[Objective] Permanent magnet synchronous motor (PMSM) is widely used in wind power generation and electric vehicles due to its high power factor, simple structure and good dynamic performance. However, PMSM may experience phase loss faults during operation due to reasons such as drive failure or loose stator winding connections. When operating with phase loss, the PMSM generates noise and vibration, leading to a reduction in output power. Prolonged phase loss operation can also damage electrical equipment, making accurate fault diagnosis crucial for ensuring the normal operation of the equipment. [Methods] This paper proposed a fault diagnosis strategy based on an improved empirical wavelet transform (IEWT) and categorical boosting (CatBoost) algorithm, and applied it to the phase loss fault diagnosis of six-phase PMSM. First, the basic principle of the IEWT algorithm was introduced. The IEWT algorithm performed spectral segmentation on the Welch power spectrum, effectively suppressing modal aliasing compared