Motor Fault Diagnosis And Prediction Model Based On Fuzzy Neural Network

The motor is widely used in industry, national defense industry because of its characteristic of simple structure, good performance, easy to maintain. As the motor is frequently started, loaded unsteadily, it works in poor working conditions in application, motor fault is hard to avoid in the life cycle. In order to avoid the motor failure, ensure the safety in production, prevent the accidents and economic losses caused by the motor fault, the motor fault diagnosis and prediction are of great importance, it has been attracted the attention of the various related fields.The research work is based on above background in this paper. Blind Source Separation method is used to separate mixed motor vibration acceleration signals, then the separated signal is extracted main feature by using wavelet packet decomposition, reduction of the feature band and strengthen the characteristics are accomplished, the energy ratio is obtained which will be input vector for the fault diagnosis model. Then the structure design of the motor fault diagnosis model, algorithm design, programming implementation, model training and testing method are described, according to the principle of fuzzy comprehensive evaluation, a kind of motor fault fuzzy forecast module to evaluate and forecast performance of motor is set up.Research results show that the proposed fault feature extraction and enhanced method are effective, the fault diagnosis model structure can be flexibly composed according to the concrete application situation which will greatly simplify the difficulty of global convergence of network train, it improves the accuracy of the fault diagnosis and prediction to some extent.