Adaptive-fuzzy-based stator-winding fault diagnosis of PM brushless DC motor drive by monitoring supply current

An adaptive neuro-fuzzy inference system (ANFIS) for the fault diagnosis and prediction of stator winding interturns of PM brushless DC motor drive systems is presented. A transient model of the drive system is set up and run under different healthy and faulty conditions. The steady-state supply current is monitored and its Fourier transform is used to compare the healthy operation frequency spectrum to the faulty operational spectra under different loads. Diagnostic indices were extracted as the frequency locations and component magnitudes of two specific samples of the frequency domain sequence. A Sugeno-based ANFIS was set up and trained using the simulation results to differentiate between healthy and faulty operations. This system was tested by simulation results under different interturn percentages and load torques from those used to train it; the test results were quite acceptable.