Stator Inter-turn Faults Detection and Localization Using Stator Currents and Concordia Patterns - Neural Network Applications

Kit inter-turn fault is one of the most difficult failures to detect. Depending on the motor protection, the motor may continue to run but, soon or later, the heating in the short-circuited turns will cause severe failures. In this paper, the stator inter-turn faults are detected by observing the Concordia patterns and the behavior of the effective current values (RMS). These fault indicators have a specific behavior corresponding to each fault and can be used detect, locate and evaluate faults. Also two strategies are proposed to use multilayer feedforward neural networks (FFNN) in the fault detection, and both strategies are sketched and the best configuration is used. The network output can be interpreted, using a table, as a diagnosis report of the machine health. In addition, the use of two neural networks at the same time is proposed to improve the prediction accuracy.