Motor current signal analysis using deep neural networks for planetary gear fault diagnosis

Abstract

Failures in planetary gearboxes can cause accidents, downtime, and high maintenance costs. Motor current signal analysis (MCSA) offers a non-intrusive method for detecting mechanical faults in rotating machinery. Recently, many intelligent diagnostic methods based on deep learning have been proposed, providing an effective way to quickly process a large amount of fault data and automatically produce accurate diagnostic results. However, most current intelligent methods for diagnosing faults are based on vibration signals. Owing to the characteristic differences between current signals and vibration signals, ideal diagnostic results cannot be obtained by directly using the spectrum signal as the sample. This study diagnosed faults in planetary gears by preprocessing current signals and using deep neural network algorithms. The effectiveness of this method is proven using experimental data, which contain substantial measurement signals covering different health conditions under different loading conditions. Furthermore, this method is compared with other methods to demonstrate its superiority.