Fault diagnosis of squirrel-cage induction motor broken bars based on a model identification method with subtractive clustering

Abstract

Fault diagnosis in electric motors is a field that evolves and grows constantly, aiming at their effective maintenance and protection scenarios under the lowest possible cost. Especially for induction motors, since they are of fundamental importance to the industry worldwide, many techniques and methodologies for the early fault detection-diagnosis have been proposed so far. In this paper, an attempt is made to develop a mechanism in order to diagnose faults in a three-phase squirrel cage induction motor rotor bars. The concept is implemented by primarily taking into account the information extracted from the classical motor current signature analysis (MSCA) and then a model identification method approach is formulated using data set manipulation known as subtractive clustering. The method is based on adaptive neuro fuzzy inference system (ANFIS). An investigation on the validity of the proposed method is performed, through experimental data taken from a healthy motor operation as well as those from the same motor with 1, 2 and 3 broken bars. From the derived results it is shown that they present satisfactory sensitivity and accuracy characteristics and thus the proposed method may be a suitable candidate mechanism in the early rotor bar fault detection phase of induction motors.