A novel Pareto-based Bayesian approach on extension of the infogram for extracting repetitive transients. (June 2018)

• Record Type: Journal Article
• Authors: Gu, Xiaohui; Yang, Shaopu; Liu, Yongqiang; Hao, Rujiang
• Is Part Of: Mechanical systems and signal processing. Volume 106(2018)
• Page Start: 119
• Page End: 139
• Publication Date: 2018-06
• Subjects: Fault diagnosis -- Infogram -- Bayesian inference -- Pareto set -- Repetitive transient; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2017.12.034
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: Informative frequency band selection is defined as a multi-objective problem. A Pareto-based Bayesian approach as an enhanced extension of infogram is proposed. It can synthesize and balance the impulsive and cyclostationary signatures. Abstract: Two most important signatures of repetitive transients in the vibration signals of a faulty rotating machine are impulsiveness and cyclostationarity. In the newly proposed infogram, the time-domain and frequency-domain spectral negentropy were put forward to characterize these two aspects, respectively. However, in extension of the infogram to Bayesian inference based optimal wavelet filtering, only one spectral negentropy was employed in identifying the informative frequency band. To overcome its drawback, a novel Pareto-based Bayesian approach was proposed in this paper. The Pareto optimal solutions which can simultaneously maximize the time-domain and frequency-domain spectral negentropy were utilized in estimating the posterior wavelet parameters distributions. Moreover, the relationship between the impulsive and cyclostationary signatures was established by the domination. It can help balance the contributions due to these two aspects other than simply synthesize by the average weight in the infogram. Three instance studies including simulated and experimental signals were investigated to illustrate the effectiveness of the proposed method by challenging different noises and interferences. In addition, some comparisons with the aforementioned peer methods were also conducted to show its superiority and robustness in extracting the repetitive transients.