Research On Control Of The High-frequency Induction Heating Power Based On Fuzzy RBF Neural Network

Induction heating is a kind of high efficiency, high speed, low energy consumption andenvironmental protection way of heating in the existing heating mode. Using inductionheating technology of high frequency induction heating power supply can be heated forartifacts as a whole, and can be targeted for local heating, in turn, to deep diathermy artifacts,or to heat work-piece surfaces. And the heating work-piece objects of high frequencyinduction heating power are not limited to metal materials, non-metallic materials can also beindirect heating, this greatly increases the range of application of high frequency inductionheating power supply.High frequency induction heating power supply is a kind of variable voltage variablefrequency device in essence. It is composed of the rectifier circuit, inverter circuit, loadcircuit and control circuit mainly. The control circuit is the main research contents in thispaper, because, load parameter of high frequency induction heating power supply will changeas heating material and quantity of the work-piece, the specific changes are involvingmagnetic field, electric field, heat transfer and other related physical process, there are a lot ofinfluence factors, the precise mathematical model is not set up in the current research level,the application of traditional PID control can’t achieve precise control to meet the needs ofthe growing industrial development. So in this paper, the fuzzy RBF neural network will beintroduced into the traditional PID control of high frequency induction heating power supply,it will realize online adjustment of PID parameters to achieve precise control effect.According to the actual circuit structure of high frequency induction heating power supplythis paper establish a fuzzy RBF neural network, the network structure parameters of theinitial value is not given randomly during the training of fuzzy RBF neural network, thetraining data is classified by K-means hierarchical clustering, the single value resulting fromthe clustering as the initial value of network structure parameters, and by this way, theconvergence speed of the error is faster in neural network training, and will not converge tolocal minima. At the same time, in view of the high frequency induction heating powersupply in high frequency switching loss problem under the working state, based on the phaseshift PWM inverter in this paper, using nonlinear inductance component in the inverter circuit,to make the inverter circuit switch tube work in soft switch state.This paper studies the fuzzy RBF network adjustment of PID parameters mainly, usingMATLAB simulation software, and completing the simulation of high frequency inductionheating power supply system. By comparing the simulation diagram, we can see the controlmethod can better realize the control of the supply power in high frequency induction heatingpower in this paper, in order to ensure that the output is stable.