Design of control system for steel strip-rolling mill using NARMA-L2

Abstract

A Simulink and mathematical models have been proposed in this study in order to control the thickness in a rolling mill process. The simulation results show that the thickness oscillation can be manipulated with high accuracy by using NARMA-L2, since it can remove the non-linearity of servo system and other disturbances complexities. The proposed NARMA-L2 controller model has been compared with a more popular PID controller and demonstrated high efficiency. The model also demonstrated its ability to be resilient under any sudden changes in system dynamics. This efficacy continues to increase even in linear or nonlinear disturbances. For proving the efficiency of the model, it had been tested with different cases such as uniform or variable thickness, large thickness, small thickness... etc, and it had appeared acceptable robustness when subjected to sudden disturbances. Very small rise time and overshoot can be obtained. Also, with precision optimization of the main parameter of NARMA-L2 the steady-state error may be removed. Accurate profiles can be manufactured for the steel plates in a hot rolling mill by consideration of the presented strategy. More sources of nonlinearities of rolling systems are taken into account in this work. The results of the nonlinear case have been implemented for uniform and non-uniform thickness of the steel plate, where, it appeared very acceptable performance of the proposed model with high efficiency to avoid the sharp edges of the sheet profile. Recent manufacturing requirements need high precision of milling products, therefore, the present study explains the intelligent strategies that can be considered to reach this object. The result shows that fast control behavior with low uncertainties can be achieved. In some rolling mill applications, variable thickness products cannot be constructed with conventional approaches, theretofore, artificial intelligent with hard learning should be introduced.