Research On Several Methods Of Fault Diagnosis For Non-stationary Signals Of Rolling Element Bearing

Machine condition monitoring and fault diagnosis are of great significance in guaranteeing equipment operation security and reducing casualties and economic loss. Our research focuses on feature extraction of bearing based on vibration and acoustic signal analysis. Different methods are proposed to deal with the acquired signal, meanwhile the merits of these methods are assessed, which provides certain solutions for condition monitoring and fault diagnosis of bearing.First, we introduce main failure forms and fault characteristic frequency of bearing, analyze the vibration mechanism, and reveal the relationship between the acoustic signal and vibration signal. Respective experiments aiming at vibration and acoustic fault diagnosis of bearing are conducted, which enables the research to be proceeded smoothly. The vibration signal is collected by acceleration sensor, while the wayside acoustic signal is acquired by microphone. In view of the characteristics of wayside acoustic signal, the experiment scheme of static collection and dynamic transmission is proposed. The acquired signals provide the base for the improved algorithm in this thesis.Afterwards, two strategies based on vibration fault diagnosis are presented. The first one refers to adaptive variational mode decomposition based on artificial fish swarm algorithm. The parameters are optimized so that we could acquire the optimal output of variational mode decomposition. The second strategy is based on component screening singular value decomposition. The threshold is introduced to choose the component signals after utilizing the difference spectrum of singular value decomposition, which leads to the rise of signal to noise ratio. Both simulated and experimental results show the effectiveness of the two proposed methods in feature extraction and fault diagnosis of bearing under noisy environment.Finally, in consideration of steep distortion and strong noise problem, the combination of time interpolation resampling and maximum correlated kurtosis deconvolution is proposed in this thesis. The acquired noisy signal with Doppler effect is first processed by time interpolation resampling method. Then the weak defective information is enhanced by an adaptive maximum correlated kurtosis deconvolution method. The proposed method takes advantages of signal resampling for Doppler effect elimination and information enhancement for noise treatment, thus is expected to recover the inherent characteristics. The proposed method is verified to be effective and feasible by simulated and experimental signals in wayside defective bearing detection.