A novel feature extraction method based on weighted multi-scale fluctuation based dispersion entropy and its application to the condition monitoring of rotary machines

• Published in: Mechanical systems and signal processing 2022-05, Vol.171, p.108909, Article 108909
• Main Authors: Sharma, Snehsheel, Tiwari, S.K.
• Format: Article
• Language: English
• Subjects: Chipping; Classifiers; Empirical analysis; Ensemble empirical mode decomposition; Entropy; Feature extraction; Flexible analytical wavelet transform; Gear trains; Gearboxes; Machinery condition monitoring; Multilayer perceptron; Multilayer perceptrons; Planetary gearbox; Rotary machines; Support vector machines; Wavelet analysis; Wavelet transforms; White Gaussian noise
• ISSN: 0888-3270; 1096-1216
• DOI: 10.1016/j.ymssp.2022.108909

Features describing the state of industrial gearboxes and their extraction from the mixed noisy signal are always an issue of concern. Unfortunately, traditional feature extraction methods are not equally useful for both adaptive and non-adaptive decomposition techniques. In the present study, a novel feature extraction method named weighted multi-scale fluctuation-based dispersion entropy (wtMFDE) is proposed for the condition monitoring of planetary gearbox (PGB). It is equally useful for adaptive (ensemble empirical mode decomposition) and non-adaptive (flexible analytical wavelet transform) signal denoising techniques. Four state-of-the-art classifiers viz. support vector machine (SVM), random forest (RF), K-nearest neighbor (KNN), and multilayer perceptron (MLP), which have extensive applications in machine condition monitoring, are used for the classification of the chipped sun gear, pitted planet gear as well as the combined effect for these faults. It is observed that the extracted features are highly sensitive and effective in a way that they can recognize faults with comparable accuracy with all the selected classifiers. •A novel feature extraction method is developed by associating the weighted parameters with the multi-scale fluctuation based dispersion entropy (MFDE).•Method is equally effective for both adaptive and non-adaptive techniques.•Experimentally, chipped sun, pitted planet gear faults, and their combinations are investigated.