Ensemble Learning Methods for Machine Fault Diagnosis

Recently, industries have been focusing more on machine fault diagnostics to avoid downtime. Induction motors (IM) are widely employed in the manufacturing and process sectors; however, they are susceptible to various faults. Bearing failure is one of the most prevalent IM faults that affect the production process. This paper proposes an Ensemble model for detecting bearing faults in IM using vibration signal analysis. Decision Tree (DT), Random Forests (RF), Support Vector Machine (SVM), K-nearest neighbors (KNN), and XGBoost (XGB) are considered as base models. The real-time vibration data is acquired using the data logger from healthy and faulty IMs. Fault detection is performed using the base models with time and frequency domain features. The ensemble models developed using machine learning base models and voting classifier improved fault detection accuracy. The KNN+XGB+SVM model provided an accuracy of 99.2%, performing better than other ensemble models with frequency-domain features.