The Intelligent Prediction Of Rolling Force Of Reversible Cold Rolling Mill

The steel industry is a very important part of China’s economic development,and cold rolled strip is widely used because of its many advantages.The research object of this paper is the rolling production process of reversible cold rolling mill.Rolling force is a very important setting parameter in the production of strip steel.It directly determines the quality of the product and the control precision of the rolling mill.Therefore,how to accurately predict the rolling force has important research significance.In this paper,the calculation model of rolling force is analyzed based on the basic theory of cold rolling process firstly.Then,the simplified Bland-Ford model is selected as the calculation method of rolling force.In order to solve the problem that friction coefficient μ and deformation resistance K are not measurable in the process of rolling force calculation,two calculation models of μ are deduced by Stone rolling force formula and forward slip neutral angle formula at first,then the calculation model of K is deduced by Bland-Ford formula,the exact values of the calculated μ and K are taken as samples.Then,three friction coefficient prediction models and a deformation resistance prediction model are established based on the theoretical analysis of the rolling process.The coefficients of the prediction model are calculated by using the calculated samples of μ and K,and a model coefficient library for actual prediction is established based on the strip modeSecondly,because in the actual rolling production,the state of the rolling mill and the rolling environment are constantly changing,the prediction model established according to the theory can’t fully meet the prediction accuracy.Therefore,it is necessary to continuously correct the model according to the actual rolling parameters in the production.Based on the model adaptive theory,combined with the measured data after each rolling,different model adaptive algorithms are designed for the steady and variable speed sections of the rolling mill.The simulation experiments show that the error of calculating the rolling force is lower with the predicted values of μ and K after the adaptation.Finally,in order to improve the prediction accuracy of rolling force in variable speed rolling process,the single hidden layer and multi hidden layers neural network models are established based on the analysis of traditional rolling force calculation models to predict rolling force.The network structure is determined by empirical formula and experimental method.The improved PSO(Particle Swarm Optimization)algorithm is used to train the single hidden layer neural network,and the prediction result is better than the gradient descent method.On the basis of the single hidden layer network model,the number of hidden layers of the network is increased to improve the learning ability of the model,and the LM algorithm is used to train the neural network model.The experimental results show that the neural network model with more number of hidden layers has lower prediction error,which provides a theoretical basis for the rolling force setting in actual rolling production.