Research Of Inner Model Control Based On Neural Networks For Pickling Acid Temperature

In the industry of ferrous metallurgy, the process of pickling technology is very important for the quality of the light cold-rolled sheet and striped steel. It is mainly effected by the function of mechanical and chemistry, and it removed the iron oxide and impurity which attached to the surface of striped steel. As a result, the surface of striped steel is very clear. In the process of pickling the striped steel, the too high or too low acid temperature will be lead to overpickling or underpickling as well as it effects the pickling speed. Therefore, whether controlling acid temperature well or not, it will directly influence the quality of striped steel.This article is based on a background of production of stainless steel pickling of cold rolled. By studying the process of technology of acid temperature control system, Building the mechanism model of controlled plant of the control system. As a result, we can conclude that this system's process of controlling is belong to the large time delay process. According to traditional PID control theory, combining the technology requires, we design this control system to cascade control system. Meanwhile design the PID controller of inner loop to the PID controller which is attached to low-pass filter. So the cascade control system is enhanced the talent of anti-interference. Through the simulation and on-site commissioning, we can find the cascade control system can basically satisfies the requirement of on-site production technology. However, as the main controlled plant----heat exchanger is nonlinear, large delay, multi-disturbance complex system. Designing the inner loop is not enough to resist the disturbance of external loop which is in large delay time. So the cascade control system make the acid temperature regulation time quite long. The robust and dynamic of the system is not very well, therefore, it restricted the potential of improving the striped steel's production in the future.As internal model control can compensate the delay time of the large delay control system, and greatly increasing the talent of system's anti-interference. Therefore, combining the internal model control with the cascade control system, through the least square method, we can fitting the inner model of generalized controlled plant, combined Maclaurin formula, the main controller is designed to internal model PID controller. Through the result of simulation, the robustness and dynamics of internal model control system is better than the cascade control system of the traditional PID arithmetic. It can decrease the regulation time of acid temperature, improving the speed of pickling.However, the internal model PID controller is based on the ideal condition which the controlled plant's parameter is stable and constant. But in the process of actual production, with the change of physical and chemical's operating model, the controlled plant's parameter which is working under the stable condition will also change, meanwhile the adjustable parameter of internal model PID controller will change. So this kind of controller isn't adaptability. With the development of intelligent control theory, we can combine the neural network with internal model control, take advantage of online learning's function of neural network to adjust the parameter of internal PID controller at any time. Make the neural network internal model PID controller adaptability. Through the result of simulation, the neural network internal model control system has the perfect robustness and dynamic, so that it can minimize the regulation time of acid temperature and largely improve the speed of pickling.