A mechanical fault diagnosis method of on-load tap-changer（OLTC）is proposed based on MPE and im-proved K-means algorithm. First

different mechanical faults of OLTC are simulated

and the vibration signals arecollected. Second

in order to realize the fault diagnosis of OLTC under nonlinear vibration signals

the multi-scalepermutation entropy（MPE）is used to extract the features of OLTC mechanical fault state. Third

the K-means clus-tering algorithm optimized by particle swarm is used to diagnose the mechanical fault of OLTC. Finally

the proposedmethod is applied to the mechanical fault diagnosis of OLTC

and is compared with the methods based on traditionalK-means algorithm and BP neural network. The results show that the proposed diagnosis method based on MPE andimproved K-means algorithm is suitable for mechanical fault diagnosis of OLTC

and it has high stability and betterdiagnosis effect than the both methods based on traditional K-means algorithm and BP network.