The objective of this high-risk, low cost feasibility study was to obtain data to which a concept for detection of early stage insulation failure in PWM controlled low voltage motors could be applied. The project scope included preliminary analyses to determine what parameter (current, voltage, or power) and feature extraction methods (spectra, wavelet transforms, time domain statistics, etc.) appeared to provide the greatest sensitivity to early stage insulation degradation.

The approach chosen to generate insulation failure was to remove a section of the motor housing and locally heat the first end turn of one of the coils. Electrical insulating materials rapidly lose their insulating ability with increasing temperature. The end turn of the first coil may be particularly vulnerable location for insulation failure, because of the additional mechanical bending stresses in this region. A 440 V, 3 phase, 37.3 kW (50 Hp), 1750-RPM motor and drive were used. Current and voltage were measured only on the faulted phase using a 1.25 MHz bandwidth data acquisition system (although, the current sensor had a 1 dB down bandwidth from DC to only 150 kHz).