Automated Gearbox Fault Diagnosis Using Entropy-Based Features in Flexible Analytic Wavelet Transform (FAWT) Domain

Abstract

Purpose: Gearboxes are important for the mechanical power transmission in rotary systems. Gearbox failure may lead to an increase in downtime and production loss. Hence, effective and reliable working gearboxes are needed for regular health monitoring and controlling of the excessive vibration of the system. The purpose of this work is to use a vibration-based technique to automate the bevel gear wear fault diagnosis. It is thus expected that our novel systematic and procedural analysis would help to accurately identify multi-class gearbox faults. Methods: In this study, a flexible analytic wavelet transform method was used to decompose the bevel gear wear signal into sub-band signals. Various entropies, such as cross-correntropy, log energy entropy, Stein’s unbiased risk estimate entropy, Shannon entropy, norm entropy, and threshold entropy were used for feature extraction from all of the sub-band signals. The Kruskal–Wallis test was also used to obtain statistically meaningful results. Subsequently, these quantitative features were fed to the Least-Squares Support Vector Machine (LS-SVM) classifier. Results and Conclusions: These methodologies are found to produce the most accurate results by using the log energy entropy-based multi-class LS-SVM classifier and the RBF kernel function. The results obtained here are compared with the previous results obtained by different methods, such as the continuous wavelet transform (CWT), discrete wavelet transform (DWT), wavelet packet transform (WPT), dual-tree complex wavelet transform (DTCWT), and tunable-Q wavelet transform (TQWT). © 2021, Krishtel eMaging Solutions Private Limited.