Digital Twins for Critical Infrastructures: Simulation Methods for Contaminant Dispersion

Abstract

We aim to establish a highly automated digital twinning workflow to effectively simulate air contaminant dispersion. With that we strive to provide assistance and informed real-time decision-making processes in the context of evacuation scenarios. Provided by open-source geo-referenced databases an initial simulation domain is set up from up-to-date urban geometries. Based on that, a wind flow simulation is coupled with the computation of contaminant concentration. To accelerate the assessment of the wind flow field, a Reduced Order Model (ROM) is desired to use instead of the high-fidelity simulation. It is shown, that ROMs can approximate the high-fidelity output with a high accuracy while leading to a good speed-up of the simulation. In future developments, we aim to include ROMs for more complex geometries as well as investigating the application of scientific machine learning methods to merge physics with data.