Constitutive modeling and experimental analysis of asynchronous multiaxial fatigue in 304LN stainless steel using Bayesian optimization

Abstract

This study proposes a computational framework for modeling the multiaxial cyclic plasticity behavior of 304LN stainless steel under both synchronous and asynchronous strain-controlled loading. A series of biaxial fatigue experiments on thin-walled tubular specimens revealed pronounced nonproportional hardening and path-dependent plastic deformation, particularly under asynchronous loading conditions. To capture these effects, an advanced constitutive model was formulated by integrating multi-component kinematic hardening, a memory-dependent isotropic hardening rule, and a nonproportionality-sensitive internal variable. Model calibration was carried out using Bayesian Optimization, with kinematic hardening parameters trained on eighteen uniaxial hysteresis loops representative of various cyclic conditions to enhance the accuracy of simulated loop shapes. The finalized model was implemented in a finite element framework and validated across seven multiaxial loading paths. Simulations demonstrated high fidelity in reproducing hysteresis loop shapes, peak stresses, strain ranges, and stress path trajectories. Qualitative analysis indicates that asynchronous multiaxial loading in 304LN stainless steel results in elevated plastic strain accumulation and significant nonproportional hardening, effects that are closely associated with the newly introduced Desynchrony Index (DI). Experimental observations further demonstrate that asynchronous loading paths can produce cyclic hardening comparable to, or exceeding, that of classical out-of-phase loading, underscoring the importance of time-varying phase interactions in cyclic plasticity. The evolution of the nonproportionality parameter also shows strong agreement with experimentally derived desynchrony indices, reinforcing the robustness of the proposed approach for fatigue analysis under complex multiaxial loading conditions.