A comprehensive evaluation of conventional methods for estimation of fatigue parameters of steels from their monotonic properties

A number of conventional and machine learning-based methods for the estimation of strain-life fatigue parameters from monotonic properties have been proposed in the literature and new ones are continuously being developed. With the development of new steels having improved properties such as higher ultimate strength, selecting appropriate estimation method becomes both increasingly difficult and important. In this study, according to the proposed detailed evaluation methodology, 10 conventional estimation methods have been evaluated regarding their accuracy and applicability for estimation of low- and high-cycle fatigue lives of unalloyed, low-alloy and high-alloy steels which were further divided into low- and high-strength subgroups. For this purpose, detailed, fully populated material dataset was assembled, including monotonic properties, cyclic stress–strain, and strain-life fatigue parameters of 75 unalloyed steels, 104 low-alloy steels and 44 high-alloy steels. With separate analyses, averaging of evaluation results was avoided and more detailed and precise ratings of individual methods were obtained. Complete material data and full results of the evaluations are made available in the paper which may further facilitate development and objective evaluations of new estimation methods and other predictive models.