The prediction model for strip crown in hot rolling based on BES-LSSVM

In the hot rolling field, the traditional control models limit further improvement of finished product quality. To enhance the control level of strip crown in hot-rolled strip steel, this study develops a crown prediction model based on Least Squares Support Vector Machine (LSSVM). Two different methods, Whale Optimization Algorithm (WOA) and Bald Eagle Search (BES), are employed to optimize the control parameters of the model. Ultimately, the predictive performance of three models—LSSVM, WOA-LSSVM, and BES-LSSVM—is evaluated. The results indicate that the BES-LSSVM model achieves the highest prediction accuracy, fastest convergence speed, and closest convergence to the global optimal solution, with R-squared of 0.92 and Root Mean Square Error (RMSE) of 1.2. It effectively addresses the challenge of low-precision crown prediction caused by frequent changes in strip specifications. Implementing the proposed model in industrial applications optimizes rolling process control parameters in reverse, significantly increasing strip crown hit rates. The hybrid intelligent model BES-LSSVM demonstrates strong generalization and robustness, offering high prediction accuracy applicable to engineering practice, thereby providing valuable guidance for hot-rolled strip steel production.