Exploiting switching transients for broken rotor bar detection in inverter-fed induction machines at all operating conditions

In comparison to rotor bar fault detection for mains fed induction machines the detection in inverter-fed drives is especially challenging. The most important reasons are the disturbances introduced by the fast switching of the inverter. Moreover the fault indicator is influenced also by control dynamics and load changes. In addition most known fault detection technologies need a certain load level to ensure proper accuracy of the fault indicator. The proposed approach overcomes these drawbacks and allows accurate detection of rotor bar faults at all operating conditions. By applying special voltage pulse patterns using the inverter switching and identifying the currents reaction it is possible to calculate the machine's transient reactance. Due to rotor fault the distribution of the transient flux linkage is altered. This leads to a fault induced asymmetry in the spatial distribution of the transient reactance that can be used as a fault indicator. However, load condition influences the fault indicator. To eliminate this influence a compensation strategy using artificial neural networks is presented. Measurement results for healthy and faulty condition show the capability of the novel rotor bar fault detection approach at different load conditions.