Abstract: Planet bearing kinematics involve spinning-revolution coupling, resulting in complex and weak fault vibration signals. This poses a tremendous challenge to fault diagnosis. Under time-varying speed conditions, the frequency characteristics of gear mesh vibrations overlap with those of planet bearing faults, significantly interfering planet bearing fault diagnosis. To address this issue, an order-frequency spectral correlation of non-stationary signals is proposed in this paper. The time-varying low-frequency amplitude envelope of the vibration signal is removed, and angular domain resampling is performed to stabilize the order characteristics of gear vibrations. The discrete random separation in the order domain is applied, to eliminate gear vibrations and retain residual random components. The random components are inverse resampled in the time domain to restore the original amplitude envelope, and planet bearing fault features are extracted through their order frequency spectral correlation or coherence. This method enhances planet bearing fault features, and improves the diagnosis capability under time-varying speed conditions. The principle is demonstrated through numerical simulation analysis, and the performance is validated experimentally. The localized fault on the inner and outer race and rolling element of planet bearings are successfully diagnosed.