Physics-Enhanced Digital Twin based solution to control process state in a Steel Manufacturing plant

Industrial Maintenance activities are currently sustaining the continuous, local and remote, monitoring of critical facilities by determining cumulative damage and, consequently, by enabling fault diagnosis. Applying data driven models is possible to correlate features, consequently, to predict downtime and to manage, direct and indirect, costs (relevant for large scale processes) of stoppages. In this context, the timely monitoring of the relevant features, “sustained” and “populated” by physics-enhanced models clustered for dimensionality reduction, can be used to prevent anomalies. Augmented data across Physics-enhanced Digital Twin (PeDT) models can add scenarios, enlarge knowledge and select optimal maintenance initiatives. The paper proposes a Decision Support Tool (DSS) - based on PeDT - for maintenance prediction with application in a rolling mill process (located in South of Italy).