Permanent magnet synchronous machine(PMSM)often works in complex and variable working conditions. To address the problem that by conventional diagnostic methods based on air gap flux signals it is difficult to identify multiple types of demagnetization faults. A multi-type classification and identification method of demagnetization faults were proposed based on the feature extraction of flux density waveform area signal and particle swarm optimization support vector machine algorithm. Firstly

the PMSM finite element model was established

the air gap flux density signal was extracted by finite element simulation

and the air gap flux density was analyzed. Secondly

using the sliding window method to extract the air gap flux density waveform area difference as a feature quantity

and conducting simulation analysis of various local demagnetization fault types and uniform demagnetization fault types based on the simulation results

a demagnetization fault sample library was constructed according to the simulation results. Then

a support vector machine model optimized by particle swarm optimization was established to accurately diagnose and locate the demagnetization fault of the PMSM. Finally

a prototype test platform was established to diagnose the demagnetization faults of one healthy motor and three faulty motors under various working conditions. The experimental results show that the fault identification rate reached 100%.