Research On Intelligent Optimal Control Of Hot Strip Rolling Process

Strip steel plays a very important role in modern life and is widely used in various light and heavy industries.With the rapid development of various industries,the corresponding requirements for strip rolling accuracy are increasing.The thickness of strip steel is an important mark to measure whether the plate is qualified or not.Whether the Strip thickness reaches the standard or not depends on the control effect of the rolling mill to a great extent.Most of the existing rolling mill control is based on PID deviation control.The traditional PID control is widely used in the actual factory production,especially in the process control field,because of its simple operation,excellent robustness and high Operability.In the PID Controller,the control parameters of the controller have a decisive effect on the control performance.Under normal circumstances,the parameter tuning of PID control is complicated,and the tuning methods of the controlled object are different.In the field operation,the tuning of the parameters of the controller depends on the expert experience in many cases.In the actual rolling process,there are a lot of interferences because of the changeable production environment.Because the rolling process is a system with time-varying delay,it is difficult to control the rolling process accurately by traditional PID,and the control parameters of the controller should be adjusted adaptively according to the optimal state of the controlled system,however,the control parameters of traditional PID controller can not be changed after it is determined.Therefore,the traditional PID controller can not completely achieve the control effect for the system which has high control requirements.With the rapid development of artificial intelligence technology in today's society,the technicians have realized the parameter self-adaptation of the PID controller by combining the artificial intelligence algorithm with the traditional PID control.In order to improve the quality of rolled products.The 2250 hot strip rolling line of a steel plant is a nonlinear system with variable time delay and strong coupling.The traditional PID control has some problems such as poor anti-interference and lack of real-time control ability when dealing with this kind of nonlinear system.A design method of PID controller parameters optimization based on reinforcement learning is proposed.The basic principle of PID control and the basic idea of reinforcement learning are expounded systematically.The actor-critical structure is introduced into the reinforcement learning,and then the design of the network structure,the number of layers and nodes of the neural network are selected,the activation functions are selected and the parameters are updated,the actor-critical structure and RBF neural network adaptive PID control strategy are analyzed in detail.The idea of combining reinforcement learning with traditional PID is proposed to optimize the control algorithm of PID parameters.Compared with traditional PID and PSO,the better PID controller is obtained.The experimental results show that the controller can quickly return to the steady state in the presence of external disturbances and respond rapidly when the time delay changes.Compared with the traditional PID controller and PSO,the optimized controller has the advantages of small overshoot,fast response time,good robustness and self-adaptation.