Data-Driven Digital Twin Development

Our team and collaborators are among the first to work on this cutting-edge technology for diagnosis, prognosis, and strategy assessment. “Diagnosis” is key to planning decision support. “Prognosis” (the ability to foresee potential failures) is key to cost-effective operation and improving the resiliency of complex infrastructure systems, such as nuclear plants, oil, and natural gas facilities, chemical plants, hostpitals, etc. Artificial Intelligence/Machine Learning (AI/ML) enabled Digital Twins (DT) to facilitate both diagnosis and prognosis. Conceptually, a DT is a virtual replica of a specific infrastructure system that “assimilates” its operating and performance history. This history, together with appropriate models, helps us determine the state of degradation on key components, as well as its expected response to operating stressors and natural hazards. One key challenge in the development of a DT lies in the ability to automate the collection of meaningful data to train the DT, and the flexibility to obtain new data as the need arises. There is very little recorded data for actual emergency and accident scenarios, and the training relies heavily on advanced simulation tools. An effective DT must continuously assimilate a large amount of current data on a system’s operation conditions and must age with it.