Stacking ensemble surrogate modeling method based on decomposed- coordinated strategy for structural low-cycle fatigue life reliability estimation

Traditional reliability analysis methods for the low-cycle fatigue (LCF) of mechanical structures consistently adopt one-step modeling approach, which tends to lack satisfactory precision when dealing with highly nonlinear problems. To address above issue, the Stacking ensemble surrogate modeling method based on decomposed-coordinated strategy (DCS-SESMM) is proposed. In this study, the Manson-Coffin (M-C) formula serves as the foundation, and the LCF life estimation problem is transformed into a two-step data-driven modeling approach through the decomposed-coordinated strategy (DCS). Firstly, the sub-surrogate models are developed to accurately capture the stress-strain responses under random load variables. Then, the predictions of sub-surrogate models are combined with fatigue parameters to construct the main LCF life model based on the M-C formula. Additionally, the Stacking ensemble learning method is introduced to improve the robustness of surrogate model by integrating data features from different training rules. Finally, the general applicability of DCS-SESMM is validated through the probabilistic analysis of a nested function and the LCF life reliability estimation of turbine blisk, demonstrating its excellent modeling characteristics and simulation performance. The proposed method can be applied to the LCF life reliability estimation of various mechanical structures, providing valuable insights for the LCF life reliability assessment.