Simulation building blocks for predicting critical system changes

Abstract

At the DLR Institute for the Protection of Terrestrial Infrastructures (PI) we develop digital twins to increase the resilience of critical infrastructures (CIP). This requires comprehensive knowledge of the physical system under consideration, coupled with a deep understanding of its dynamical behavior. The department Simulation Methods for Digital Twins (PI-SDZ) aims to integrate physics-based simulation methods, reduced-order models and physics-informed machine learning into a digital twin framework. In particular, PI-SDZ is developing new research contributions in four areas: (i) high-fidelity simulation methods are applied to critical infrastructures to obtain detailed analyses of the system’s behavior, impairment and possible failure due to crisis’ onsets; (ii) virtual sensing techniques and optimal sensor placement strategies are developed and used to maximize the information collected from real infrastructures; (iii) model order reduction techniques are investigated to narrow down the complexity of real structures, while preserving the essential systems characteristics; (iv) methods of physics-informed machine learning are addressed to combine all the aforementioned modelling and simulation building blocks into a predictive tool capable of identifying critical system changes in real-time or near real-time. Thus, once integrated into a digital twin, the presented methods are promising building blocks to increase the resilience of critical infrastructures components.