Deviation Prediction And Correction Control In Aluminum Hot Continuous Rolling Based On Data-drive

In aluminum hot continuous rolling, deviation of a strip is represented by the distance between the center of the strip and the center of a roll, is external performance of rolling state parameters lose lateral symmetry. Deviation phenomenon will lead to product quality problems of aluminum strip pull off and misalignment in coiling strip, but also may lead to uneven wear of work roll, cutting machines and other equipment failure problems, seriously affecting the production efficiency. It is one of the problems of modern rolling enterprises need to solve. Therefore, this paper based on data-drive discusses the law of strip deviation process, builds aluminum deviation state prediction model, and then designs deviation process control strategies,which have a very important significance to improve the product quality and the life of rolling equipments in the rolling process. The main work are as followsBy analyzing the "1+4" hot rolling production process and on-site monitoring parameters,get the data of rolling process and analysis process deviation rule. To reduce the complexity of predicting the amount of deviation,principal component analysis is used to reduce the dimension of all relevant factors affecting the deviation of the process in the premise of preserving the original information and the contribution of the factors in the main ingredients is analyzed.Based on principal component analysis, the deviation process neural network prediction model is established. On the selected network nodes in the hidden layer, the best hidden nodes is chosen by using trial and error method. The LM algorithm is chosen in network training for the convergence speed and network errors and the test data in field is used to test the model.Based on a large number of process data in rolling scene, multi-rack fuzzy control model including F1-F3, F2-F4rack fuzzy controller model is established by using FCM clustering algorithm for the model structure and foreside coefficient identification,using the least squares method for the Rear coefficient identification. Furthermore, the simulation based on rolling operating data on the industrial scene is conducted to validate the accuracy of this deviation control model.