A comparative study between Empirical Wavelet Transforms and Empirical Mode Decomposition Methods: Application to bearing defect diagnosis. (15th December 2016)

• Record Type: Journal Article
• Authors: Kedadouche, M.; Thomas, M.; Tahan, A.
• Is Part Of: Mechanical systems and signal processing. Volume 81(2016)
• Page Start: 88
• Page End: 107
• Publication Date: 2016-12-15
• Subjects: Bearing fault -- Acoustic emission -- Empirical Mode Decomposition (EMD) -- Ensemble Empirical Mode Decomposition (EEMD) -- Empirical Wavelet Transform (EWT); Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2016.02.049
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: The Ensemble Empirical Mode Decomposition (EEMD) is a noise assisted method that may sometimes provide a significant improvement on Empirical Mode Decomposition (EMD). However, the amplitude and number of added noise need to be selected when applying the EEMD method. Furthermore, the computation time which depends on the number of ensemble trails is very high compared to the EMD. In this paper, a new way for choosing the appropriate added noise is presented. Conversely, a recently-developed method called the Empirical Wavelet Transform (EWT) is investigated. A comparative study between the EMD and EWT methods is conducted. The results show that the EWT is better than the EEMD and EMD on mode estimates and computation time is significantly reduced. An experimental study on bearing diagnosis is conducted. The EWT is applied to experimental data coming from damaged bearings. In the paper, an index selection is introduced that allows for the automatic selection of the Intrinsic Mode Functions (IMF) that should be used to perform the envelope spectrum. It is shown that choosing all the IMF selected by the index is more efficient than only choosing the best one. The envelope of the sum of the selected IMF clearly reveals the bearing frequencies and its harmonics which are excited by the defect. This approach seems to be an effective and efficient method for processing bearing fault signals. Highlights: A new method called EWT is presented to decompose the signal on a multiple components. The EWT is better than the EEMD and EMD on modes estimates and computation time is significantly reduced. A new way for choosing the appropriate added noise for EEMD is presented. A new way for automatic selecting (Index_selection) of the IMF is introduced. It is shown that choosing all the IMF selected by the index is more efficient than only choosing the best one. Allows early detection of defect.