Using reduced-order and data-driven techniques, mathematical modelling and real-time optimization of digital twins for cyber-physical systems

Reduced-Order Models and Data-Driven Digital Twin Mathematical Representations and Real-Time Optimization Abstract digital twins offer a way of real-time monitoring, control and optimization of a physical asset, hence transforming technologies in cyber-physical systems (CPS). By merging mathematical modelling, reduced-order, and data-driven modelling approaches, this work addresses how CPS digital twins might be more efficient and accurate. Complex physical systems are captured using ROMs, which also help to preserve significant dynamics by lowering computation. Using data gathered on real-world systems, machine-learning and other statistical methods are utilized to enhance these models to maintain the digital twin updated with the actual system. Combined with data-driven approaches, ROM makes real-time optimization conceivable in which parameters of the system are continuously modified to acquire optimal system performance under evolving conditions. For industries including manufacturing, energy, and transportation where real-time judgments are required, this is particularly helpful. The paper offers analysis of how such integrated models developed, their uses in CPS, some of the difficulties these models offer for the future and chances for improvement. These contributions comprise building data-driven physics-based digital twins using libraries of component-based RMIs and leveraging parallel reduced-order modelling for high-performance computing pipelines. Proceeding toward more accurate, scalable, and efficient digital twin to support industry and smart manufacturing.