Breakout Prediction Based on Twin Support Vector Machine of Improved Whale Optimization Algorithm

Abstract

Breakout is one of the hazardous industrial accidents in continuous casting production, which has adverse effects on operational stability, product quality, personal safety and equipment life. Therefore, the study of the breakout prediction model is of great significance. A hybrid intelligent algorithm is proposed based on twin support vector machine of improved whale optimization (LWOA-TSVR) by introducing the Levy flight algorithm. It uses the LWOA algorithm to solve the optimization problem of the objective function in the TSVR algorithm, so as to obtain a new breakout prediction model. The simulation and industrial trials results show that the model has higher recognition accuracy than the traditional detection method, and can predict all the breakout signals accurately and timely. Five advanced algorithms are compared to further verify the performance of the model. The results prove that LWOA-TSVR has the advantages of fast convergence speed and high prediction accuracy compared with the other four algorithms. The model is applied to the practical production of a steel mill. Finally, the reported ratio of the model is 100% and the prediction accuracy is 98.2%, which is found to be more effective than the practical application system used in continuous casting production. Hence, this model provides a theoretical basis for breakout detection technology.