Research On The Controlled Temperature Of Induction Heating Power Based On The Fuzzy Neural Network

With advancing levels of technology, greater precision in controllingthe temperature of workpiece heated by induction heating power is required.The control method for induction heating powers is crucial to determine thetemperature precision for the heated workpiece, therefore, the research ofcontrol methods is of great value. In this thesis, using fuzzy control and neuralnetwork theory, a Proportion and Integral (PI) control method based on afuzzy-neural network is studied, and a model of an induction heating power isbuilt. The experiment results show that the control method proposed in thispaper gives promising results in improving the temperature precision for theworkpiece heated by induction heating power.Firstly, a 150kW/30kHZ parallel induction heating power and itsmathematical model are investigated. When a general PI control method isused in an induction heating power, the varying temperature of the heatedworkpiece's surface is plotted;the disadvantage of the general PI method canbe found based on the analysis of the given temperature curve.Secondly, a fuzzy-neural network PI control method is proposed on thebasis of the analysis of general PI methods. Then a Back Propagation (BP)algorithm is introduced, and by employing the structure of neural network,fuzzy control and neural network are combined in such a way that the fuzzymembership function and fuzzy rules from the traditional expert system areconverted into the neural network. As a result, a distributed knowledge systemis constructed. Due to its self-learning ability, this system can update theconnection weights in neural networks continuously, and consequently makethe fuzzy membership function and fuzzy rules more accurate. According tothis distributed knowledge system, the PI parameters for an induction heatingpower can be adjusted properly when the workpiece's temperature varies, andfinally the expected temperature curve can be acquired.In the end, a visualized software package is developed under theenvironment of Visual C++ 6.0. This system can perform training andsimulation of the induction heating power based on the fuzzy-neural networkPI control method. A fuzzy-neural network PI controller can be trained off-lineby this system, and the performances of the fuzzy-neural network PI controllerand general PI controller can be observed. Finally, the advantages of newcontrol method can be shown by analysis and comparison with the simulationresults.