Analysis and detection of broken rotor bars in induction motor under fluctuating load by means of stray flux signals

Abstract

The early detection of faults in induction motors is an important task to avoid unexpected stops and financial losses at industrial facilities. Such fault identification becomes a hard task when the induction motor operates under the effect of fluctuating loads because they introduce undesired spectral components that may overlap with other fault-related frequencies such as broken rotor bars frequency components. Therefore, the common methodologies based on spectral analysis are prone to fail when discerning between faulty conditions (e.g. broken rotor bars) and operation under fluctuating loads. In this work, a diagnosis methodology based on the analysis of stray flux signals acquired in the vicinity of an induction motor that operates under fluctuating load conditions is proposed. The presented method is based on a statistical feature calculation that is followed by a linear discriminant analysis along with a neural network classifier, for the automatic detection and classification of different fault severities associated with broken rotor bars in the induction motor. The proposed method is evaluated with experimental data and the obtained results demonstrate the capability of the proposed method to detect the occurrence of broken rotor bars in an induction motor that operates under fluctuating loads conditions. This proposal is an adequate strategy to be implemented in industrial environments.