polym. Sci. 6, 497 (1962). 74 A. A. Griffith, Phil. Trans. Roy. Soc. London, Series A 221, 163 (1920). 75 R. S. Rivlin, A. G. Thomas, J. Polym. Sci. 10, 291 (1953). 76 J. J. C. Busfield, A. G. Thomas, M. F. Ngah, “Constitutive Models for Rubber,” A. Dorfmann, A. Muhr, Ed., A. A. Balkema (1999) p. 249, 77 J. A. C. Harwood, A. R. Payne, J. Appl. Polym. Sci. 12, 889 (1968). 78 G. R. Hamed, RUBBER CHEM. TECHNOL. 64, 493 (1991). 79 E. H. Andrews, J. Mech. Phys. Solids 11, 231 (1963). 80 I. S. Choi, C. M. Roland, RUBBER CHEM. TECHNOL. 70, 202 (1997). 81 A. R. Payne, R. E. Whittaker, RUBBER CHEM. TECHNOL. 45, 1043 (1972). 82 M. L. Williams, Int. J. Frac. Mech. 1, 292 (1965). 83 V. E. Panin, Theor. Appl. Frac. Mech. 30, 13 (1998). 84 G. J. Lake, A. G. Thomas, Proc. Royal Society of London, Series A. 300, 108 (1967). 85 L. Mullins, RUBBER CHEM. TECHNOL. 42, 339 (1969). 86 C. M. Roland, RUBBER CHEM. TECHNOL. 62, 880 (1989). 87 L. R. G. Treloar, “Physics of Rubber Elasticity,” Oxford University Press, London (1958). 88 Handbook of Molded and Extruded Rubber,” 3 rd Edition, Goodyear Tire and Rubber Company (1969). 89 J. R. Luchini, Ph.D. Dissertation, University of Michigan (1977). 90 V. A. Coveney, S. Jamil, D. E. Johnson, M. A. Keavey, C. Menger, H. T. Williams, “Finite Element Analysis of Elastomers,” D. Boast and V. A. Coveny, Eds., IMechE, United Kingdom (1999). 91 J. R. Luchini, R. H. Arthur, J. M. Peters, Tire Sci. Technol. 22, 206 (1994). 92 A. N. Gent, K. W. Scott, “Engineering with Rubber, How to Design Rubber Components,” A. Gent, Ed., Carl Hanser Ve rlag, Munich (1992) ch. 4. 93 L. Mullins, Trans. I. R. I. 35, 213 (1959).

[ Paper 6, presented at the Spring, ACS, Rubber Division Meeting (Savannah, Georgia), April 29-May 1, 2002 ]