Study On The Fault Diagnosis Of Variable Frequency Speed Regulation Based On Fuzzy Neural Network

The structure of the variable frequency speed regulation system is complicated, and it doesn't exist distinguish features between the fault symptoms and causes. According to this, a fault diagnosis method based on the fuzzy neural network (FNN) is used.The method also combines the features of the fuzzy system and neural network.The fuzzy system lacks of self-study abilities, its membership functions and fuzzy rules are chosen by experts subjectivity, and there was no clear physical significance between input and output of the neural network, so a fault diagnosis method which combines the fuzzy system and the neural network is used. An improved BP algorithm of increasing the momentum or adjusting the adaptive learning rate which based on the traditional BP algorithm is also proposed. The fault mechanisms of the variable frequency speed regulation system are researched. To analysis and extract the characteristics of the output signal of the variable frequency speed system, The Fourier transform and wavelet analysis method is used.The absolute value of skewness, kurtosis and root mean square of the output current of inverter are used as the input of the FNN. And the vibration band energy and deviation in current, voltage and vibration are used respectively as the input of FNN to diagnosis the motor fault. The results show that the fault diagnosis method which based on the FNN has a higher accurative rate comparing with the general diagnosis method, and it also can compensate the disadvantages of the fuzzy system and the neural network using singly. The errors of fault diagnosing which based on the deviation in current, voltage and vibration is smaller than it only based on vibration band energy. Therefore, in the variable frequency system fault diagnosis, the diagnosis method based on FNN is a useful diagnostic method.