MODWT and VMD Based Intelligent Gearbox Early Stage Fault Detection Approach

Gearbox, a crucial constituent of any plant machinery, always requires special attention as it has to perform under considerable environmental conditions throughout its service life. Hence, the tracking of gearbox performance degradation is paramount to ensure the reliability and availability of the whole system. This performance degradation assessment is often based on the vibration-based condition monitoring program which extracts the fault signatures from the raw vibration signals. Then, based on the previous known values from ISO standards a comparative analysis is done to depict the health status of the gearbox components. However, an effective signal processing methodology is always required to detect incipient faults at a very early stage as the actual fault signature is generally masked under environmental noise and considered to be difficult to extract. Hence, this paper proposes an intelligent gear fault diagnosis methodology based on Maximal Overlap Discrete Wavelet Transform and Variational Mode Decomposition (VMD) to identify the incipient fault signatures at a very early stage. To check the performance of the proposed methodology, different classifiers performance such as Support Vector Machine, Decision Tree, Ensemble Tree, Naive Bayes, and k-Nearest Neighbor are also depicted and compared. Results shows that the VMD-based signal processing technique extracts the hidden faulty signature and helps to accurately classify the fault stages.