A Multidomain Feature Fusion-Based Method for Stator Winding Fault Detection Equipment of Permanent Magnet Synchronous Machine

Inter-turn short circuit, a prevalent electrical fault in permanent magnet synchronous machines, can lead to serious consequences without prompt diagnosis. A multi-domain feature fusion based diagnostic method is proposed, which only requires two current sensors to measure the phase currents. The essence of the proposed method lies in using signal decomposition algorithms to extract instantaneous phase from phase currents, followed by constructing direct and quadrature axis currents for diagnosis. This paper proposes using the rime algorithm to optimize the bandwidth constraint and the number of modes of the signal decomposition method to improve accuracy. A data pre-truncation method and phase reconstruction are proposed to avoid errors caused by substantial estimation errors near data boundaries. Subsequently, fault characteristic components can be extracted by integrating the calculating order tracking and cumulative phase. Finally, experimental validation is conducted using a fault prototype. The results show that the maximum errors between the estimated and measured quadrature-axis current are only 0.6% and 2.4% under steady-state and transient conditions, respectively, thus achieving precise extraction of fault characteristics.