Stator winding turn-fault detection for closed-loop induction motor drives

Sensorless diagnostics for line-connected machines is based on extracting fault signatures from the spectrum of the line currents. However, for closed-loop drives, the power supply is a regulated current source and, hence, the motor voltages must also be monitored for fault information. In this paper, a previously proposed neural network scheme for turn-fault detection in line-connected induction machines is extended to inverter-fed machines, with special emphasis on closed-loop drives. Experimental results are provided to illustrate that the method is impervious to machine and instrumentation nonidealities, and that it requires lesser data memory and computation requirements than existing schemes, which are based on data lookup tables.