Applications Of Electromechanical Equipment Fault Diagnosis Based On Dual-tree Complex Wavelet

To ensure the electromechanical equipment run effectively, credibly and safely,fault diagnosis technique is the indispensable part of modern industry, which canmonitor equipment operation conditions. The extraction of fault feature information isthe key to the fault diagnosis technique. The working environment ofelectromechanical equipment is complicated, and the fault vibration signal is usuallynonlinear and nonstationary. At the same time interference of strong noise iscontained in the vibration signals. So it is important to extract the fault featureinformation through the effective method of signal processing. As a powerful tool fornonlinear and nonstationary signal, wavelet transform has been widely used in thefield of fault diagnosis. However, the traditional wavelet transform has manyshortcomings. For example, orthogonal discrete wavelet transform (DWT) lacks shiftinvariance. Complex wavelet transform cannot reconstruct perfectly. Stationarywavelet transform brings redundancy and other wavelet methods have insufficiency insome ways. Dual-tree complex wavelet transform (DT-CWT) can overcome theseshortcomings by contrast. In this paper, DT-CWT is deeply investigated andsuccessfully applied to the electromechanical equipment fault diagnosis. The maincontents are as follows:(1) The basic principle of DT-CWT is described. The characteristics of theapproximate shift invariance and the smaller the frequency aliasing about DT-CWTare verified. A new fault diagnosis method is proposed based on DT-CWPT andthreshold de-noising. The simulation signal and rolling bearing fault vibration signalsare processed by DT-CWPT and the traditional discrete wavelet packet transformationrespectively. Compared to the traditional discrete wavelet packet transformation,DT-CWPT has an advantage.(2) It is difficult to set threshold accurately, so a new fault diagnosis method isproposed based on DT-CWT and singular value decomposition (SVD).The number ofeffective singular value is determined automatically by the singular value differencespectrum and curvature. Noise contained in signal is eliminated as far as possible, andthe effective information of signal is retained at the same time. The fault characteristicinformation is extracted effectively from fault signal of electromechanical equipment.(3) When the signal is decomposed by DT-CWPT, the layer of decompositionand the component signal of decomposition should be chosen reasonably. A new fault diagnosis method is proposed based on DT-CWPT and spectral kurtosis(DT-CWPT-SK). The layer of decomposition and the component about the signal canbe chosen automatically and accurately.(4) A new fault diagnosis method is proposed based on DT-CWPT and supportvector machine (SVM), the high recognition rate of single fault about rolling bearingcan be obtained. Compared to the single fault of rolling bearing, the compound faultrecognition accuracy is not ideal. So a new fault diagnosis method is proposed basedon DT-CWPT and independent component analysis (ICA), and the independentcomponent signal of ICA is de-noised by AR spectrum at same time. The fault featureof rolling bearing can be separated effectively and the fault feature are extracted.