Research Of The Flatness Control System On Tandem Cold Mill

With the development of modern industry, the much higher request is bring up for the quality of sheet and strip. The flatness control and pattern recognition has become a hot topic studied in cold rolling, because it is the one of the key technologies of improving the quality of sheet and strip. In addition, because having many advantages in establishing and identification of the model, Artificial Neural Networks(ANN) had been applied extensively in identification and control of flatness in recent years. This paper, regarding 6-h CVC Cold Rolling Mill of a sheet metal company as the research subject, with the purpose to exact and automatic control flatness, established the shape pattern recognition mode and the shape prediction model with the Error Back Propagation(BP) neural network.At first, the thesis discussed the knowledge of rolling mill and flatness, basic methods and strategy of flatness control. The theory of flatness measure device and shape measurement is discussed in detail.In addition, the paper analyzed the defects of traditional shape pattern recognition, and established the shape pattern recognition model with the BP neural network. The model is established by Matlab7.0 neural network toolbox. Input of the model is the deviations of 52 determined values and the mean value of them. Output is the coefficients of the flatness curve of the fourth degree. The coefficients can be directly used to flatness control.Finally, flatness control system is a complex and multiple variable coupling control system, which made it not easy to attain control effect by means of general control method. In order to perform Precise Control of flatness, a shape prediction model based on BP Neural Networks is established by Matlab7.0 neural network toolbox. The model can give effective set value of the actuator in flatness control system. Using experimental data to train and simulate the networks, the result is satisfactory.The establishment of these two models can identify and predict flatness effectively, so it is of significance for the precise control of flatness.