Generalizing multiaxial vibration fatigue criteria in the frequency domain: A data-driven approach

Random vibration fatigue of structures is intrinsically multiaxial, driven by complex stress states. Non-proportional multiaxial stresses exacerbate fatigue damage compared to proportional stresses. Frequency domain fatigue analysis is generally more efficient for structures operating near natural frequencies. However, handling non-proportional multiaxial stress effects in the frequency domain remains challenging. This paper introduces a data-driven framework that reformulates time domain multiaxial fatigue failure criteria into their frequency domain counterparts, explicitly addressing non-proportional effects. The method introduces a non-proportional correction factor into the power spectral density of equivalent stress, determined through a data-driven approach, converting various time domain multiaxial fatigue criteria into their frequency domain equivalents. Combined with modal decomposition analysis, this approach efficiently evaluates multiaxial vibration fatigue for engineering structures, effectively accounting for non-proportional stress effects.