Simulations of fluid dynamics, such as the flow of air over an airplane wing, are vital for understanding the fluids, but cannot be computed fast enough for humans to interact with the simulation. This project will develop an efficient, accurate, integrated environment for the simulation, visualization, and analysis of time-dependent turbulent flow fields using advanced physical and numerical models. The investigators will create low-order (relatively cheap and fast) approximations to an existing Navier-Stokes code, and use the approximations to provide interactive explorations of the simulation database while the main code continues to run. They will also add a graphical user interface to the existing code to allow remote visualization. The key subtasks in this process will include:
* Developing physically-based models, called surrogate models, of the compressible Navier-Stokes equations that sharply reduce the cost of the simulation yet capture the dominant flow structures.
* Using the surrogate models to visualize particle paths, streaklines, and pathlines interactively while the full simulation continues.
* Using the flow features to assess the error of the surrogate models.
* Developing efficient parallel algorithms for the surrogates, and assessing more advanced strategies for incrementally constructing and adapting them to simulation data.
