Enhanced vibration separation technique for fault diagnosis of sun gear

The complex structure of a planetary gearbox weakens the gear fault characteristics in measurement signals. Vibration separation (VS) technology can address this problem by eliminating the impact of time-varying transfer paths on the signal. However, existing methods for determining the first window position (FWP) of VS are not highly accurate. To address this issue, a novel method for determining the FWP has been proposed in this paper, thereby creating an enhanced VS (EVS) technology. First, the envelope signal, which contains the amplitude-modulated signal (AMS) due to time-varying transfer paths, is separated using a zero-phase bandpass filter and Hilbert transform. Then, adaptive chirp mode decomposition is employed to accurately estimate a harmonic of the AMS. A sequence is formed by identifying the maximum points of the harmonic signal, indicating the moments when the planetary gear passes beneath the accelerometer. Finally, the VS method is applied based on the selected FWP. The results from both simulation and experimental signal analyses indicate that the FWP error of the EVS method (1.57%) is lower than the errors (1.83, 20.06, and 12.04%) found in the comparison methods. Additionally, the fault characteristic amplitude in the envelope order spectrum using EVS (9.78 m/s²) is higher than the amplitudes (6.14 m/s², 6.18 m/s², 8.56 m/s², and 0.95 m/s²) recorded by the comparison methods. These findings confirm the effectiveness and superiority of EVS over other methods in terms of FWP accuracy and fault characteristic enhancement.