Special Issue on Reduced Order Modeling, Generative AI, and SciML in Digital Twins

The emergence of digital twins facilitate “personalized” control and optimization decisions by taking into account the unique characteristics of individual assets or processes. Two of the most critical aspects of digital twins are (1) real-time updating of a digital model using data collected from the physical system (physical-to-digital) and (2) real-time control, optimization, and decision making of/for the physical twin based on insights from its digital twin (digital-to-physical). While digital twins have great potential to revolutionize design optimization and operation of engineering systems, the real-time coupling and synchronization between a physical system and its digital counterpart—through model updating, control, inference, and optimization—poses significant challenges for the widespread adoption in industry. Reduced-Order Modeling (ROM), generative Artificial Intelligence (AI), and Scientific Machine Learning (SciML) are key enablers that address these challenges. They play a vital role in significantly reducing the time required for the bidirectional coupling between the physical system and its digital model, thereby enabling the potential of digital twins.