Model-based design of experiments and measurement optimization frameworks based on scalable and tractable algorithms and software

Mathematical models have become increasingly critical due to the rapid advances in computational methods in recent decades. However, the validation of these models often demands extensive and costly data, leading to time-consuming processes. Traditional design of experiments (DoE) methods struggle to choose informative experiments, especially for the typically large-scale, nonlinear, and dynamical science-based models in chemical and biomolecular engineering (CBE). In this dissertation, I propose a sequential model validation workflow powered by novel DoE and measurement optimization (MO) frameworks to improve data acquisition efficiency and accelerate the model building and validation process. The workflow relies on two scalable and tractable frameworks, along with their generalized open-source software tools: Measurement optimization: determines what to measure for experiments to maximize the experimental information content. It guides apparatus preparation during the experimental setup stage, balancing the information content with practical constraints such as budgets Model-based design of experiments: quantifies experimental information content statistically and optimizes experiment selection based on updated model information. This framework is used throughout the model validation process, recommending new experiments after each iteration to update the model Both frameworks focus on addressing the challenges of DoE and MO techniques leveraging large-scale, nonlinear, and dynamical models in CBE, providing user-friendly open-source software tools for widespread applications. In this dissertation, I describe the development of the model-based DoE and the MO frameworks and their generalized tools to streamline the model validation workflow for complex models such as partial differential algebraic equations (PDAEs). I briefly discuss how these frameworks and the open-source software tool contribute to the broad DoE technique paradigm and applications. I demonstrate the tractability and scalability of the frameworks with laboratory and pilot-scale carbon capture experiments. Moreover, generalized open-source software tools are developed and applied to carbon capture experiments, highlighting their versatility and practicality. This dissertation lays the groundwork for a sequential MO and MBDoE workflow that can be readily applied to various challenging problems in CBE and beyond, offering potential benefits to broader science, technology, engineering, and math (STEM) communities. I conclude with a discussion of the future directions, and provide preliminary works for some future directions as a starting point.