Predicting the Yield Strength Ratio of Spring Steels via Machine Learning With Experimental Validation

ABSTRACT Spring steels are widely used in automotive and railway systems, in which service reliability is strongly affected by the yield strength ratio (YSR). However, accurate YSR prediction remains challenging because of the nonlinear coupling between alloy composition and heat‐treatment parameters, particularly when limited data are available. This study involves the development of a machine‐learning framework to predict the YSR of spring steels. A hierarchical feature‐selection strategy combining Spearman correlation analysis, SHapley Additive exPlanations analysis, and best‐subset optimization is employed to identify key variables and reduce redundancy. After feature selection, the eXtreme Gradient Boosting model achieves the coefficient of determination of 0.841 and the root mean square error of 0.017. In addition, multi‐objective optimization and Pareto front analysis are integrated to guide the design of the composition process. Two distinct alloy‐process combinations that satisfy the target requirements (YSR = 0.9 and elongation ≥ 10%) are identified and experimentally validated. The proposed framework provides an interpretable and efficient strategy for the data‐driven design of high‐strength spring steels.