An encoder information-based anomaly detection method for planetary gearbox diagnosis

To keep the reliability of the planetary gearbox, anomaly detection has been widely investigated for its health monitoring. To this end, a novel approach is presented in this paper to extract fault features based on the merits of built-in encoder signals. Considering that collected encoder data is accumulated in angular positions, instantaneous angular acceleration (IAA) is firstly calculated to highlight the characteristic components. And then time synchronization average (TSA) is applied on an estimated multi-period for denoising, which improves the robustness of the TSA to the feature attenuation effect caused by the round-off error of the basic period. In this paper, we explore the distinguishing properties of regular components and the fault anomaly to impose different restraints on them, which is embodied as a periodicity-enhanced model of robust principle analysis. And objective features are further separated by solving this optimization model. The validation analysis of the proposed framework is applied on both the simulation and experimental cases. The results show that the proposed method is of good performance to deal with encoder signals from the planetary gearbox for fault diagnosis.