Prediction of Mechanical Properties of Hot-Rolled Strip Based on KKT-ISVM Algorithm

Hot-Rolled Strip Steel (HRSS), as a critical material in the industry, plays a decisive role in enhancing product performance and improving engineering solutions due to its excellent mechanical properties. This paper proposes an innovative prediction algorithm, the Karush-Kuhn-Tucker (KKT)-based Incremental Support Vector Machine (ISVM) algorithm to predict the mechanical properties of hot-rolled strip steel, including tensile strength, yield strength, and elongation. In the ISVM framework, only samples that violate the KKT conditions are updated, avoiding redundant training on the entire sample set and significantly reducing computational complexity. The predictive results of the KKT-ISVM algorithm are compared with those of the classic ISVM and SVM algorithms. The results show that the KKT-ISVM not only requires less training time but also offers a significant improvement in prediction accuracy. Additionally, the algorithm demonstrates superior performance on four key evaluation metrics, Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), and R-squared ($R^{2}$) proving its efficiency and accuracy in predicting the mechanical properties of hot-rolled strip steel.