Abstract

Conventional methods used for fault detection of planetary gears are usually targeted for use with gears under constant speed conditions. Recently, some methods have been developed using an order-tracking method, or wavelet transform (WT), to detect planetary gear faults under variable-speed conditions. However, these previously studied techniques are computationally inefficient. Inefficiencies in these methods exist because 1) the ordertracking method of these techniques requires additional signal processing procedures, along with angle measuring devices, and/or 2) the WT process requires a lengthy computation time. Therefore, this study proposes a more efficient method, specifically a variance of energy residual (VER) method, for planetary gear fault detection under variable-speed conditions. The proposed VER method does not need angular information, and offers the potential to reduce computation time by using short-time Fourier transform (STFT) instead of WT. To demonstrate the performance of the proposed VER method, we calculated the fault sensitivity and computation time of the VER method using both simulation and experimental signals. From the results, it can be concluded that the VER method shows better fault sensitivity e with less computation time e than the previous method that uses WT.