Fault Detection Of Six-Phase Permanent Magnet Synchronous Motor Drive System

With the development and improvement of modern technology and manufacture level, electric machines play a more important role in modern industrial production. The faults of electric machine drive system can destroy the machine itself and impact the normal operation of the system, furthermore harm the security of worker. In other words, one may avoid very costly expensive downtime of plant by proper scheduling of machine replacement or repair if warning of impeding failure can be obtained in advance.Firstly, this dissertation introduces a number of intelligent fault diagnosis methods,in particular, the application of wavelet analysis and neural network theory in the motor drive fault detection system. Wavelet transform as a time-frequency analysis method, which has the capacity of local features at time-frequency domain characterization of signals, through the window of time-frequency transform to highlight the flexibility of the different frequency components of signals. While the neural network represents a new approach to the system, which stored information in distribution, using the network topology and the right to realize the value of the distribution of non-linear mapping, and take advantage of the overall implementation of parallel processing from the input space to output space of non - linear transform information. Secondly, the choice of six-phase 20kW permanent magnet synchronous motor for the experimental prototype drive system to its working principle make a note, and from software and hardware of the drive system to conduct a comprehensive introduction for the fault signal extraction and the cause of the malfunction prepare job classification. Furthermore, the use of wavelet transform, by selecting the appropriate wavelet function and decomposition level on the motor stator current break down, eventually be able to extract fault features of the signal. Finally, choose a three-tier structure of BP network, the use of pattern recognition and sub-best technology on the fault signal is analyzed, and came to the reasons for failure happened.In addition, the fault signal in the analysis, the use of Matlab7.1 software, by selecting the DB4 wavelet function of stator current wavelet analysis, to obtain the corresponding coefficient of wavelet decomposition. Choose Appropriate neural network model, in the already normalized wavelet coefficients on the basis of the training by entering the sample finally completed the entire system fault detection. And to have been built to control the DSP2407 core experimental platform on the completion of the relevant basic experiments, the method using at fault detection effectiveness is verified.