Multidimensional feature extraction based on vibration signals of rolling bearings
暂无分享,去创建一个
In order to reduce the wrong fault diagnosis probability of rolling bearing using single feature, this paper, being aimed at detecting three working states, which are normal running state, rolling element damage fault, and outer-race damage fault, uses two methods, Shannon entropy computation of wavelet coefficients, and wavelet packet combined with bi-spectrum analysis, to extract multidimensional features from bearing vibration signals. A group of feature variables totally reflecting the different working states of rolling bearing are searched for easy and effective fault diagnosis. Shannon entropy can show the inner uncertainty of system, and bi-spectrum analysis can reflect the nonlinear and non-Gaussian features of system. Simulation shows that combination of two methods can distinguish three working states more effectively, which helps to implement accurate fault diagnosis in time.
[1] Nitish V. Thakor,et al. Wavelet Entropy for Subband Segmentation of EEG During Injury and Recovery , 2003, Annals of Biomedical Engineering.
[2] Kuihe Yang,et al. Application of Wavelet Packet Analysis and Improved LSSVM on Rotating Machinery Fault Diagnosis , 2008, 2008 Workshop on Power Electronics and Intelligent Transportation System.
[3] Osvaldo A. Rosso,et al. Alterations of thalassemic erythrocytes detected by wavelet entropy , 2007 .