Research On Fault Diagnosis Of Motor Based On Probabilistic Neural Network

Asynchronous motor as the driving force of modern industry, to the people’s daily life and social production has brought great help. With the rapid development of modern industrial technology and science and technology, the number and type of motor are increasing, its structure and properties also become increasingly complex. Normal operation of the motor to ensure its efficient production process quality, safe and reliable operation characteristics, improve the social and economic benefits. Motor running in an abnormal state, not only damage the structure and properties of the motor itself, will cause a huge impact on manufacturing and personal safety, and can cause huge economic losses. Therefore, the common motor faults occur during operation, the use of advanced science and technology for diagnosis and treatment, there are very important practical significance. However, with the development of intelligent control technology, neural networks, fuzzy control, artificial intelligence systems in fault diagnosis of the motor have been extensive practice and application. This paper uses neural network algorithm design motor common fault diagnosis system to improve the efficiency of fault diagnosis.This paper studies the common fault diagnosis neural network under asynchronous motor, by mastering the structure and mechanism of network failure, network model BP network were established and PNN network, and make a diagnosis based on comparing the two. The main target is Y801-4 asynchronous motor, by building test platform, collecting data signal, extract fault characteristic frequency, and verify the feasibility of BP network PNN network fault diagnosis system.First, the short-circuit fault between the stator and the rotor turns the bars fault, failure, and the inner eccentric bearing fault a detailed analysis, the main analysis of the causes and mechanisms of these four faults generated by the motor, the motor and vibration caused by and the change of the stator current.Secondly, the principle of BP network in-depth understanding, using MATLAB software development, neural network algorithm, by constantly changing training and training function parameters to achieve the best efficiency of fault diagnosis.Then, grasp the PNN neural network based principle and network structure, the motor appears four common fault diagnostic analysis and comparison of BP network training results to determine the diagnostic method has a simple structure, fast training, high stability.Finally, the failure to build a motor test platform, and set artificially broken rotor bar fault and rotor eccentricity fault. Vibration motor running signal acquisition and fault runtime and flow through the motor stator current signal. The vibration signal in time domain waveforms and frequency domain waveforms simple analysis and judgment. According to the failure mechanism of the motor, the stator current signals to extract the fault characteristic frequency as the input sample neural network, a test train in the MATLAB platform, finally achieved satisfactory diagnosis.