Sequential physics-data coupling framework for multi-model life prediction of fatigue, creep, and creep-fatigue

Accurate life prediction for high-temperature components subjected to fatigue, creep, and their interactions presents considerable complexity, driven not only by coupled degradation mechanisms but also by the inherent intricacy of high-precision models requiring rigorous parameter fitting. In practice, model selection is often guided by application-specific requirements through resource-intensive trial-and-error processes, potentially overlooking synergistic insights from complementary approaches. This study introduces a sequential physics-data coupling framework designed to reconcile these challenges. Classical physical models are harmonized through multi-objective optimization to generate robust baseline predictions. A data-driven correction module further refines predictions by adaptively correcting model biases via confidence-guided multi-task learning. Validated on five alloys including GH4169, TC4, MAR-M247, 9Cr1Mo, and 304HCu, the framework demonstrates enhanced generalization across creep-fatigue, high-cycle fatigue, and creep rupture scenarios. By synergizing physics-based interpretability with adaptive corrections, it reduces over-reliance on single-model selection while maintaining computational efficiency, offering a practical tool for engineering reliability assessments.