Study On Fault Diagnosis Method For Roller Bearing Based On Wavelet Analysis And BP Neural Network

Rolling bearing is an important part of rotating machinery, and one of the main sourcesof mechanical equipment faults, a lot of equipment faults is caused by the bearing every year,and it generates great economic losses, the rolling bearing have a direct impact on equipmentand the whole production process, thus the research on faults monitoring and diagnosistechnology of rolling bearing has an important theoretical value and practical significance.Rolling bearing faults monitoring and diagnosis is usually based on the collected sampleswith bearing vibration signal, through the analysis of the collected bearing signal, extractedsignal characteristic as the basis to determine bearing running, and judged the running state ofthe bearing through calculation and analysis. On the one hand, the collected signal of bearingcontains a lot of noise and nonlinear pulses, so it is difficult to distinguish bearing featureinformation which is drown on the noise; on the other hand, it is also crucial to select theappropriate method of fault diagnosis.On the basis of analyzing the basic principle of the diagnostic algorithms which is usedin the faults diagnosis, the experimental data of SKF6205-2RS bearing as sample, de-noisedthe fault signal of bearing using wavelet analysis and Empirical Mode Decomposition (EMD)method, the results show that the method can effectively restrain the noise and strengthen thecharacteristics of the signal. Using the characteristic of high resolution of wavelet packetanalysis, then the de-noising signal is decomposed into different frequency bands, andextracts the signal characteristics of each band as fault feature vector, the results show thatwavelet packet can extract the local characteristics of signals, the feature vector can representdifferent characteristics of signal effectively.In order to identify and determine the running state of bearing, using BP neural networkas bearing faults diagnosis network, choosing Particle Swarm Optimization (PSO) algorithmto optimize and improve BP neural network, using a combination of empirical formula andexperimental methods to determine the hidden layer nodes of the network, the results showthat PSO algorithm can effectively improve the convergence speed and improve the accuracyof the network of BP neural network, optimized neural network has good reliability.Establishing faults diagnosis system based on wavelet analysis and BP neural network,selecting the bearing data to test and verify the network, simulation results show that faultsdiagnosis system has a better ability to identify faults, which can effectively identify differentoperating state of the bearing.