Optimization Of Slab Shape Control Model For Unsteady Rolling Process Based On Data Analysis

In the actual production of hot continuous rolling,the shape control of stable production process has reached a very high level,but in the production process,there are a lot of unsteady rolling processes such as specification switching,stop rolling and roll changing,the actual shape control effect is low,which seriously affects the quality and yield of products.In order to solve this problem and improve the control accuracy of strip shape,this paper takes a 1580mm hot rolling production line as the research object.Based on the analysis of production process data,from the two aspects of load distribution optimization and strip shape prediction,the load distribution strategy aiming at the optimization of strip shape is obtained,and the prediction model of strip crown is established based on depth learning,which improves the unsteady rolling process finally The control accuracy of plate crown in.The main contents of this paper are as follows:(1)Analysis of the influencing factors of plate shape in unsteady rolling process.Based on the control model of hot rolling process,the influence law of temperature,rolling force,roll wear and roll thermal expansion on the shape of roll was systematically analyzed.Compared with the steady rolling process,the load distribution,rolling force,bending force and interroll amount fluctuate greatly in the unsteady rolling process,such as specification switching and roll changing,resulting in a low plate crown hit ratio.By analyzing the data of the unsteady rolling process,such as specification switching,stop-rolling and roll-changing,the main factors influencing the low plate protruding life rate were found out.(2)Study on the optimization strategy of load distribution aiming at plate shape.In the case of unstable load distribution in unsteady rolling process,the objective function of plate shape and rolling force balance was established,and the objective function was solved by GA algorithm and improved NSGA-II respectively,and the optimal load distribution scheme was obtained.(3)Based on the data analysis and mining of the unsteady rolling process in the field,the multi-layer network structure was determined and the DBN-a and DBN-b prediction models were established for the specification switching and roll changing processes.By improving the learning rate and momentum terms and introducing the conjugate gradient method,two models IDBN3-a and IDBN3-b were obtained,and simulation experiments were carried out to verify the predictive performance of the improved model.The research results of this paper have strong practicability for the shape control in the unsteady rolling process of hot rolling,and have theoretical guidance significance for the shape control in the unsteady rolling process.