Machine Learning-Based Improved Creep Life Prediction of 316 Austenitic Stainless Steel with Add-on Chemical and Microstructural Features

Abstract 316 austenitic stainless steel (AusSS) is extensively used in industrial and structural applications due to its excellent corrosion resistance, high strength, durability, and resistance to creep at elevated. This study advances the understanding of creep life mechanics in 316 AusSS by integrating a comprehensive set of sixteen hitherto unconsidered chemical (wt.% of C, Si, Mn, P, S, Ni, Cr, Mo, Cu, Ti, Al, B, N, and Nb + Ta), and microstructural (austenitic grain size number and non-metallic inclusion) features alongside two key physical features (test temperature and stress). By employing eight classical empirical models, three machine learning (ML) approaches, and shallow neural networks, the research provides a robust comparison against unseen test data. Notably, the XGBoost model demonstrates the highest accuracy (98.4%) and lowest prediction error (2.3%) in creep life prediction, underscoring its effectiveness. Through SHAP analysis, the expanded feature set's influence on creep life prediction is elucidated, revealing how chemical and microstructural properties play a pivotal role in more accurate forecasting. This interdisciplinary approach emphasizes the integration of computational methods with data-driven techniques, advancing materials science through novel computational insights and predictive modeling of material creep behavior. The study underscores the synergy between computational and experimental data, offering valuable improvements in predictive models for inorganic materials like 316 AusSS.