Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
A novel approach for anomaly detection in edge analytics of vibration data in rotating machinery
Submission Author:
Lucas Costa Brito , MG
Co-Authors:
Julia Duarte, Lucas Costa Brito, Marcus Antonio Duarte, Aldemir Ap Cavalini Jr
Presenter: Julia Duarte
doi://10.26678/ABCM.COBEM2023.COB2023-0363
Abstract
Due to the increasing use of Edge analytics techniques, made possible by the development and lower cost of IoT (Internet of Things) sensors, together with the assumptions of Industry 4.0, real-time vibration monitoring of rotating machinery has become a reality in modern industry. In order to be able to manage a large volume of data and information, in addition to reducing high storage costs, new monitoring tools have been seeking to offer integrated anomaly detection. However, it is noted that these tools still generate many false negative results, leading to breaks or storage of undesirable data. In this paper a new approach based on Autoencoder neural networks and statistical histogram comparison techniques for detecting anomalies in dynamic data streams is presented. Validation in three real datasets demonstrate the viability of the approach, with Recall results superior to 90%.
Keywords
real-time monitoring, Vibration Analysis, rotating machinery, Anomaly detection, Edge Analysis
