Induction machine broken bar detection using neural networks based classification

This paper addresses the developing of a fault diagnosis system for detection of broken rotor bars, a common mechanical fault in cage induction machines, using efficient feature extraction techniques and a neural network classifier. The proposed algorithm uses the stator current and motor speed as inputs. Fast Fourier Transform is utilized to obtain the frequency spectrum of the current signal. An efficient algorithm is then used to extract suitable features out of the frequency spectrum of the signal. The relevance of the features for the purpose of fauIt detection is investigated and verified. A neural network classifier is then developed and applied to distinguish different motor conditions. A series of data collected from experiments on a three phase 3 hp cage induction machine performed in different load and fauIt conditions are used to provide data for training and then testing the classifier. Experimental results confirm the etkiency of the proposed algorithm for detection of broken bar faults.