Simulation Research On Cable Diameter Control Of Polymer Extruder Based On Reinforcement Learning

Polymer extrusion is a complex process.There is not only a strong coupling phenomenon between process parameters,but also a large hysteresis problem in the control process because of the existence of a certain distance between production and measurement.For the hysteresis problem of polymer extruder,this paper firstly introduces and analyzes the process flow of the polymer extruder,analyzes and simplifies the mathematical relationship between the parameter,and establish the mathematical model of the polymer extruder.At the same time,the boundedness of the solution of the mathematical model and the stability of the linear system are analyzed.Finally,the input and output relationship of the polymer extruder was obtained and then simulated.In terms of control strategy,this paper combines the advantages of dynamic programming method and temporary difference method,and proposes a control method based on step-by-step updating and model-free.This method mainly uses the more advanced Sarsa(?)algorithm of temporary difference method as the core control algorithm of polymer extruder to precisely control the cable diameter in the extrusion process;In the prediction part,this paper analyses the importance of sample model to improve the control effect of polymer extruder,and designs a suitable fitting method to solve the parameters of sample model efficiently,so as to predict the cable diameter after the delay time in real time.Finally,combined with control and prediction to form a polymer extruder controller marked SM-Sarsa(?),with two important processes: one is the decision process with Sarsa(?)as the core,and the other one is the fitting process with Sherman Morrison as the key technique.The decision process can provide inputoutput data for fitting process to get a prediction model,in order to track the transfer function of the system.The fitting process provides the predicted value for the decision process through the prediction model,so as to carry out real-time control of equivalent no delay.In order to prove the superiority of SM-Sarsa(?),This paper also introduces and designs traditional controllers,including PID,fuzzy PID and dynamic matrix,and then the SM-Sarsa(?)was simulated and compared with these traditional controllers under different lag time.The results show that the control effect of SM-Sarsa(?)is better than that of these traditional controllers at the expense of certain training time,especially in indicators such as overshoot,response speed,and time range of qualified wire diameters.The cable insulation extruder with this controller has the advantages of faster response speed and convergence rate,and almost no overshoot.It is easy to deal with control systems with unknown internal environment and hysteresis problem,which can be extended to similar control systems with these features.