Application Research Of BP Neural Network PID Control In Gas Evolution Tester

Gas evolution tester is an important equipment of foundry technology, and thetemperature dynamic-characteristic has directly influenced the quality of production.Electric heater is an object featuring in nonlinear, time variability, large time lag andasymmetry, general PID controller can not obtain high control precision, parameter isn’teasily adjusted and badly adaptive ability. In order to improve the system’s adaptive abilityand anti-jamming capability, this paper propose BP neural network PID control algorithmto realize precise temperature control of gas evolution tester.Intelligent PID control includes expert-type smart self-tuning PID control, fuzzy PIDcontrol and neural network PID control，which have characteristics of self-learning, selfadaptive and self-tuning parameters online. However, the expert knowledge base ofexpert-type PID control is not easy to be established, the fuzzy rules of fuzzy PID controlis not easy to be determined and the control accuracy is not high. With comprehensivecomparison of several intelligent control algorithm, the neural network PID controlalgorithm was chosen in this thesis.Artificial Neural Network is a simple units(neurons) for the node, using a networktopology consisting of active network and can be used to describe virtually any nonlinearsystem. General PID controller has a simple algorithm, easy to adjust, etc. Artificial neuralnetwork with a combination of traditional PID control to constitute intelligent PIDcontroller, it can auto-tuning controller parameters, adapt to changes in controlled processparameters. It can easy to solve the online real-time tuning parameters, the process isdifficult for some complicated and slow time-varying system parameters for the lack ofeffective control.Firstly, this paper describes the common Gas evolution tester and development trendof domestic and foreign. Secondly, this paper introduced the PID control theory andcompare the various algorithms and intelligent control algorithm advantages anddisadvantages. Thirdly, this paper describes the artificial neural network PID controltheory and software design.Finally, this paper provides real-time temperature control curves. The results show that BP neural network PID control temperature control curve of the overshoot is only1℃, the temperature curve without oscillation, no static error. And in half an hour thetemperature had stabilized to meet the accuracy requirements, gas experiment can be made.Neural PID control achieve the desired control effect.