Advancements in stress‐based multiaxial fatigue prediction: A data‐driven approach and a new criterion

Abstract

AbstractA highly accurate iteration of the Liu & Zenner multiaxial fatigue criterion is presented and validated against the FatLim dataset. The proposed criterion modifies the mean stress parameters of the original criterion while keeping the amplitude parameters unchanged. The investigation of the best‐performing stress amplitude parameters revealed that existing multiaxial fatigue criteria that rely on dimensional weight parameters exhibit a notable prediction bias. This bias leads to inaccurate fatigue assessments. The paper sheds light on the extent of this bias and its implications for common engineering practices. By introducing the refined criterion, the study addresses these limitations and provides a more reliable framework for multiaxial fatigue analysis. The comprehensive validation against the FatLim dataset demonstrates the superiority of the proposed criterion in accurately predicting fatigue behavior under diverse loading conditions. The findings underscore the significance of considering both amplitude and mean stress parameters in multiaxial fatigue criteria, offering valuable insights for enhancing the accuracy of fatigue assessments in engineering applications.