Detection of Eccentricity Faults of Induction Motors Based on Decision Trees

Abstract

One of the common faults of induction motors is the eccentric faults. In this study, the eccentricity faults of a 3 kW induction motor were analyzed by using Decision Tree (DT) method. The eccentricity fault was implemented in laboratory by expanding the bearing housing located inside endshields of the induction motor. The rotor rotation center was shifted by 0.2 mm. The induction motor was first tested at healthy condition and then with eccentric fault at full load. The three-axis vibration signals of healthy motor and motor with eccentricity fault were collected. Various statistical features of each vibration axis were analyzed. The results show that the statistical features extracted for each axis vibration signals can distinguish a healthy state of the motor from a faulty state. A simple machine learning (ML) algorithm, Decision Trees (DT) is employed to distinguish the faulty motor. The results show that by using DT eccentricity faults can be distinguished with accuracies nearly up to 100 %.