Intelligent Machine Learning based Fault Prediction Model for Induction Motor

Abstract

Industrial sector has taken revolution with the induction motors (INM) playing a crucial role in automating many industrial processes. In this context monitoring, minimization of downtime of induction motors becomes very essential. In this work, all possible faults of induction motors such as bowed-rotor fault, broken-end ring, broken-rotor bar, cage fault, coil- to-coil, eccentricity fault, inner-race fault, misaligned-rotor fault, outer-race fault, phase- unbalancing, phase-to-ground, phase-to-phase, rolling-element fault, single-phasing, turn-to- turn, unbalanced-rotor fault are reported and intelligent machine learning (ML) algorithms are employed to diagnose the faults. A comparative analysis on the performance of the ML models for the fault study is carried out and the ML models executed are Random Forest, Support Vector Machine, Decision Tree, Naïve Bayes, Logistic-Regression and Gradient Boosting. The INM system is analyzed for its healthy case and faulty cases to collect the data which is prepared and then feature engineering is exerted to extract the prominent features using Principal Component Analysis (PCA). The comparative analysis of the performance metrics such as accuracy, precision, recall, and F1-score indicate the efficacious performance of the Random Forest (RF) model in comparison to other ML models.