Edge-Enabled Fault Prediction Of Motor Bearings

Abstract

Our project focuses on implementing a comprehensive solution to reduce the risk of major breakdowns in electric motors caused by bearing faults. At its core, we developed an innovative edge cog device that uses vibration sensors for early prediction of different bearing faults in motors. The collected vibration signal data can give us a better knowledge of the various faults happening in mechanical systems. In this work, we analyze vibration signal data by integrating several signal processing methods and connecting them with machine learning approaches to categorize various types of bearing failures, therefore avoiding considerable downtime and costly repairs. The procedure begins with these sensors gathering real-time vibration data, which is then analyzed at the edge node using various signal processing algorithms to detect bearing faults. This data is then transferred to the cloud, where a deep learning neural network made up of 1-D convolution, pooling, and dense layers, and trained on a variety of datasets containing both healthy and defective bearings is examined to precisely determine the fault type. Furthermore, we provide a user-friendly dashboard that enables real-time monitoring and notifications, allowing maintenance teams to resolve any irregularities as soon as they are noticed. By combining these advanced techniques, our edge device not only prevents catastrophic failures but also promotes timely maintenance plans and operational efficiency in general.