Dynamic Rolling Force Prediction of Reversible Cold Rolling Mill Based on BP Neural Network with Improved PSO

In order to improve the prediction accuracy of the rolling force in dynamic rolling process of reversible cold rolling mill, the two rolling force prediction methods combining improved PSO(Particle Swarm Optimization) and back-propagation(BP) neural network are proposed in this paper, they are LPSO-BP model and MPSO-BP model respectively. Firstly, we train the BP neural network with PSO instead of gradient descent method to solve the local minimum problem. Then, in order to enhance the search ability of PSO-BP algorithm, we combine the two improved PSO algorithms: LPSO and MPSO with BP neural network to predict the rolling force. Finally, four neural network models: BP, PSO-BP, LPSO-BP and MPSO-BP are constructed in MATLAB environment, and the practical production data are used in the rolling force prediction experiment. The final experimental results show that the MPSO-BP model has the lowest prediction error 2.12 % compared with the other three models. In other words, the MPSO-BP model has higher prediction accuracy and better stability, which is suitable for practical applications.