Research And Application On Multivariable Decoupling Control Based On BP Neural Network PID

With the development of the industrial control, control system is becoming more and morecomplex, the controlled plants of most industrial production are MIMO systems which are relatedwith each other, it is rather difficult to get an ideal decoupled effect by using common controllingmethod, so by means of multivariable coupling system, intelligent controlling method is applied torealize decoupled control. In order to get a satisfying outcome, neural network, is used to combinewith PID control in different ways in this paper, which has strong ability of self-adapting andself-learning, therefore it can solve the nonlinear problems of multivariable coupling system.The object researched in this paper is multivariable coupled system. Firstly, diagonal-matrixdecoupling control is studied, and decoupling principle of Diagonal Matrix is analyzed, thesimulation model is constructed by Simulink to get the simulation curve; Secondly, PID neuralnetwork is studied and a multivariable decoupled controller is set up. The system, which isprogrammed, is simulated by Matlab and attains good effect; finally, multivariable decouplingcontroller is built based on BP neural network, which is simulated by Matlab and the result issatifying. Compare the above three kinds of simulation curve, result turns out the curve got fromBP neural network is better than the other two ones.This study is based on process control system of Key laboratory of Shenyang University,SIMATIC NET industrial network, softwares and hardwares of SIEMENS are used as applicationplatform, the coupling system of liquid and temperature are application objects. The mathematicalmodel of this coupling system is calculated by means of step response curve, it will be used as thesimulation model of the three methods mentioned above, and BP neural network multivariablecontroller is implemented on PLC equipments to achieve decoupling control. The experimentalresult indicates that BP Neural Network PID multivariable control algorithm can control practicalsystem effectively.Form the simulation and experimental result we can see that the effect of BP neural networkmultivariable coupling is not only better than the other two methods, but also has a good practicaleffect, which owns a certain applicative value.