Spectral Correlation Analysis for Planetary Gearbox Fault Diagnosis Under Time-Varying Speeds

Abstract

Planetary gearbox vibration signals feature amplitude modulation (AM) and frequency modulation (FM) and exhibit cyclostationarity. It is essential to identify the signal carrier and modulating frequencies for gear fault diagnosis, but this is very challenging under practical time-varying speed conditions. To address this issue, this article proposes an order–order spectral correlation (OOSC) analysis method, by leveraging the advantage of cyclostationary (CS) signal analysis in modulation feature extraction. First, the nonstationary signal under time-varying speeds is resampled in angular domain, making it order-stable relative to the rotating frequency. Then, the discrete random separation (DRS) is applied to the resampled signal, to cancel interferences due to rotating frequency irrelevant components. Finally, the separated resampled signal is analyzed by spectral correlation (SC) in order–order domain. This proposed method generalizes CS analysis from constant speed to time-varying speed conditions and can reveal both the signal carrier and modulating frequency orders simultaneously, thus facilitating gear fault diagnosis under practical nonstationary conditions. It is experimentally validated on a wind turbine drivetrain test rig, and the sun gear faults in both planetary stages are successfully diagnosed.