Extraction of weak fault transients using variational mode decomposition for fault diagnosis of gearbox under varying speed

Abstract Non-stationary vibration signals of a gearbox under varying speed display complicated modulations, which lead to intense sidebands thereby resulting in difficulty to identify the presence of a fault. Variational mode decomposition (VMD) being highly adaptive, effective in attenuating mode-mixing problem, low computational time requirement and therefore it is suitable to decompose a modulated multi-component non-stationary gearbox vibration signal. This research work thus utilizes the merits of VMD for demodulation and to diagnose localized gear tooth faults under real-time speed variation. In the present study, the vibration signals of a gearbox for different faults were acquired under complete speed variation (i.e., run-up, random fluctuation, and coast-down). To identify the symptoms confirming the presence of faults, the vibration signals were decomposed by VMD thereby demodulating the raw vibration signal. The decomposed VMFs exhibited the presence of transients due to faults and were analyzed statistically. To state the effectiveness of VMD in exhibiting fault by extracting fault transients, the performance of VMD was compared with the performance of recently developed wavelet based empirical wavelet transform (EWT) and flexible analytic wavelet transform (FAWT). The fault detection performance of VMD outperforms EWT as demonstrated by simulated signal and validated by the experimental investigations. FAWT was found ineffective in decomposing the vibration signal under varying speed.