Hybrid-driven probabilistic damage assessment of creep-fatigue-oxidation interaction

Abstract

This paper presents a hybrid-driven probabilistic damage assessment approach by considering creep-fatigue-oxidation damage interaction (CFO-DI). Based on generalized strain energy density exhaustion (GSEDE) framework, the hybrid-driven concept integrates the strengths of both physics-based models and machine learning, exploring the frontier from deterministic evaluation to probabilistic assessment. Experimental investigations involving generalized creep-fatigue loading tests are conducted to establish a comprehensive dataset in Inconel 718 at 650 °C. Deterministic models for fatigue, creep, and oxidation damages are developed, and their interactions are analyzed using the GSEDE framework. To tackle limited experimental data, a divide-and-conquer strategy employing machine learning models is implemented for data augmentation. Probabilistic assessments are performed incorporating uncertainties from material properties, loading conditions, and model parameters using Monte Carlo simulations and Latin Hypercube Sampling. The results demonstrate accurate life prediction accuracy and reliable probability distributions in the presence of oxidation damage. Finally, a novel three-dimensional probabilistic CFO-DI assessment diagram quantified by the confidence level is developed, providing a technical pathway for safe-life design in high-temperature structural applications.