Due to the difficulty of collecting planetary gearbox fault data, applications of data-driven methods for fault diagnosis of planetary gearboxes have always been a challenge in the case of insufficient data. To address this issue, a fault diagnosis method of planetary gearbox based on digital twin of virtual and real data consistency is proposed. A digital twin of the gearbox was created in UG and dynamically simulated by ADMAS, resulting in simulation data for four health states, which were compared with the measured data for subsequent fault diagnosis. The results of the study highlight that the proposed methodology promises significant progress in fault diagnosis under conditions of limited availability of fault data.