Study On Vibration Fault Diagnosis Of Asynchronous Motors Based On Wavelet And Neural Network

Asynchronous Motors is most used in the productions and lives. It have more advantage, such as simple structure, low cost, high reliability, long lifetime and convenient maintain. It is extremely significant that motors run normally because this will ensure manufacture process safe and effective with higher quality and lower consumption. It can not only damage motors itself but also affect the normal work of the whole system and even endangers people if motors didn't work and were in faults. Furthermore, this will result in huge economic loss. Thus it is urgent for diagnosis and the warning of fault in early stages.This article has elaborated the fault mechanism of motor. Author has also analyzed the reason which results in breakdown and concluded frequency performance of several kinds of common faults. Through experiment, we obtain vibration signals and the feature of wavelet packet energy. It makes wavelet analysis become a pre-processor of Neural Network. Finally, it created a fault diagnosis system based on wavelet transform and Neural Network.As the fault signals are non-stationary transient ones, the traditional signal analysis methods, such as Fourier Transform, are not so efficient and useful for the fault signal detection. However, Wavelet Analysis has the excellent time-frequency local performance; it can detect the different frequency components of the fault signals by its adjustable time-frequency window. This paper has proved that Wavelet Analysis is efficient and useful in the fault diagnose area by compared the results which use Fourier Transform and Wavelet Analysis. And as Wavelet packet Transform has transfer feature to useful signals and restraint feature to noise, just as a band-pass filter, this paper has proved that Wavelet packet Transform is most suitable for removing the noise form non-stationary and stationary signals by compared the results which use Wavelet Transform and Wavelet packet Transform. Beacuse Wavelet packet Transform has time-frequency feature and multi-resolution feature, not only the whole signal but the partial signal can be analyzed, the fault feature of non-stationary transient signals can be caught correctly. So Wavelet packet is always used in the extract the feature of fault.When motors are in faults, its amplitude and the vibration form and the frequency component are vary form normal stages. Different fault can cause different vibration form, so, the vibration signals can exactly reflect the fault information. This paper carefully analyze the vibration signals of motor which decomposed into six levels using wavelet packet, we get the feature energy of each frequency-band , which will input the Neural Network.The paper choose six frequency bands which have close connection with fault and simulate three fault, BP Neural Network will be created by determine input and output units and suitable cryptic level and parameter. then it should be trained, when the reach Precision that we need, we can input the specimen data to diagnose fault and input the test data to test Net. The results of diagnose Proved that Wavelet and Neural Network is correct and precise, it makes motor fault diagnosis more intellectualized.