Recent Awards

Although mathematical and computational models are central to the development and operation of complex engineered systems, there are many open modeling challenges when we are faced with the scale and complexity of modern cyber-physical-human systems. In particular, engineering requirements often make incompatible demands on models. Detailed models facilitated by big data and high-performance computing make highly accurate predictions but are difficult to interpret and may over-explain observed phenomena. Coarse/reduced models are easier to interpret and offer robust explanations but at the cost of numerical accuracy. These competing demands are succinctly summarized as abstraction and fidelity.

This project uses data science tools to reconcile the competing demands of abstraction and fidelity for modeling cyber-physical systems in the age of big data. The motivating hypothesis of this project posits that systems-level behaviors can be categorized into distinct equivalence classes or phenomena. The project aims to develop a novel multi-model approach to understanding the relationships among models of the same phenomenon, each with varying levels of complexity. This approach involves identifying physically interpretable, abstract models and accurate, high-fidelity models of the same phenomenon. The next step is to earn the calibration curve that connects these different models. This two- step procedure aims to combine the benefits of both types of models: useful abstractions enabled by physical models and high-fidelity predictions of big data. The project will rigorously demonstrate the conditions for the existence of this calibration curve within an equivalence class. This innovative method addresses a fundamental challenge shared by all engineering disciplines, with a specific focus on energy conversion systems as its validation domain. The project also tackles three specific types of operational challenges in interconnected energy

networks.