Based On Improved Particle Swarm Optimazation Of Rolling Force Model And Self-learning

With the progress of science and technology, the aluminum alloy products are becoming more and more widely used, and the precision requirement of aluminum products is becoming more and more highly. In the process of rolling, rolling force plays a very important role in the quality and performance of the strip. At present, in the actual production, we get the traditional rolling force model with the theoretical calculation, and the accuracy cannot satisfy the contemporary high quality requirements of aluminum products.Aluminum strip rolling process is relatively complex, and not so easy to control, like steel rolling force, and it has complex relationship between the parameters such as flattening radius, tension, rolling speed, so it does not have high accuracy of forecasting on the basis of traditional formula and experience value. High precision of the predicted rolling force is the powerful guarantee of the quality of aluminum plate, so after analyzing the traditional theory of rolling mathematical model, this article optimize the clustering radius of nearest neighbor clustering algorithm by using an improved particle swarm optimization(PSO) algorithm, and then using the nearest neighbor clustering algorithm we calculate the center vector of the RBF neural network hidden layer, and the powerful reasonable combination of them decides the hidden layer structure of RBF neural network. Finally, based on the nonlinear approximation ability of RBF neural network and its self learning characteristics, we propose the modeling method based on RBF neural network.The algorithm we put forward is used in rolling force prediction of the part of finishing mill group in the scene of the "1 + 4" aluminum strip in some factory. And according to the actual data from the field, we will train and test the established model. Compared with the basic particle swarm algorithm optimization RBF neural network, it has greatly progress in the prediction precision and convergence speed in the simulation result, and it effectively improves the prediction precision of rolling force. Finally, in order to make the established model better to adapt to the actual conditions, we also briefly introduce the application of model self-learning.