Machine Learning Model and Prediction Mechanisms of Bainite Start Temperature of Low Alloy Steels

Abstract

The random forest regression (RFR) model was proposed to predict the bainite start temperature (B_s) using alloying elements, such as C, Mn, Si, Ni, Cr, and Mo, as well as the prior austenite average grain size (AGS). RFR demonstrated a performance improvement of approximately 1.2% over the empirical equation. Cr, C, Mo, Mn, Si, AGS, and Ni were assigned importance, in that order, in the RFR using Shapley additive explanation (SHAP) analysis. The detailed prediction mechanisms of the RFR are discussed using the SHAP dependence plot.