Fault diagnosis of broken rotor bars in induction motor using multiscale entropy and backpropagation neural network

Interruptions in any process industry due to machinery problem induce a serious financial loss. And as we know that induction motors occupy a major area in machinery and process industry, detection of faults beforehand is a key to avoid the state of financial or production crisis in future. The present work proposes a novel algorithm for the detection of broken rotor bars in induction motor. Stator current in addition to rotor vibration in an induction motor was measured and employed for fault detection of broken rotor bar. Multiscale entropy (MSE) is used as statisticbased approach in order to tackle the nonlinear behavior existing in rotor bar using vibration and current as the diagnostic media, as both cumulatively considered describe the regularity in the diagnostic information. The proposed work presents an approach to analyze features that distinguish the rotor vibration and stator current samples of normal induction motor from those of the broken rotor bar. Further, backpropagation neural network classifier is applied over the resultant feature set which distinguishes the faulty data set from the healthy with an accuracy level of 15.5 % for vibration and 14 % for current. © 2015, Springer India.