PCA and ANN Based Induction Motor Fault Classification

Abstract

Issues and malfunctions in induction motors can result in significant periods of inactivity and substantial financial losses due to maintenance and decreased revenue. This underscores the importance of monitoring motors, detecting potential faults in their early stages, and facilitating diagnoses. Among the recurrent problems encountered in induction motors, turn-to-turn short circuits, bearing degradation, and cracked rotor bars are the most prevalent. This study introduces an approach for detecting multiple faults in induction motors. The Principal Component Analysis (PCA) is recommended to condense the feature dataset extracted from motor currents under both normal and faulty conditions. From a pool of fourteen statistical parameters derived from stator currents, the application of PCA reduces the dimensionality to ten new features. These newly formed features are subsequently harnessed to categorize faults in induction motors using the Artificial Neural Network (ANN) classifier. Furthermore, a comparison of classification outcomes is conducted before and after employing principal component analysis. Empirical findings validate that by employing PCA in conjunction with ANN, it is possible to accurately classify various faults present in an induction motor.