The purpose of this study was to predict rolling contact fatigue life in traction drives. For this purpose, a time series of cracks inside the roller was first observed. As a result, no independent cracks were observed inside the roller, and only cracks originating from the surface were observed. Based on these observations, the crack growth rate and stress intensity factor at each evaluation point were obtained, and the fatigue life Np was calculated by substituting them into the fatigue life prediction equation. As a result, the obtained fatigue lives were longer than the past fatigue test results. Furthermore, the analysis focusing on the Paris law suggested that the tendency of crack propagation might be underestimated. Therefore, the fatigue lives obtained by replacing the longest crack with a hypothetical crack at a specific evaluation point were generally consistent with the past fatigue test results. Therefore, extreme value statistics were applied to properly evaluate the crack growth trend, and the time series of cracks in the continuous growth zone was observed by reducing the pressing force applied to the rollers.
HIRAI et al. (Wed,) studied this question.