Experimental Design, Uncertainty Quantification and Decision Making for Complex Systems

Modern engineering and scientific research often involve carrying out experiments with complex systems to solve critical decision-making problems. For example, a complex system characterizing a manufacturing process can be used to determine the optimal material choices and manufacturing procedures in different stages. Since these complex systems may include large-scale computer models and expensive experiments across multiple platforms, it is critical to provide high-quality decisions under limited experimental resources. This award supports fundamental research on an experimental design and uncertainty quantification framework for utilizing structural information of complex systems to address decision-making problems. This framework has the potential to lead to significant improvement in the efficiency and effectiveness of learning and decision-making of complex systems in various engineering and scientific fields. The computational outcome will be made available as open-source software. Additionally, the results from this project will be used in educational activities to demonstrate how experimental design can enhance efficiency in information collection and decision-making for future engineers and scientists. Motivated by real applications, the project will investigate three specific types of structural information that appear in complex systems: (1) systems that include coupled and/or multiple experimental platforms; (2) systems that exhibit scientific connections among different outputs; and (3) systems that consist of a combination of subsystems characterizing different stages of a complex process. This project will incorporate these types of structural information into statistical surrogates and data acquisition for new sequential experimental algorithms to extend Bayesian optimization approaches developed for “black-box” systems. Theoretical results on optimal experimental design strategies for different components of the complex systems will be established to guide sequential data acquisition. The project investigator also plans to develop new decision uncertainty quantification methods to effectively assess the quality of the decisions. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.