Hybrid prediction model for strip width based on improved mechanism and data-driven model

Accurate prediction of strip width is a key factor related to the quality of hot rolling manufacture. Firstly, based on strip width formation mechanism model within strip rolling process, an improved width mechanism calculation model is delineated for the optimization of process parameters via the particle swarm optimization algorithm. Subsequently, a hybrid strip width prediction model is proposed by effectively combining the respective advantages of the improved mechanism model and the data-driven model. In acknowledgment of prerequisite for positive error in strip width prediction, an adaptive width error compensation algorithm is proposed. Finally, comparative simulation experiments are designed on the actual rolling dataset after completing data cleaning and feature engineering. The experimental results show that the hybrid prediction model proposed has superior precision and robustness compared with the improved mechanism model and the other eight common data-driven models and satisfies the needs of practical applications. Moreover, the hybrid model can realize the complementary advantages of the mechanism model and the data-driven model, effectively alleviating the problems of difficult to improve the accuracy of the mechanism model and poor interpretability of the data-driven model, which bears significant practical implications for the research of strip width control.