A new fatigue damage prediction model applicable to multi-slope S-N curves

Abstract

The S-N curve methodology, deeply rooted in fatigue data derived from testing specimens, continues to be a prevalently utilized and foundational approach for fatigue evaluation within engineering disciplines. Nonetheless, the occurrence of fatigue limits frequently induces multi-segment patterns within S-N curves, thereby instigating discrepancies in both the slope and intercept values. This study introduces a fatigue damage prediction model tailored for multi-slope S-N curves. It employs a novel rainflow stress range distribution function, integrating two Rayleigh distributions to improve the accuracy of stress distribution moments while minimizing unknown parameters. Initially, the research delineates four bimodal power spectral density configurations and formulates a parameter search strategy for the identification of standard spectra that correspond with a wide array of bandwidth parameters. Subsequent to the parameter filtration process, a selection of 2241 standard spectra, encompassing a diverse range of bandwidth parameters, is made. Through the application of these meticulously selected standard spectra, a series of comparative analyses are executed amongst the predictive outcomes derived from the DK, ZB, JB, MN, ZL models, and the newly proposed model, across varied multi-slope S-N curves and differential fatigue limits. The insights gleaned from the relative error assessments underscore the commendable predictive precision of the novel model across multiple bandwidth processes. Notably, the model exhibits significant enhancements in predictive performance for processes delineated by a normalized spectral width parameter below 0.8.