Transient signal analysis based on Levenberg–Marquardt method for fault feature extraction of rotating machines. (March 2015)

• Record Type: Journal Article
• Authors: Wang, Shibin; Cai, Gaigai; Zhu, Zhongkui; Huang, Weiguo; Zhang, Xingwu
• Is Part Of: Mechanical systems and signal processing. Volume 54/55(2015)
• Page Start: 16
• Page End: 40
• Publication Date: 2015-03
• Subjects: Fault diagnosis -- Feature extraction -- Levenberg–Marquardt method -- Parameter identification -- Transient modeling; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2014.09.010
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: Localized faults in rotating machines tend to result in shocks and thus excite transient components in vibration signals. An iterative extraction method is proposed for transient signal analysis based on transient modeling and parameter identification through Levenberg–Marquardt (LM) method, and eventually for fault feature extraction. For each iteration, a double-side asymmetric transient model is firstly built based on parametric Morlet wavelet, and then the LM method is introduced to identify the parameters of the model. With the implementation of the iterative procedure, transients are extracted from vibration signals one by one, and Wigner–Ville Distribution is applied to obtain time–frequency representation with satisfactory energy concentration but without cross-term. A simulation signal is used to test the performance of the proposed method in transient extraction, and the comparison study shows that the proposed method outperforms ensemble empirical mode decomposition and spectral kurtosis in extracting transient feature. Finally, the effectiveness of the proposed method is verified by the applications in transient analysis for bearing and gear fault feature extraction. Highlights: Double-side asymmetric transient model is built based on parametric Morlet wavelet. The Levenberg–Marquardt method is used to identify the model parameters. An iterative extraction technique is proposed for transient signal analysis. The performance is verified by a simulation signal and some practical applications. The comparison study shows that the proposed method outperforms EEMD and SK.