Study Of The Motor Fault Diagnosis System Based On Wavelet Analysis And Neural Network

The motor is one of the most important power-driven facilities in modern industry, not only does the automation technology for motor drive need improving but also the reliability, security and stability of running motor is quite important.In this paper, DC motor fault diagnosis is researched by monitoring its vibration signals. As the motor vibration signals are non-stable and random, the signal analysis based on traditional Fourier Transform can't completely meet the requirements for feature extraction. Also, the fault mathematical model of the motor is very complicated. The fault diagnosis method of motor based on wavelet analysis and neural network is suggested in this paper. Wavelet transform is used to extract the feature from vibration signal and the neural network is used to identify the feature and then outputs the running state of the motor.First, this method uses the technology of wavelet time-frequency for the noise cancellation and filtering of motor vibration signals, and strikes the energy of frequency band through the wavelet packet coefficients, and gains the fault characteristics from various changes in the energy of each frequency band, and the algorithm is realized with MATLAB software. Second, the vibration signal acquisition system and experiment scheme are designed for the diagnosis based on three typical running states of DC motor rotor system such as non-middle, axletree bump and normality. Finally, BP neural network is designed based on energy feature vector extracted with wavelet packet and three typical running state of DC motor, and the algorithm is realized with MATLAB software.The simulation experiment has been carried out to verify the validity and accuracy of the algorithm. The sample and testing vibration signals are collected separately. The off-line analysis of signals has been carried out to extract the feature with wavelet packet and the eigenvectors of sample and testing vibration signals have been calculated respectively. Then train BP network with sample eigenvectors, and the network has been tested with testing eigenvectors after the successful training. The testing results accord with corresponding states for actual testing signals. The results prove that the fault diagnosis system based on wavelet analysis and neural network suggested in this paper can effectively diagnose the faults of the DC motor and improve the fault diagnosis accuracy in the motor.