Abstract
Reduced-order Modeling Enables Digital Twins for Smart Process Engineering
Smart process systems heavily rely on virtual process models for optimizing the process yield and the product purities, designing feedback controllers to guarantee stable operations and attenuate disturbances, and for process surveillance during plant operation. Hence, digital or virtual twins of all the processing steps are needed. As many of the process steps include exothermic reactions, complicated reaction kinetics, fluid-structure interactions, particle aggregation, nucleation, breakage and growth described by complex population balance models, etc., the mathematical models often consist of coupled systems of time-dependent nonlinear partial differential equations (PDEs). Hence, surrogate models are needed that allow optimization and control design with limited time budgets, and real-time computations for fault diagnosis and surveillance. Model order reduction of dynamical process models is an enabler to design computationally efficient surrogates for real-time operation. Many recent advances in model order reduction of nonlinear dynamical systems allow the application to the coupled PDE models occurring in process engineering. We will briefly sketch the underlying mathematical methodologies and demonstrate the power of these techniques using several problems from chemical engineering.