Entropy and K-Nearest Neighbors-Based Feature Extraction for Bearing Fault Detection

Abstract

Bearing failures in rotating machines can lead to significant operational challenges, causing up to 45-55% of engine failures and severely impacting performance and productivity. Timely detection of bearing anomalies is crucial to prevent machine failures and associated downtime. Therefore, an approach for early bearing failure detection using entropy-based machine learning is proposed and evaluated while combined with a classifier based on K-Nearest Neighbors (KNN) and Support Vector Machine (SVM). Entropy-based feature extraction should be able to effectively capture the intricate patterns and variations present in the vibration signals, providing a comprehensive representation of the underlying dynamics. The results of the classification carried out by KNN-Entropy have an accuracy value of 98%, while the SVM-Entropy model has an accuracy of 96%. Hence, the Entropy-based feature extraction giving the best accuracy when it is coupled with KNN.