A Data-Driven Approach Based on Artificial Neural Networks for the Detection and Classification of Bearing Anomalies in Power Generation Plants

Abstract

Power generation plants play a crucial role in modern societies, but they are vulnerable to different types of anomalies and faults that can have serious economic and environmental consequences. Bearing anomaly detection is an effective approach to recognize potential failures beforehand and avoid their occurrence. Recently, artificial neural networks (ANNs) have emerged as a promising approach for detecting anomalies in power generation plants, owing to their capability of acquiring intricate patterns and adapting to diverse operating circumstances. The presented study proposes a novel method to detect bearing anomalies in power generation plants using artificial neural networks. The approach aims to enhance the precision and dependability of anomaly detection by incorporating diverse features extracted from bearing data signals. Experimental validation was carried out on vibration data obtained from a real-world power generation plant to demonstrate the effectiveness of the proposed approach for detecting bearing anomalies. The results indicate that the proposed approach surpasses conventional methods, emphasizing the potential of ANNs for detecting vibration anomalies in power generation plants with higher accuracy and reliability.