Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 4, 39, 2001 FATIGUE CRACK GROWTH PECULIARITIES AND MODIFICATIONS OF THE WHEELER RETARDATION MODEL (PART 1) Sylwester Kłysz Materials Engineering Division, Air Force Institute of Technology, Warsaw e-mail: sklysz@itwl.waw.pl What has been given consideration in the paper, are peculiarities of fati- gue crackgrowth, ones referredto curvesof a crack lengthvs. thenumber of cycles or the rateof propagationvs. the rangeof stress intensity factor, i.e. the curves typical of tests with overloads applied in cycles – for the PA7 aluminiumalloy and 18G2Asteel.Difficulties with themodel-based description (based on the Wheeler retardation model) of fatigue crack growth have been shown as due to the peculiarities discussed. Key words: crack growth, overloads, retardation,Wheeler model 1. Introduction At present, it is generally accepted that fatigue tests under a constant amplitude or program loading insufficiently represent realistic service load histories. With respect to fatigue crack growth under a variable amplitude loading, the term load interaction effects (also referred to as load sequence or load history effects) is used to label the phenomenon that the crack growth increment in a given cycle can be different than that in a constant amplitude test under the same maximum and minimum stress intensity. Random load tests and a statistically equivalent program load test do not give the same re- sults (Jacoby, 1970; Schijve, 1973).Moreover, the effect of alternative designs, materials, manufacturing techniques, etc. on the fatigue life under a variable amplitude loading can only be judged correctly from variable amplitude tests. In some cases, constant amplitude tests can even give qualitatively wrong an- swers with respect to the fatigue crack growth resistance (Bucci et al., 1980; Schulte et al., 1984). This, together with development in data procesing me- thods and equipment capabilities (computer controled servohydraulic testing 826 S.Kłysz machines), has caused the fatigue testing under a realistic, random loading to become widely appreciated nowadays. Because a deeper interest in the effects of load interaction on fatigue crack growth was initiated by aircraft manufacturers, aircraft materials were ory- ginally of primary importance in fatigue tests under a variable amplitude loading. Reviews of these early experimental results obtained under simple variable amplitude sequences and also under a program and random loading were prepared by Schijve (1972, 1985). Later on, the same autor presented an ample survey of trends observed in the analysis fatigue crack growth under flight-simulation loadings (Schijve, 1985). Following aircraft, other industrial branches, e.g. pressure vessels, nuclear reactors, bridges and offshore struc- tures have also placed more reliance on damage tolerance designs. Though introducing the latter philosophy in awider scale has prompted investigations on load interaction effects in a variety of metals, the coresponding data bank still remains by far more meagre than in the case of aircraft materials. Imposition of a single overload duringbaseline cycling causes a retardation of the crack growth effect. The severity of this effect is usuallymeasuredby the normalized number of retardation (delay) cycles, ND/NCA (Skorupa, 1996). Here, ND (reffered to as the delay distance) is the number of cycles between the overload application at the crack length aov and the recommencement of the steady state fatigue crack growth rate after the overload-affected crack growth increment ∆aov. The NCA is the number of baseline cycles that would have elapsed over ∆aov in the absence of the overload. The fatigue crack growth rate da/dN response to a single overload can vary depending on the load parameters. In general, higher overload ratios kov = Fov/Fmax give larger retardation in the crack growth. Increasing the overload ratio results in an increase in ND/NCA and ∆aov and yields a lower minimum da/dN value (Skorupa, 1996; Blom, 1989; Schin and Fleck, 1987). For high overloads (overload ratio kov > 2.5), the initial acceleration was absent and immediate retardation was observed in a Ti-alloy (Ward-Close et al., 1989) and in mild steel (Damri and Knott, 1991). The retardation decreases if the stress ratio R is increased (Damri andKnott, 1991). For a stainless steel the fatigue crack growth increment between the overload application and the occurrence of the minimum da/dN to decrease and the initial acceleration to vanish as R increased from 0.1 to 0.6 until, at R = 0.65, retardation became immediate (Skorupa, 1996; Shin andHsu, 1993). The ratio between ∆aov and the size of the plastic zonemeasured on the surface of specimens decreased for increasing R from values larger than 1 (3.5 atR=0) to values ranging between 0.2 and 0.3 for R > 0.45. The influence of the stress intensity factor ranges at the Fatigue crack growth peculiarities... – part 1 827 baseline level, ∆K is complex since the ND vs. ∆K and ∆aov vs. ∆K plots are typically U-shaped curves, thus indicating the least retardation effect at some intermediate ∆K value (Skorupa, 1996; Ward-Close et al., 1989). For certain combination of loading parameters, crackmay retard by several orders ofmagnitude or even arrest though ∆K is far above threshold (Skorupa, 1996; Bernard et al., 1977; Bertel et al., 1983). For an equivalent reduction in K values, a multiple overload gives rise to a more severe fatigue crack growth retardation during subsequent baseline cycles than a single overload (Ward-Close et al., 1989). The retardation at the step-down in load occurs more rapidly than for a single overload, and even immediately if during an overload block the stationary fatigue crack growth rate value is attained (Skorupa, 1996; Ward-Close et al., 1989; Sehitoglu and McDiamid, 1980). The retardation effect is amplified by elevating the overload ratio value, similarily to a single overload case (Chang et al., 1981; Chehini et al., 1984). Increasing the number of overload cycles Nov gives rise to a more severe retardation effect through extending the ND period, while the ∆aov distance remains the same as for a single overload (Skorupa, 1996; Chehini et al., 1984; Chen and Roberts, 1985). An intermittent overload can produce either retardation or acceleration in fatigue crack growth rates depending on the combination of load andmaterial parameters. With a relatively frequently applied single intermittent overload (Nov =1) crack growth acceleration has been observed by several authors. In stress controlled tests the acceleration phenomenon appeared to bemore pro- nounced in the low ∆K region and todiminishor evenpass into retardation at higher ∆K values. The longer the interval DN between single overloads, the more severe retardation is produced, as exemplified by Skorupa (1996), Zhang et al. (1987). For a sufficiently large DN period, each overload application is visualized by a discontinuity in the corresponding a vs. N (crack length vs. number of cycles) plot. For a multiple intermittent overload (DN > 1) the authors observed the retardation to increase by Nov when DN was kept con- stant. Further extending of Nov, however, resulted in the acceleration effect. Like in the case of a single intermittent overload, the retardation effect for a multiple intermittent overload sequence was enhanced when the DN interval was increased (Skorupa, 1996; Zhang et al., 1987). For engineering practice the difficulties of crack growth propagation ana- lysis are related, among other things, with choice of adequate calculating mo- del, well fitted to geometrical, loading andmaterial requirements of analysed problems. A model-based (theoretical) description of fatigue crack growth is usually a problem of great complexity for the following reasons, at least: 828 S.Kłysz • it calls for a properly determined stress intensity factor adequate to the way of applying loads to a component under tests, geometry of the component with a propagating crack, occurrence of internal stresses, effects of environment, e.g. of a corrodingmedium, upon the component, etc. • it requires verifying whether coefficients of equations of propagation re- main constant for various modes of loading (loading mode I, II, III; the random nature of the applied load), or how the coefficients change with changes in stress ratios, mean loading levels, frequency, temperature • it needs taking into account thephenomenasuchas effects of retardation, creeping, etc., i.e. ones that accompany the fatigue cracks growth. Moreover, a gooddescription of a selected case of fatigue crack propagation does not ensure similar description for other one – in another condition of te- sting and type of thematerial. Investigatorsmustkeep this restriction inmind. Existence of some osculate elements should be worthy of notice in the aspect of developing the modelling in fracture mechanics. Such elements are in fact mechanisms contributing to crack propagation at the microstructural level. Unfortunately, testing such phenomena is difficult, laborious, time-consuming and expensive. Theoretically, the crack growth rate retardationmodels can be a consequence in this area. The purpose of this paper is to present peculiarities of fatigue crack pro- pagation for variousmaterials and various load conditions (part 1).Moreover, modifications of the Wheeler crack growth retardation model are proposed to cover more possibilities of modelling the crack propagation (part 2). The extension presented have deals with (among other things) the peculiarities mentioned above. In this paper results of fatigue tests of 18G2A steel and the aluminiun alloyPA7are presented.The experimentswere performedunder cyclic loading with overloads. The results obtained during the examinations and presented in fatigue curves a= f(N) and da/dN = f(∆K) are analysed. The genesis of changes in theWheelermodel and features of themodifiedmodel describing the fatigue crack propagation in a wide range are illustrated. 2. Examination of PA7 duralumin The Wheeler model (Wheeler, 1972; Fuchs and Stephens, 1980; Kocańda and Szala, 1985) is a pretty good tool to describe fatigue crack propagation in Fatigue crack growth peculiarities... – part 1 829 materials forwhichachange in thecrack length (or crackopeningdisplacement –COD),after the loadhasbeenapplied, against anumberof loadcycles, shows the nature as shown in Fig.1: Fig.1a – initially, the rate of crack growth (or increasing of COD) ismore or less intensively retarded in comparison to the crack growth rate prior to the application of the load, then systematic (monotonic) growth can be observed up to reaching the crack growth rate equal to that from before the load application, or Fig.1b – intensive retardation of the crack length increase (no change of COD) for a long time after applying the load (or up to applying a sub- sequent load, at least). Both diagrams have resulted from fatigue tests on crack propagation. The tests were carried out using compact-tension (CT) specimens, made of the PA7duralumin.Theconstant-amplitude loadingwith overloads (featuredwith the constant overload factor kov = Fov/Fmax) applied in cycles, i.e. every DN number of cycles, determined the testing conditions. An initial fatigue precrack, approximately 2mm long as fromthebottomof the notch,wasmade due to 30-40 thousands cycles of the constant amplitude Fmin −Fi (where Fi –maximum load applied at the stage of the precrack initiation) and stress ratio R=0.3. The numbers of the applied load cycles and values of the crack openingdisplacement (converted then into the length of the propagating crack – with the compliance method) were recorded. TheWheelermodel describes fatigue crack propagation phenomenaby the relationship da dN =CpC(∆K) m (2.1) and defines the retardation coefficient Cp in the following way Cp = ( rp,i aov+rp,ov−ai )n (2.2) where rp,i,rp,ov – radii of plastic zones of current and overload cycles, re- spectively ai,aov – crack lengths of the current and overload cycles, respecti- vely n – exponent in theWheeler model. In the first case (Fig.1a), corresponding to a relatively low value of the exponent n (if description made with the Wheeler model), the propagating 830 S.Kłysz Fig. 1. Experimental records of crack opening displacement in CT specimenmade of PA7 duralumin under fatigue tests with overloads applied in cycles Fatigue crack growth peculiarities... – part 1 831 crack overcomes – partially or completely – the overload plastic zone, and completely escapes from the phase of retardation due to the overload – before the subsequent overload is applied. In the second instance (Fig.1b), corresponding to a relatively high value of the exponent n (the model-based description), the propagating crack does not surmount the overload plastic zone. It continues propagating under severe retardation conditions due to the overload, up to applying the subsequent overload. Considerable increments are gained in the course of applying the overload, rather than between the overloads. 3. Tests of the 18G2A steel Fig. 2. Crack length a vs. number of cycles N for CT specimens made of 18G2A steel under fatigue tests with overloads applied in cycles For comparison, Fig.2. shows some results of tests aimed at investigating fatigue crack propagation in the 18G2A steel and, Fig.3, a sample of their theoretical description.The tests were conducted usingCT specimens, aswell, under a cyclic constant-amplitude loading, without overloads at all, and with ones applied in cycles. The testing conditions and numbers of cycles up to the failures of the selected test pieces have been presented in Table 1. 832 S.Kłysz Fig. 3. Theoretical description of crack propagation in CT specimens made of 18G2A steel Table 1. Testing conditions and results of tests with specimens made of 18G2A steel Specimen Fmin−Fmax Fov DN 2Nf No. [kN] [kN] [cycle] [cycle] 1 2.02-6.72 – – 80500 2 2.02-6.72 8.40 20000 69000 3 2.02-6.72 8.40 10000 77500 4 2.02-6.72 10.08 10000 120000 The similarity between the plots of a = f(N) relationships (i.e. crack length against a number of cycles) as well as between the theoretical descrip- tions of specimens 1, 2, and 3 are peculiar to the instances of fatigue crack propagation under loads of a constant amplitude with small overloads applied (up to approx. kov =1.2, as in the cases under discussion) or with overloads applied within long time intervals (e.g. longer than 20 per cent of the fatigue life typical of a given range of the loads Fmax−Fmin), although it is not the rule and depends on thematerial tested. These are the cases of rather mono- tone crack propagation or of propagation when the effects of retardation fade away before a subsequent overload is applied (analogous to the case illustrated in Fig.1a). In the cases of application of larger overloads (e.g. kov > 1.4) (like in specimen No.4), the relationships a= f(N) do not show a systematic (or fully monotonic) increase. Characteristic are the step-by-step increments in Fatigue crack growth peculiarities... – part 1 833 the crack lengths at every moment of overload occurrences, whereas between the overloads the crack growth is explicitly retarded (analogous to the case illustrated in Fig.1b). The case shown in Fig.3 presents a result of theoretical description of crack propagation, with retardation models incorporated, (for various values of the retardation exponent n) which reveals the averaging of the a = f(N) relationship – arriving at the minimum of the sum of squ- ares of deviations in experimental and computational values for values of the exponent n approaching zero or unity, rather than of matching it up to the real (retarded) crack growth. The asteriskmarks the instance of theminimum of the sum of squares of deviations in the experimental and computational values. 4. Analysis of results and their theoretical description Thematchingmade at values of the exponent n approaching zero or unity means in practice nothing but curvilinear regression of a low order (with no point of inflexion); therefore, one cannot recognised thematching that correc- tly reflects the experimentally determined relationship. There is no chance to simultaneously provide a severe retardation in the crack propagation betwe- en the overloads and considerable crack growth for the overload cycle (those step-by-step increments shown inFig.1 andFig.2) using the classicalWheeler model (because of its mathematical and physical fundamentals). Tomake the increment for one overload cycle be one or two orders of magnitude higher than that reached in the course of several thousands of cycles between the overloads, the exponent n must assume a high value. It would make the Cp coefficient reach some low value (which, in turn, would result in some con- siderable retardation in the crack increasing between the overloads); the C and m coefficients should be at the same time high enough to ensure this substantial increment in the crack length for a single overload cycle (when Cp = 1). However, the cracks within the tested specimens with no overloads (or at least within the range of up to the first overload, or within all adequate growth periods when Cp = 1) would then show the crack growth rate 10 up to 100 times as high as in the case of cracks within the overloaded specimens (at the utmost, the specimenswith no overloads would show fatigue life many times shorter). In practice, it does not take place on such a large scale. Numerical analysis of the already obtained results has been carried out to determine the parameters of equations that describe propagation rate of fatigue cracks within the tested material. Mathematical models described by 834 S.Kłysz Kłysz (1991) have been used, as well, with the Paris equation da/dN = CpC(∆K) m employed. Owing to the fact that the initial fatigue precracks within the tested specimens were started up with applying the load higher than the base load of the test (Fi = 9kN instead of Fmax = 6.72kN), the model-based description takes also into account the effect of retardation of the crack growth due to a change (decrease) in the level of the loads applied between the stage of precracking and the exact propagation test (in particular, for the specimen tested with no overloads applied in cycles). Values of the coefficients C and m of the Paris equation have been determined (for the pre-set value of n), onesmatchedwith the least squaresmethodup to the a= f(N) relationships found experimentally for individual specimens. Initially, assuming values of the exponent n of theWheelermodel to be expressedwith integers, the following values of the coefficients of interest have been arrived at – see Table 2. The Sn column includes values of the sum of squares of deviations of the least squares method, ones suitable for individual matches. Fig.3 show the essence of themodel-based (according to theWheeler mo- del) description of crack propagationwith the effect of retardation due to over- loads taken into account (starting up with the extreme instance for n = 0, when no retardation, Cp = 1 is considered). What has become evident from Table 2 and the figures, is that both the values of the coefficients C and m as well as the approximating curves show pretty wide divergence. The visible mismatch for some values of the exponent n proves that there is an optimum value of the exponent, forwhich the sum Sn reaches theminimum, and values of C and m adequate to the values of the exponent n forwhich themismatch has been observed give the best ever possible –with a given retardationmodel assumed – theoretical description of the experimental data. It is also evident that the optimum values of the coefficients n,C,m are different for particular specimens (testing conditions). However, this is not an optimistic statement, because such a model-based description of the fatigue crack propagation and the effect of retardation in the crack growth (for a givenmaterial) would not be uniform. It would depend on loading conditions. What is observed for specimens No. 1, 2, and 3 are some monotonic changes in values of the coefficients C and m depending on the exponent n of the retardation model, which is not the case with specimen No.4. Besides, the changes in values of these coefficients for the specimen No.4 are different. Fatigue crack growth peculiarities... – part 1 835 Table 2. Coefficients in the Paris equation describing experimental data for CT specimens made of 18G2A steel, for various values of the exponent n of theWheeler retardation model Specimen n C m Sn No. (·10−13) 1 0 0.271 4.59 67.90 1 0.345 4.53 62.80 2 0.579 4.39 56.31 3 1.096 4.21 47.80 4 3.071 3.93 37.82 5 13.033 3.53 28.53 6 82.802 3.03 24.51 7 947.062 2.35 31.50 2 0 0.598 4.39 29.65 1 1.061 4.24 27.06 2 1.992 4.07 23.56 3 5.198 3.81 18.95 4 16.504 3.50 14.00 5 50.931 3.21 10.91 6 352.378 2.69 18.63 3 0 3.965 3.81 15.68 1 7.539 3.65 15.28 2 17.378 3.43 15.76 3 48.398 3.16 19.66 4 196.043 2.79 26.50 5 534.804 2.54 34.82 4 0 112.992 2.64 11.30 1 154.190 2.67 10.06 2 129.298 2.86 13.63 3 169.516 2.93 28.65 4 40.695 3.51 60.24 5 0.086 5.48 86.33 Another peculiarity of the fatigue crack propagation has been shown in subsequent figures, which represent experimentally determined curves of pro- pagation, ones gained using the SEN (single-edge notch) specimens made of the 18G2A steel, also tested under a load of constant amplitude (stress ratio R = 0.3), with overloads applied in cycles, i.e. every DN number of cycles, on the levels of kov = 1.2, 1.6, 1.75 (Fig.4a to Fig.4c, respectively). The 836 S.Kłysz Fig. 4. a= f(N) relationship for specimen tested with overload k ov =1.75, 1.6, 1.2 Fatigue crack growth peculiarities... – part 1 837 plots confirm the generally known properties of fatigue cracks propagation under overloading conditions: both single overloads and those applied in cyc- les increase the fatigue life as compared to the life gained with no overloads imposed at all. In general, the fatigue life increases with increase in the level of the applied overload (at least in the range of the discussed levels of the overload). There is, however, an optimum value of the time interval between the overloads, at which the life reaches the maximum (i.e. overloads applied too often or too rarely, as compared to this optimum time interval, do not give such strong effect of retardation of the crack growth). However, there are distinct differencesbetweenparticular plots of the crack propagation: dynamics of crack length growth, convexity of the curves over the sections between the application of the overloads are completely different in subsequent figures (comparing to the previous figures). For the loads on the level of kov = 1.2 the increment of the crack length are almost monotonic (only at final segments some collapses are observed during occurrences of the overloads). In the case when larger overloads are applied, the plots a= f(N) represent (with their collapses) nearly every overload, but in a very distinct manner. The increments in the crack lengths within the exact cycles of over- load application are greater, as well. Characteristic are also the plots of the corresponding relationships of the crack propagation rate da/dN against the range of the stress intensity factor ∆K. The experimentally found and re- corded relationships da/dN = f(∆K) can give – in the course of tests with overloads applied in cycles – a plot as shown in Fig.5a. Between subsequent overloads and with systematically increasing ∆K (together with increasing of the crack length) the crack growth rate decreases – see the vertical columns in Fig.5a with slightly negative inclination towards the X axis – which se- ems to contradict the Paris relationship. The shown plot corresponds to the overload kov =1.75 applied every DN =15000 cycles. The propagation rate decreases as much as by 2-3 orders of magnitude. The decrease in the crack growth rate (its level and range of the occurrence) depends mainly on the value of the applied overload. Subsequent figures show relationships analogous to that in Fig.5a, but arrived at with specimens subjected to tests at smaller overloads, i.e. kov =1.6, 1.2, respectively. It is evident that with the level of the overload getting lower, the degree of retardation growth gets reduced (the rate of propagation does not get so much reduced); the crack starts escaping the retardation phase: for the overload kov = 1.6 the statement is true for higher values of the crack length, whereas for the overload kov = 1.2 even for evident predominance of the non-retarded growth between the overloads. Obviously, the level of retardation and the range of its occurrence substan- 838 S.Kłysz Fig. 5. da/dN = f(∆K) relationship for specimen tested with overload k ov =1.75, 1.6, 1.2 Fatigue crack growth peculiarities... – part 1 839 tially affect the final fatigue life of the specimens under tests. In this case, with even (i.e. every DN number of intervals between the overload applica- tion) the fatigue life was changing for particular levels of the applied overloads as 150000:120000:100000:71560 cycles. It has become apparent that the ana- lysis of fatigue crack growth can be carried out frommany and various points of view; there are, however, no unequivocal settlements – in particular, when various kinds of materials and effects of specific kinds of the overloads upon them are analysed. 5. Conclusion Inmost practical engineering situation the applied stresses fluctuate (often in a random manner) and, under this condition, the failure occurs at lower stress levels than it would be expected when a steady stress were applied. This phenomenon is called the fatigue and causes the vast majority of in- service failures. The service life of a structure is limited by the critical element whose fatigue life falls either to low cycle fatigue, high cycle fatigue, thermo- mechanical fatigue or creep damage. It is essential to understand how their greating is related with the fatigue life limit and the rate of approaching it. This paper is concerned with corresponding aspects of theoretical description of fatigue crack evaluation under the condition of overloads occurrence. The experimentally determined basic relationships of fatigue crack growth a= f(N) and da/dN = f(∆K) are analysed in the paper. They show diffe- rent features for both different and the same materials. It refers, first of all, to dynamics of the crack length growth convexity of the curves after appli- cation of the overloads. This is especially interesting because they contradict each other under the same test conditions. They depend on many different factors, and investigators do not know all relations between them as well as the crack growth rate or fatigue life. Admitteally, very important problems, which could put some light and explain this characteristic are conected with some microstructural aspects and mechanisms of the fatigue crack growth – which are not described in this paper. The above mentioned features make creation of uniform theoretical de- scription rather difficult. They sometimes preclude simple mathematical mo- dels from being applicable to the involved problems. TheWheeler retardation model is one of the oldest and simplest used in the analysis of crack growth phenomena in specimens undergoing overloads. It is very useful inmany cases 840 S.Kłysz but not for all purposes. To avoid in engineer practice the unacceptable risk of a catastrophic failure it is necessary to monitor the fatigue life of the critical elrments and retire them from service before their allocated life has been exceeded. Because the ability of structural elements to resist the effects of failure mechanisms is a functionofmaterial properties,designandoperatingconditions, then thorough undestanding of the failure mechanisms is essential if the failure modes, the safe-life and the service life of the elements are to be accurately determined and safelymonitored.Therefore, in part 2 of this paper themodification of the Wheelermodel improving the description of experimental results is presented. Because of its good mathematical and physical fundamentals it seems that more complete description of different behaviour of cracks in variousmaterials. Usually, results of analysis of fatigue crack growth are used to identify the critical locations of a fracture in the analysed elements. References 1. Bernard P.J., Lindley T.C., Richards C.E., 1977, The Effect of Single Overloads on Fatigue Crack Propagation in Steels,Metal Science, Aug./Sept., 390-398 2. Bertel J.D., Clerivet A., Bathias C., 1983, On the Relation Between theThreshold and the Effective Stress Intensity FactorRangeDuringComplex Cyclic Loading, Fract. Mech.: Fourteenth Symposium – I. ASTM STP, 791, 336-379 3. Blom A.F., 1989, Overload Retardation During Fatigue Crack Growth Pro- pagation in Steels of Different Strengths, Scandinavian J. Metall., 18, 197-202 4. Bucci R.J., ThakkerA.B., SandersT.H., SawtellR.R., Staley J.T., 1980, Ranking 7XXXAluminium Alloy Fatigue CrackGrowth Resistance Un- der ConstantAmplitude and SpectrumLoading. Effects of Load SpectrumVa- riables on Fatigue Crack Initiation and Propagation,ASTM STP, 714, 41-78 5. Chang J.B., Engle R.M., Stolpestad J., 1981, Fatigue Crack Growth BehaviourandLifePredictions for 2219-T851AluminiumSubjected toVariable Amplitude Loadings, Fract. Mechanics: Thirteenth Conference, ASTM STP, 743, 3-27 6. Chehini C., Schneider M.L., Robin C., Pluvinage G., Lieurade H.P., 1984, Influence of Multiple Overloads on the Propagation of Fatigue Cracks, Proc. 2nd Int. Conf. on Fatigue and Fatigue Thresholds, 915-925, Univ.of Bir- mingham Fatigue crack growth peculiarities... – part 1 841 7. ChenG.L.,RobertsR., 1985,DelayEffects inAISI 1035Steel,Engng Fract. Mech., 22, 201-212 8. Damri D., Knott J.F., 1991, Transient Retardations in Fatigue Crack Growth Following a Single PeakOverload,Fatig. Fract. Engng Mater. Struct., 14, 709-719 9. Fuchs H.O., Stephens R.I., 1980, Metal Fatigue in Engineering, A Wiley- Interscience Publication 10. Jacoby G., 1970, Comparision of Fatigue Lives Under Conventional Program Loading andDigital RandomLoading,ASTM STP, 462, 184-202 11. Kłysz S., 1991, Modelling of Fatigue Crack Growth Within the Stress- Concentration Regions (in Polish), The PhD Dissertation, Military Academy of Technology,Warsaw 12. Kocańda S., Szala J., 1985, Fundamentals of Fatigue Calculations (in Po- lish), PWN,Warsaw 13. Schijve J., 1972, The Accumulation of Fatigue Damage in AircraftMaterials and Structures,AGARD, Proc. Conf., Lyngby, Denmark, 118 14. Schijve J., 1973, Effect of Loading Sequences on Crack Propagation Under Random and ProgramLoading,Engng Fracture Mech., 5, 269-280 15. Schijve J., 1976, Observations on the Prediction of Fatigue Crack Growth PropagationUnder Variable Amplitude Loading,ASTM STP, 595, 3-23 16. Schijve J., 1985,The Significance of Flight-Simulation Fatigue Tests, LR-466, Delft Univ. of Technology, The Nederlands 17. Schulte K., Trautman K.H., Nowack H., 1984, New Analysis Aspects of the Fatigue Crack Propagation Behaviour by SEM-in Situ Microscopy, AGARD, Proc. Conf., 376, 16-1÷16-10 18. Sehitoglu H., McDiamid D.L., 1980, The Effect of Load Step-Down on Fatigue Crack Arrest and Retardation, Int. J. Fatigue, 2, 55-60 19. Shin C.S., Fleck N.A., 1987, Overload Retardation in a Structural Steel, Fatig. Fract. Engng Mater. Struct., 9, 379-393 20. Shin C.S., Hsu S./H., 1993, On the Mechanisms and Behaviour of Overload Retardation in AISI 304 Stainless Steel, Int. J. Fatigue, 15, 181-192 21. SkorupaM., 1996,Empirical Trends and PredictionModels for Fatigue Crack Growth Under Variable Amplitude Loading, Netherlands EnergyResearchFun- dation, ECN-R-96-007 22. Ward-Close C.M., Blom A.F., Ritchie R.O., 1989,MechanismsAssocia- ted with Transient Fatigue CrackGrowthUnder Variable Amplitude Loading: An Experimental and Numerical Study,Engng Fract. Mech., 32, 613-638 842 S.Kłysz 23. Wheeler O.E., 1972, Spectrum Loading and Crack Growth, Trans. ASME, J. Basic Eng., 94, 181 24. Zhang S., Marrisen K., Schulte K., Trautman K.H., Nowack H., Schijve J., 1987, Crack Propagation Studies onAl 7475 on the Basic of Con- stant Amplitude and Selective Variable Amplitude Loading Histories, Fatig. Fract. Engng Mater. Struct., 10, 315-332 Osobliwości rozwoju pęknięć zmęczeniowych i modyfikacja modelu opóźnień Wheelera – część 1 Streszczenie Przedstawiono osobliwości przebiegu procesu pęknięć zmęczeniowychw odniesie- nu do krzywych długości pęknięcia w funkcji liczby cykli lub prędkości propagacji w funkcji zakresu współczynnika intensywności naprężeń – to jest krzywych typo- wychdlabadańzprzeciążeniamizadawanymicyklicznie.Badano stopaluminiumPA7 i stal 18G2A.Omówionotrudnościwopisiemodelowympropagacjipęknięć (woparciu o model opóźnieńWheelera) w związku z omawianymi osobliwościami. Manuscript received March 19, 1999; accepted for print May 9, 2001