Nonlinear Autoregressive with Exogenous Model to Diagnosis Open Switch Fault in PMSM

Abstract

This study aimed at presenting a technique to diagnosis faults of electronic switch in permanent magnet synchronous motor (PMSM). The current output of both thyristor bridges and the diode of system excitation is monitored under healthy and faulty operations. Features extracted at different operations using Multi-scale wavelet decomposition (MSWD). MSWD features are used to train Nonlinear Autoregressive with Exogenous Model (NARX) which sequentially operated to evaluate the fault level in case open circuit that developing across a switch. The two models have been tested and designed due to the simulated data, wherein the findings showed acceptable effectiveness in the diagnosis of various types of fault.