A Feature Extraction Method for Aircraft Engine Rotor Vibration Diagnosis. (2015)

• Record Type: Journal Article
• Authors: Zhang, Cui; Wang, Keming; Zhao, Pengran
• Is Part Of: Procedia engineering. Volume 99(2015)
• Page Start: 1576
• Page End: 1581
• Publication Date: 2015
• Subjects: aeroengine -- vibration analysis -- empirical mode decomposition -- information entropy -- feature vector -- fault diagnosis; Engineering -- Congresses; Engineering -- Periodicals; Engineering; Conference proceedings; Periodicals; 620.005
• DOI: 10.1016/j.proeng.2014.12.709
• Languages: English
• ISSNs: 1877-7058

Abstract

Abstract: Based on study of aeroengine vibration mechanism and analysis of characteristics of vibration signals corresponding torotor faults, empirical mode decomposition method is used to decompose the vibration signals measured on engine cases. It is used in time-frequency domain to extract vibration features, because time-frequency properties of vibration signals can reveal rotor faults more effectively. The rotor vibration signals are first compared with the energy state of intrinsic mode function of EMD and its relevance to the original signal. Then, on the basis of determination of the main fault information included by IMF, the three information entropies are evaluated in the time domain, frequency domain and time-frequency domain respectively. Finally, the feature vector for rotor fault diagnosis is composed of the three information entropy values calculated from each IMF and the spectral entropy of wavelet packet space characteristics obtained by wavelet packet decomposition. The results show that empirical mode decomposition method based on time-frequency analysis can extract feature vectors of the non-stationary fault signals effectively. This provides a systematic method of quantitative feature selection for aeroengine rotor fault diagnosis through vibration analysis.