How to Fix Some of Our Errors Via Data Science

Over the past two decades, multiscale modeling has advanced tremendously, and several algorithms currently exist. Yet, our ability to apply first principles modeling to process design is seriously limited due to multiple challenges. One of the challenges relates to error of first principles calculations, reaction mechanisms, active sites, etc. and its impact on catalyst discovery and reactor design. In this talk we will discuss sources of errors and new statistical learning methods that are transferable and explainable to quantify and correct some of the errors. We will discuss data quality and probabilistic graph models as a framework that integrates data, knowledge, and physical models to enable uncertainty quantification.