Rolling force prediction of hot-rolled transverse thickness-varying plates based on stochastic configuration network

Rolling force determines the quality and production stability of hot-rolled products, so it is important to be predicted accurately. Nowadays, rolling force prediction is mostly focused on flat rolls, and there are few reports on rolling force prediction for transverse variable thickness steel plates during MAS rolling. In this paper, a SCN-based prediction model for the rolling force of transverse variable thickness plates is proposed, and the prediction model is established with variable thickness parameters, rolling process parameters, and material composition as inputs. The model was tested using actual industrial data, and the results show that the SCN-based rolling force prediction model has the highest prediction accuracy and stability in the paper, with MAE and MAPE of 810.18kN and 2.46%, respectively. The model also has the advantages of simple modeling and less manual adjustment of parameters, which has great potential for application in the field of hot rolling.