Robust health evaluation of gearbox subject to tooth failure with wavelet decomposition

Machinery condition monitoring is a key step to perform condition-based maintenance (CBM) policy. In this paper, a novel health evaluation method based on wavelet decomposition is proposed. In the process of wavelet decomposition, a new index is defined to choose the optimal detail signal. After that, frequency spectrum growth index (FSGI) is proposed to serve as a quantitative description of machine health condition. This index is helpful for maintenance decision-making. At the same time, a semi-dynamic threshold criterion that can be used to check the existence of fault is established. In order to demonstrate the performance of this index with its semi-dynamic threshold, a comprehensive study with three sets of vibration data collected from a mechanical diagnostics test bed is conducted to validate this method. The analysis results indicate that the proposed method is insensitive to the selection of wavelet function and wavelet decomposition level, which means that FSGI has excellent performance in gear early fault detection.