Extraction Of Fault Features Of Rolling Bearing Based On Vibration Signal

Rolling bearing is the most widely used in all kinds of rotating machinery is an important component, its running state directly affects the performance of the whole machine,fault diagnosis for it has important significance. Fault diagnosis of rotating machinery including the signal measurement, feature extraction, state diagnosis and state analysis in four steps, in which the core part is the feature extraction of fault signal.The rolling bearing as the research object in this paper, the method of bearing signal feature extraction is studied.This paper introduces the mechanism of the vibration of rolling bearings, on several typical fault types and formation reasons was analyzed.At present, temperature detection, oil detection method, vibration analysis method and acoustic emission method were used fault diagnosis methods of rolling bearing, the vibration analysis method was used in this paper.First studied the vibration signal of time domain analysis method, including the extraction method of dimensionless parameters and non-dimensionless parameters characteristics.By changing the time domain statistical parameters compared with normal signals and fault signal value, get time domain signal feature.Secondly, the frequency domain analysis method of bearing signal is studied, the method of using Hilbert demodulation characteristics of rolling bearing are extracted of rolling bearing. In the actual rolling bearing fault,Generally have periodic pulse signal is generated, so that the fault signal modulation phenomenon, on both sides of the natural frequency of the presented evenly spaced modulation side bands in spectrum. The analysis method of demodulation, can extract the modulation information from fault signal, but this method need to know the frequency of bearing fault.Again, the analysis method of time domain is studied, the influence of noise in the process of bearing signal acquisition, based on the correlation of the wavelet singular entropy feature extraction method of the rolling bearing was proposed in this paper. This method firstly to sampled signal wavelet transform, get the wavelet coefficients. Use of signal and noise wavelet transform on different scales have different propagation characteristics, the wavelet coefficients to do relevant calculations, in order to achieve the aim of separation of noise. The processed coefficient was decomposed by singular value, singular value was calculated by information entropy, get the correlation of the wavelet singular entropy.Through the research on simulation of bearing signal, the method is proven to have the effect of restraining the noise. The method is used as feature vector bearing signal, using probabilistic neural network for fault classification, verify the validity of the method of feature extraction.Finally, because the EMD decomposition is adaptive, suitable for dealing with nonlinear,non-stationary signal, based on EMD instantaneous power spectral entropy feature extraction method of the rolling bearing was proposed in this paper. The vibration signal is decomposed by EMD, get several IMF components, these components is analyzed by power spectrum,calculating the information entropy of the power spectrum. The method is used as feature vector bearing signal, EMD instantaneous power spectral entropy as the feature vector, using probabilistic neural network for fault classification, verify the validity of the method of feature extraction.