Towards Practical Model-Based Decision Making

Model-based decision making is a cornerstone of modern chemical process engineering, essential for tasks ranging from process design to real-time optimization. However, its widespread application is often hindered by two major bottlenecks: the high cost of experimental data collection for model calibration and the computational burden of rigorous process simulations. To bridge the gap between theory and practice, this thesis advances two critical methodologies: Optimal Experimental Design (OED) and Surrogate Modeling.

Addressing the need for experimental efficiency, this work establishes trust in OED through a wet-lab benchmark study, demonstrating increased data efficiency over traditional factorial designs. It further introduces a novel cubature-based method for uncertainty quantification in nonlinear models. To tackle the lack of reusability in surrogate modeling, the thesis proposes a custom modelfluid representation. This enables the training of a distillation column surrogate that generalizes across chemical systems. Together, these contributions provide a synergistic framework for more efficient and scalable model-based engineering.