Fault Diagnosis of Multiswitches and Current Sensors in Dual-Three-Phase PMSM Drives Using Current Mirror Techniques

Current sensor fault (CSF) and open-circuit fault (OCF) are common faults in dual-three-phase permanent magnet synchronous motors (DT-PMSMs) drive, and their characteristics can be captured in the stator currents. However, the complexity of multiphase systems increases the difficulty of separating hybrid faults. This article proposes a strategy for CSF/OCF hybrid fault diagnosis and identification (FDI) with the current digital-twin model. Based on the current analysis of multiswitch OCF and various CSF, we provide sufficient evidence for the separability of CSF/OCF according to the stator currents. For this reason, a deep recurrent neural network (RNN), commonly used for text generation, is first applied to construct a precise current mirror model, and detailed group network planning, as well as parameter training, is accomplished. Furthermore, on the basis of the verified current model and the evidence of fault current separability, a simple k-nearest neighbor (k-NN) algorithm can be employed to diagnose and separate 12 common faults in multiphase systems simultaneously. The experimental results demonstrate that the designed strategy achieves a satisfactory level of accuracy, with the classification accuracy reaching up to 95.2%.