The goal of the proposed research is to devise rigorous mathematical techniques that utilize large databases obtained by simulations to develop predictive models of turbulent flow. Even though the proposal is focused on turbulence, the tools to be developed will be of general applicability for data-driven modeling in other areas of science and engineering.
Predictive models have not yet taken full advantage of the massive amounts of data being generated by the fluids community. New strategies are needed to extract information and modeling knowledge from data. Adjoint-driven inverse problems will be invoked to extract relevant modeling information from data. An important aspect of this approach is that the data is processed in the context in which it is needed for prediction. Domain-specific machine learning techniques will be used to convert information to modeling knowledge. In essence, the inverse solution infers functional deficiencies in the model and machine learning is used to reconstruct the missing functional form. The co-PIs plan to investigate how to identify and formulate a properly-posed data-driven-turbulence-modeling problem, the implications that these approaches have in more general data-driven computational physics applications, and the most effective ways to use machine learning in a predictive physics setting. Applications to be explored include transition to turbulence, thermal transport, and near-wall turbulent stress closures. The proposed work is expected to result in improved closure models for Reynolds-Averaged as well as hybrid Reynolds-Averaged/Large Eddy simulations. While the focus of the proposed work is on turbulent flow applications, several aspects of the formulation and tools will be of more general value to the field of data-driven physical modeling. Educational activities that would integrate machine learning into fluid dynamics courses are proposed. Tools and technologies will be shared with the community and the industry
