Discussion: “A New Approach for Fatigue Damage Modeling of Subsurface-Initiated Spalling in Large Rolling Contacts” (Walvekar, A. A., Paulson, N., Sadeghi, F., Weinzapfel, N., Correns, M., and Dinkel, M., 2016, ASME J. Tribol. 139(1), p. 011101)

Abstract

Professor Sadeghi and his coworkers have been developing a method to estimate the rolling contact fatigue life of rolling bearings in recent years using a continuum damage mechanics (CDM) approach. Development of the method has involved progression from a defect-free and inclusion-free model to ones incorporating grain boundary cracks and inclusions.The most recent paper on this topic is authored by Walvekar et al. [1]. In Sec. 2.2, the authors stated “It was hypothesised that failure mechanism for torsional fatigue and rolling contact fatigue are equivalent because the mechanism for damage accumulation is similar for both types of fatigue phenomena.” It is necessary to verify whether such a hypothesis is correct or not but this has not been carried out within the paper.This hypothesis was postulated by Raje and Sadeghi in 2009 in a paper published in the IMechE Journal of Engineering Tribology entitled “Statistical Numerical Modelling of Sub-Surface Initiated Spalling in Bearing Contacts” [2]. In Sec. 3, the authors stated “It is hypothesised that since fatigue damage under rolling contact conditions is caused purely by the action of shear stresses, the mechanism of failure is similar to that prevalent in torsional fatigue. Thus, by knowing the stress-life relationship for a material in torsional fatigue, the fatigue damage parameters can be extracted.”Torsional fatigue data on SAE 52100 bearing steel from Styri were used to provide the shear stress to failure [3], but Styri stated that in about 90% of cases, failures start below the surface, presumably at inclusions or other defects that act as stress concentrating features. However, the model being validated by Raje and Sadeghi is for a material free of defects, voids, or inclusions, while the steel used by Styri would be of 1940 s pedigree but with no statement of its cleanliness.Would the authors comment upon this observation, particularly as, intuitively, the torsional fatigue life of a pristine steel would be expected to be much longer than the one containing defects or inclusions while failures would be surface-initiated?Furthermore, results from the CDM model of pristine steel are compared to L10 bearing lives measured by Lundberg and Palmgren for NU 209 cylindrical roller bearings. The authors present a method in Sec. 4.3 for the conversion of L10 life in shaft revolutions to inner race stress cycles. A comparison of the model results and the bearing lives of Lundberg–Palmgren is then presented. The authors stated that “Stress-life results for this bearing are shown in Fig. 7 and are found to be in good agreement with stress-life results from the current model.”Would the authors comment upon this conclusion, taking into account that the results from the CDM model of pristine steel are being compared to bearings of 1940 s pedigree with no statement of cleanliness?In Sec. 4.2, the authors claim that the model is found to fit within the range of experimental data for cylindrical roller bearings reported by Harris and Barnsby [4]. Reference to this data appears to suggest that shaft revolutions have been plotted as opposed to stress cycles in Fig. 7. Would the authors confirm whether this is the case or not?Intuitively, a defect-free and inclusion-free model should predict lives that are much longer than both the experimental data from Harris–Barnsby and Lundberg–Palmgren. Once defects such as inclusions are introduced into the model, comparisons may then be made with the experimental data. It is necessary to model the size, number, shape, and composition of the inclusions (and other defects, such as carbide stringers) within the steels tested by Harris–Barnsby (1980 s/1990 s pedigree) and Lundberg–Palmgren (1940 s pedigree) in order to carry out such a comparison. Only at this point the claims can be made for validation (or otherwise) of the model. Would the authors comment upon this observation?It also appears that the CDM model has not been validated at lower contact stresses. It may be difficult to validate the model at contact stresses below 2200 MPa (the main region of interest for industrial machines) for the following reasons:Would the authors comment upon validation of the model at lower contact stresses?