Machine learning insight into the mean stress impact on fatigue life of additively manufactured 18Ni300 maraging steel under various multiaxial stress paths

Abstract

This study investigates the effects of mean axial and mean shear stresses on the fatigue life of Laser Powder Bed Fusion (LPBF) 18Ni300 maraging steel under uniaxial, torsional, in-phase, and out-of-phase axial–torsional loading conditions. A Gaussian Process (GP) model is employed to analyze fatigue life data, enabling (i) the estimation of the impact of specific mean stress components and (ii) the identification of dominant damage mechanisms through the selection of key stress-related predictors. Results reveal that static and alternating axial stresses similarly influence fatigue life, an effect captured by the maximum axial stress. This is attributed to the stress-raising geometry of surface pits, which limits axial ratcheting and reinforces the dominant role of maximum axial stress. In contrast, mean shear stress induces angular displacement ratcheting, leading to additional fatigue damage that maximum stress alone cannot account for. Under out-of-phase loading, this angular ratcheting is suppressed, significantly reducing the influence of mean shear stress on fatigue life. The GP model effectively captures the non-linear relationships between stress components and fatigue life. These results emphasize the critical role of mean stress effects and surface features in designing and evaluating AM components, enhancing the understanding of fatigue damage mechanisms in AM steels and aiding the development of predictive life models for complex loading conditions.