Research Roller Bearing Online Fault Diagnosis Based On Wavelet Analysis And SVM

An important part of the mechanical drive system, antifriction bearing should be controlled over the operation status, which has very important significance to ensure machines run on the normal functioning of its performance. The bearing vibration signal is studied in this paper, and analyzed the signal processing method is suitable for vibration signal feature extraction. Then the feature information is used to extract the bearing fault detection and fault diagnosis. The main research of this paper includes Vibration signals in time domain statistical feature extraction method, frequency domain feature extraction methods, automatic detection and diagnosis method of the bearing failure based on support vector machine (SVM).Based on the fault detection and diagnosis process of the vibration signal includes signal acquisition, feature extraction and fault state recognition, wherein the feature extraction and status recognition are two of the most important part. Therefore, this paper studies the indicators of the time domain vibration signal, including RMS value, variance, crest factor, kurtosis, skewness, and sixth-order central moments, and then studies the frequency domain characteristics of vibration signal extraction method which mainly is the peak extraction of the bearing fault characteristic frequency.At present, the common method of the extraction of peak value of bearing fault characteristic frequency is resonance envelope demodulation. The problems of band pass filter parameter settings will be encountered in the course of using this method. Finding the formant signal spectrum by artificial means to determine with pass filter parameters greatly influenced by artificial factors and have limitations and randomness. Because of this, a method based on the spectral kurtosis to automatically determine the filter parameters is proposed in this paper, which can reduce the impact of human factors. In the course of studying the method, first filter the vibration signal with a finite impulse response, and then extract the envelope of the filtered signal and calculates the envelope to get the fault characteristic frequency power spectrum peaks. Then according to the characteristics fault signal is attenuated oscillation waveform, do the vibration signal wavelet transform with the application of Morlet wavelet instead of finite impulse response filter, and take the envelope to get envelope spectrum, which can achieve better results.In order to meet the automatic and intelligent requirement of the fault detection and diagnosis of antifriction bearings, the SVM method for the fault pattern recognition of the bearing is applied in this paper, and the expected recognition rate is achieved.