Research On Tension System Identification And Control Of Tandem Cold Mill Based On Neural Network

Steel industry is the backbone of improving national economy and cold rolled sheet is one of the most important products. It is a kind of high value-added steel. Nowadays, cold rolled duality sheet products are mainly large-scale efficiently produced by tandem cold mills. The mill is composed of numerous equipments and its control process is quite complicated. Rolling with tension is one of the most important features of tandem cold rolling. The accuracy of tension control will greatly affect the stability of rolling and quality of finished strips. As the matter of this point, researching deeply on the control of tension is quite meaningful both in theory and reality.The thesis analyzes calculated tension model of tandem cold rolling and clarify the method of tension control based on the studying on the structure of tandem cold mill. Each part’s mathematical model of tension is deduced and then the mathematical model of rolling gap adjusting tension system is established based on the detailed analysis on every part of tension system of tandem cold mill.The mathematical model for control is deduced by using mechanism method. The neural network is utilized combined with system identification principle to identify mathematical model of tension control system for tandem cold mill. The better approximation model is achieved by stimulating with Matlab. The model, whose error can meet the needs, can be used to control tension system.Tension is controlled online by using the model which is identified by neural network. A system is designed which can identify and control tension at the same time. It takes a BP neural network as an identification device to approximate the real tension system and controls the tension system with a liner neural network. Because NN can adjust its weights, the identified model can adjust itself adaptively according to the real model’s changes. And in the meantime, in order to make the control result accuracy and timely, the control network changes the tragedy and parameters of control adaptively with the changes of identified model. The result is compared with traditional control for mechanism model by carrying out the Matlab simulation experiment. It turns out that the identification and control system can get a better response, as well as, the anti-interference ability and adaptability to the changes of system’s parameters are improved.According to the shortcomings of NN control system, fuzzy PID method is studied to improve the control method. The NN controller in the old system is replaced by a fuzzy PID control part to avoid the long adjusting time and considerable fluctuations in the early period. Its effectiveness is validated by the Matlab simulation, and both response time and fluctuations of the improved system are better.