Fault Diagnosis Of Motor Based On Wavelet And Neural Networks

Three-phase asynchronous motor plays an important role in the social production, their fault diagnosis is of great economic and social significance. The research status of domestic and foreign motor fault diagnosis is summarized, then a vibration-noise detection method for recognition of motor bearing fault is proposed in this paper. Finally, the simulation and diagnosis of induction motor's common fault is completed.For induction motor working principle, the fault mechanism and characteristics of bearing fault,rotor broken-bar fault,rotor imbalance, load fluctuations of the motor is analyzed systematically. In view of the motor fault signal with a large number of time-varying signals and sudden short-term impaction composition, Fourier transform is a pure frequency domain analysis method, applies only to stationary signal analysis, however, Wavelet analysis, its Time-frequency domain has a good localization characteristics, is suitable for detection of the normal entrainment of the transient signal anomalies and to demonstrate its components. So wavelet analysis is chosen to extract the signal characteristics of the motor fault in this paper.As an adaptive pattern recognition technology, neural network has a strong nonlinear mapping ability and has been applied increasing in the field of fault pattern recognition widely. The fault feature extracted by wavelet packet is input to the neural network, and the training, test and diagnosis is completed.The three-phase induction motor diagnosis test is completed, the vibration and noise data under normal and bearing fault condition is acquired, simple time-frequency domain analysis has been completed, then wavelet packet is used to extract the signals of different frequency bands as the energy-scale feature vector. We construct BP, RBF, Elman neural network to diagnose the fault, analyze three training and test results and determine the BP network for the ultimate in motor fault diagnosis system. Finally, a group of foreign motor fault diagnosis of experimental data is quoted and the verification and analysis of BP network is completed. The results show that the wavelet analysis combined with neural network approach for motor fault diagnosis is effective.Finally, three common faults of motor - rotor broken-bar fault,rotor imbalance, load fluctuations, is carried out in signal simulation, the decrease of the noise by wavelet packet is completed. Then, the use of wavelet packet decomposition, reconstructing, combined with power spectral methods complete the above diagnosis.