Aiming at the difficulty in planetary gearbox fault diagnosis, a method combining Particle Swarm OptimizationVariational Mode Decomposition (PSO-VMD) and Partial Mean of Multi-scale Entropy (PMMSE) was proposed. First, to reduce the drawback of subjectively selecting the parameters of VMD, PSO is used to optimize the decomposition scale and secondary penalty factor. Then, the signal is processed by VMD into several components and the valid components are selected by mutual information method to reconstruct signal. Multi-scale entropy can analyze the complexity of signal under different scale factors, and PMMSE based on multi-scale entropy can reflect the mean value and variation trend of multi-scale entropy. Finally, the PMMSE of reconstructed signal is applied to describe the state of planetary gearbox, and thereby to diagnose planetary gearbox faults. By analyzing the experimental signal of planetary gearbox, the results indicated the proposed method could primely distinguish different faults of planetary gearbox.