Force Model Research And Application Of Strip Cold Rolling Mill

With the increasing modernization of cold rolling mill, the technical staff of steel rolling industry, keenly aware the importance of the mathematical model of rolling. Therefore, the major manufacturers pay much attention to the development, maintenance and optimization of rolling process mathematical model. Because the accuracy of mathematical models determine the yield and quality of rolled products directly, that how to improve the forecast accuracy of rolling mathematical models become the focus and difficulty. In this paper, based on existing literature, some work has done to improve the accuracy of rolling force prediction.Firstly, this paper based on the theory of cold rolling, analysis of a variety of cold rolling force calculation models, and compared them. Focus on discussing two major factors which impact the pre-set precision of rolling force, material deformation resistance and friction coefficient. And details explain the regression model of deformation resistance. Then select the two different forms of deformation resistance model, and obtain the friction coefficient by different methods, substituting them to the rolling force formula of Bland-Ford, and compare with the measured value, to choose better ideal models of deformation resistance and friction coefficient.Then, for the uncertainty of the rolling condition, and the rolling parameters given by rolling schedule can not completely satisfy the technology and equipment. Therefore, it's necessary to continue optimize in the practice production based on the measured values. This chapter based on the model adaptive theory, using the method of exponential smoothing to study the adaptive algorithm of deformation resistance. According to the field measured data, achieve the model adaptive. The simulation study shows that deformation resistance model after adaptation can follow the measured deformation resistance very well.Finally, we do a research on the rolling force prediction of neural network in practice application. According to the characteristics of five-stand cold rolling mill, the parallel structure of BP neural network model is proposed, and LM algorithm be used to train the network. Then the material is classified according to different specifications, to determine the record number. We store the trained network structure and parameters in the self-learning data file according to the record number, and create a database. When use it online, according to the specifications of the rolling species to call. The simulation test shows that the rolling force model of neural network can effectively improve the forecasting accuracy of rolling force and provide convenience to predict rolling force on the line.