Modeling On Transverse Thickness Of Hot Finishing Rolling Aluminum Strip Based On Neural Network And Flatness Control

With the rapid development of China’s aerospace industry, high precision aluminum strip is growing demanded. China’s mill equipment of aluminum hot rolling is relatively backward and lack of flatness control means. Thickness of hot rolling aluminum is used more and more thinly, in order to reduce costs and improve efficiency, the hot rolling strip is directly used more and more to replace the cold rolling strip; and hot-rolling aluminum strip with poor flatness easily lead to cold-rolled strip with poor flatness. So it has very important significance to control flatness in the hot-rolling aluminum strip. The flatness control in hot rolling aluminum strip can be achieved by controlling the transverse thickness distribution. In order to get hot rolling strip with good flatness, the neural network modeling and intelligent control of transverse thickness distribution in hot rolling finishing mill are studied. The specific researches are as follows:(1) The mechanical model of4-high rolling mill is established by using influence function method. Combine with the model of aluminum flow stress, the rolling force is calculated by the measured data. The correctness of mechanical model is proved compared with the measured rolling force. The influences of bending force, roll crown and roll diameter are analyzed on transverse thickness distribution based on mechanical model.(2) For aluminum hot rolling process with multi-variables, strongly coupling, nonlinear, it is difficult to establish accurate mathematical model to solve transverse thickness distribution. A BP neural network is established to predict the hot rolling aluminum transverse thickness distribution while the forces are considered only. The transverse thickness distribution is obtained with the work rolls thermal crown. The factors of great influences on transverse thickness distribution are chosen as the input of the BP network and The BP network are established for single-channels. Comparing the measured data, the relative errors of the simulation data are less of1%.(3) A BP neural network is established to predict the overall transverse thickness distribution. The factors of great influences on transverse thickness distribution based on influence function method are chosen as the inputs of the BP neural network and the outputs are the overall transverse thickness distribution. The genetic algorithm is used to optimize the network weights and thresholds. The training time of BP network is reduced and convergence performance is improved. Comparing the prediction results of GA-BP model with the measured transverse thickness distribution, the relative errors are less of0.8%.(4) The hydraulic bending roll system is used to control the crown for controlling the flatness of hot rolling strip. Comparing conventional and fuzzy PID control strategy, simulation results show that the fuzzy PID and conventional PID response speed almost, but overshoot of fuzzy PID is smaller than conventional PID and fuzzy PID can overcome parameters variability and strongly nonlinear of the hydraulic bending roll system. The error between flatness uncontrolled and the standard flatness curve reach0.015mm and the error is less than0.005mm after controlled. A reference is provided for the flatness control of hot rolling aluminum strip.