CAREER: Software Frameworks to Enable Parallel, Adaptive, Multiscale Simulation of Turbulence

Rensselaer Polytechnic Institute, Troy NY

Investigators

Abstract

CAREER: Software Frameworks to Enable Parallel, Adaptive, Multiscale Simulation of Turbulence Computational Fluid Dynamics (CFD) is the numerical simulation of the behavior of fluids, such as the flow of air over a car body or airplane wing or the flow of water past a ship. This is a classic problem in computational science, with great implications for both science (understanding the fluids) and industry (improving performance of land, air, and sea vehicles). Important cases of CFD - particularly turbulence, where a broad range of scales interact - are still unsolved problems. This project will attack those problems using modern multiscale computational methods. The goal of this project is the development of computational frameworks necessary to efficiently simulate turbulent flows at different approximation and resolution levels. It will integrate existing physics-based turbulence models with modern multiscale methods, where subgrid-scale models are based on mathematical analyses of the discrete methods used to solve the underlying partial diffential equations. The project will develop a parallel and adaptive computational environment that automatically combines mesh refinement and coarsening with method-order variation to achieve high efficiency. The resulting framework will also support dynamic load-balancing, new a posteriori error estimation procedures, and appropriate data structures for high performance on a variety of modern computer architectures. Lessons learned from this framework will provide insight into frameworks for other complex phenomena. The framework will also enable rich educational rewards. Young children are fascinated by interactive computer simulations of the world (e.g. why do airplanes fly, or baseballs curve); the framework can be packaged to let them play with these concepts, and eventually attract them into a career in science and engineering. At the college level, the framework can be used to explain fluid mechanics concepts to engineering students, let CS students analyze communication behavior in a real application, and illustrate numerical analysis to undergraduate and graduate students.