(596c) Finding the Right Space for Hybrid Modeling of Chemical Processes

Traditional dynamic modeling of chemical processes typically relies on first-principles models with constant parameters, limiting their ability to capture the nonlinear and time-varying behavior of real systems. To address this limitation, we have developed a hybrid modeling framework that integrates system-agnostic first-principles dynamics with system-specific, data-driven, time-varying parameters. The objective of this approach is not to construct a perfect digital twin, but rather to improve the real-time fidelity of existing physics-based digital twins by reliably transforming them closer to reality through systematic calibration. This framework is broadly applicable across multiple modeling paradigms (e.g., DFT, MD, kMC, and CFD) and length scales (from laboratory to plant scale) and is currently being demonstrated in collaboration with leading partners in the chemical process industry.