Neural network application for fault detection in electric motors

This research describes the monitoring of the fundamental spectral features of the bearing damage through accelerated aging studies for induction motors with a power rating of 5 HP. For this aim, the bearing damage is characterized between 2-4 kHz through the spectral analysis methods applied to motor vibration signals. Also, coherence analysis approach, defined between the stator currents and vibration signals, is used for as another indicator of the bearing damage. After the computation of the coherences, a neuro-detector based on the auto-associative neural structure is trained in the frequency domain. Hence, the bearing damage detection is realized by observing the changes in the errors (residuals) generated by the neural net.