Predictive Maintenance is gathering a lot of interest both from research and industries. The combination of Digital Twin models and Machine Learning provides the mixture of past and featured values for application in the prediction of failures in correlation with production plans. In this work, we explored the use of Machine Learning to extract, through the important features selection, information on which sensors/data - used in a steel industry production line - can be considered "principal" through data obtained from the integration of real-time monitoring and Digital Twin elaboration. The analysis of the data, collected from a period of six months, provided information on anomalies and main signal correlation. The data from Digital Twin and Machine Learning predicted normal and in need of observation states along with the anomalies. Further investigation using Machine Learning, provided the sensors that reported the anomalies and gathered principal components. The sensors' signal data are currently used for real-time monitoring and Predictive Maintenance plans and integrated in a cloud based platform. Copyright (c) 2022 The Authors.This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Feature investigation with Digital Twin for predictive maintenance following a machine learning approach
Fabio Fruggiero
;
2022-01-01
Abstract
Predictive Maintenance is gathering a lot of interest both from research and industries. The combination of Digital Twin models and Machine Learning provides the mixture of past and featured values for application in the prediction of failures in correlation with production plans. In this work, we explored the use of Machine Learning to extract, through the important features selection, information on which sensors/data - used in a steel industry production line - can be considered "principal" through data obtained from the integration of real-time monitoring and Digital Twin elaboration. The analysis of the data, collected from a period of six months, provided information on anomalies and main signal correlation. The data from Digital Twin and Machine Learning predicted normal and in need of observation states along with the anomalies. Further investigation using Machine Learning, provided the sensors that reported the anomalies and gathered principal components. The sensors' signal data are currently used for real-time monitoring and Predictive Maintenance plans and integrated in a cloud based platform. Copyright (c) 2022 The Authors.This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.