Abstract: A new partial demagnetization fault diagnosis method for permanent magnet synchronous machine (PMSM) is proposed based on the kurtosis of air-gap flux density difference signals. First, the equivalent magnetic circuit (EMC) model of PMSM is established. According to the simplified EMC model, the change of radial air-gap flux density under the main magnetic circuit of each pole after partial demagnetization fault is analyzed. Then, by differentiating the prestored air-gap flux density signals in the no-load state of the healthy motor with the measured air-gap flux density signals in the online state, and calculating its kurtosis, the fault characteristic value is calculated. The maximum of kurtosis value of the difference between the air-gap flux density signal in the no-load state and the air-gap flux density signals under different working conditions of the healthy motor is taken as the reference value. Partial demagnetization is diagnosed according to the fault characteristic value and the reference value. Finally, both the simulation and experimental results validate the effectiveness of the proposed partial demagnetization fault diagnosis method.