Planetary Gearbox Fault Detection Under Speed-Varying Conditions With Angular-Domain Cyclostationary Feature Mode Decomposition

Planetary gearbox fault diagnosis under variable-speed operational conditions still presents several difficulties. To overcome this issue, this article investigates an approach, angular-domain cyclostationary feature mode decomposition (ACFMD), for fault detection in planetary gear systems using built-in encoder signals. The encoder signal is applied to capture torsional vibration characteristics and rotational phase information of the reference shaft for transient fault analysis. First, the decomposition directions of the candidate modes are determined by constructing a series of different initialized filters. Then, the angular weighting indicator of second-order cyclostationary (ICS₂) is applied as the optimization objective of the mode update. Finally, the fault impulses component can be accurately locked and extracted after iteratively performing mode update and selection processes without angular-domain resampling. The effectiveness of ACFMD in accurately identifying fault signatures under time-varying speed conditions is verified through both simulation and case studies. The proposed ACFMD significantly enhances the applicability of feature mode decomposition (FMD) theory for fault detection under nonstationary situations.