A quantitative relation for static recrystallization critical temperature model of Nb micro-alloyed steel by coupling symbolic regression

Abstract

The strain-induced precipitation, recovery, and recrystallization interactions occurring during the rolling process of Nb micro-alloyed high-strength low-alloy (HSLA) steel significantly influence the evolution of austenite, thereby determining the final properties of the product. Therefore, establishing a highly accurate static recrystallization critical temperature (SRCT) model is crucial for setting the parameters of the hot rolling process. However, to date, a mathematical model that is both highly accurate and concise, specifically applicable to Nb micro-alloyed steel, has not been established. In this study, based on the SRCT experimental data of Nb micro-alloyed steel, the SRCT prediction model was constructed with physical metallurgy guidance and combined with symbolic regression machine learning (SRML) algorithm. The model uses the contents of dissolved elements such as Nb, C, N, Mn, Si, the solubility temperature (T_s), strain (ε), strain rate ( $\stackrel{·}{\epsilon }$ ), and austenite grain size (d_γ) as input parameters. Compared to existing models, the SRCT prediction results of the proposed model shows a reduction in root mean square error (RMSE) from 37.1°C to 4.4°C and an increase in the correlation coefficient (R²) from 0.0279 to 0.9865. Additionally, the proposed model was used to analyze the effects of variables such as Si and Mn content in the steel, strain, and strain rate on the determination of SRCT, which were consistent with experimental results. Furthermore, the predictive accuracy of the SRCT model proposed in this study was further validated by combining additional experimental data.