Incipient Inter-Turn Short Circuit Fault Estimation Based on a Faulty Model Observer and ANN-Method for Induction Motor Drives

Abstract

Background: According to statistics, short circuit faults are the second most frequent faults in induction motors. Thus, in this paper, we investigated inter turn short circuit faults in their early stage. Methods: A new equivalent model of the induction motor with turn to turn fault on one phase has been developed. This model has been used to establish two schemes to estimate the severity of the short circuit fault. In the first scheme, the faulty model is considered as an observer, where a correction of an error between the measured and the estimated currents is the kernel of the fault severity estimator. However, to develop the second method, the model was required only in the training process of an artificial neural network (ANN). Since stator faults have a signature on symmetrical components of phase currents, the magnitudes and angles of these components were used with the mean speed value as inputs of the ANN. A simulation on MATLAB of both techniques has been performed with various stator frequencies. Results: The suggested schemes prove a unique efficiency in the estimation of incipient turn to turn fault. Besides, the ANN based scheme is less complex which reduces its implementation cost. Conclusion: To monitor the stator of an induction motor, the choice of the appropriate algorithm should be done according to the system in which the motor will be installed. If the motor is directing connected to the grid or fed via an inverter with a variable DC bus voltage, the observer would be better, otherwise, the ANN algorithm is recommended.