Insulation failure prediction in induction machines using line-neutral voltages

This paper presents both the experimental results and a simple theoretical validation of a novel technique used to detect turn faults in a single phase of the stator of an induction machine. The algebraic sum of the three instantaneous line-neutral voltages is shown to readily identify the occurrence of an insulation fault. Since this sum contains frequency components due to stator core magnetic saturation, slot harmonics etc., the detection sensitivity can be greatly improved by removal of these frequencies by band-pass filtering this sum around the fundamental. As the technique relies on deviations in phase impedance balance occurring during turn faults, and not changes in average phase impedance values-attributable to differing normal operating conditions-the technique is largely immune to varying load conditions, operating temperatures and source voltage perturbations. Furthermore, it is shown that this technique is not only free of relatively complicated calculations, but expectedly more so immune to numerical errors in comparison to existing techniques based upon accurate sequence calculations of voltage and current. The simplicity of this technique allows a low-cost analog or digital implementation wherein a turn fault can be reliably detected within a single cycle of the fundamental machine excitation