{"id":3352,"date":"2022-07-13T09:30:00","date_gmt":"2022-07-13T07:30:00","guid":{"rendered":"https:\/\/cirpicme.org\/?page_id=3352"},"modified":"2022-07-26T13:05:07","modified_gmt":"2022-07-26T11:05:07","slug":"digital-twins-for-predictive-maintenance","status":"publish","type":"page","link":"https:\/\/cirpicme.org\/index.php\/cyber-physical-systems-digital-manufacturing-2\/digital-twins-for-predictive-maintenance\/","title":{"rendered":"Digital Twins for Predictive Maintenance"},"content":{"rendered":"\n

by Tobias Harries, Matthew Hartnoll, Mohammadmilad Hafezianrazavi, Harry Meek, Aydin Nassehi (UK)<\/em><\/p>\n\n\n\n

<\/p>\n\n\n\n

Abstract<\/strong><\/p>\n\n\n\n

Modern maintenance strategies tend to be a mixture of different approaches, dependent on the type of machine and the component failure mode. For some machines, time-based maintenance is appropriate, however not all failure is fully age-related and other situations may require risk- based or condition-based maintenance. Predictive maintenance is a method originating from condition-based maintenance and is primarily used to predict the remaining useful life of a machine. Predictive models can be data driven, experimental, physics based or a hybrid thereof. In this paper, a virtual factory is used together with a data\/experimental driven hybrid approach through the use of statistical models encompassed in a digital twin to calculate the remaining useful life of machines in the factory. An augmented statistical model is chosen for this purpose as it relies on data being fed from the factory during operation, therefore suiting the twin\u00d5s functionality. Sensor data from previous machines or failure profiles is pre-processed and machine learning is used to extract condition indicators. These indicators are used to identify different types of fault and predict degradation paths. Such a method requires no mechanistic understanding of the process and is thus generalizable. Furthermore, with the possibility of using synthetic data, the method requires limited historic data compared to other methods, therefore suiting the scale of small to medium factories where predictive maintenance approaches are most challenging to implement.<\/p>\n\n\n\n

Keywords<\/strong>: Predictive maintenance; Digital twin<\/p>\n\n\n\n

<\/div>\n\n\n\n

Video presentation<\/strong><\/p>\n\n\n\n