ITR: Simulation of Flows with Dynamic Interfaces on Multi-Teraflop Computers

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

This Project will develop advanced parallel algorithms and software for simulating complex flows with dynamic interfaces. The development of scalable, parallel high-accuracy algorithms for simulating such flows poses enormous challenges in computational science. The project will use these algorithms for microstructural simulation of blood flow. This application provides an excellent testbed for the methods: it is extremely computationally challenging and of critical medical importance. Blood flow belongs to a class of flow problems with dynamic interfaces. Blood is a mixture of interacting gel-filled solid sells and fluid plasma. Current blood flow models are macroscopic, treating the mixture as a homogeneous continuum. Microstructural models resolve individual cells deformations and their interaction with the surrounding plasma. Because of the computational difficulties of resolving tens of thousands of dynamically deforming cells, no one to date has simulated realistic blood flows at this level. Yet such simulations are necessary in order to gain a better understanding of blood damage - which is central to improved artificial organ design - and for the development of more rational macroscopic blood models. Simulating flows with dynamic interfaces is much more difficult than flows in well-understood fixed domains. The central challenges are to develop numerical algorithms that stably and accurately couple the moving fluid and solids, and geometric algorithms for computing the resulting dynamic meshes. This project takes the approach of treating both fluid and solid domains as collections of grid points, with associated meshes, that evolve over time and devising numerical algorithms that couple the domains seamlessly. It will attack the difficulty of creating and managing the evolving mesh by developing scalable parallel algorithms for the convex hull, Delaunay triangulation, and mesh partitioning components. With careful attention to fundamental algorithmic issues, these cheap geometric computations will enable these dynamic flow simulations to scale to thousands of processors as on mult-teraflop systems. This research will benefit a wide community of scientists and engineers. The computational algorithms will be widely applicable to a variety of fluid-solid and fluid-fluid interaction problems. More generally, the core parallel computational geometry kernels will provide generic support for the geometric computations underlying many dynamic irregular problems. The project will distribute a portable library of efficient implementations of these algorithms. Also, the project will undertake a broad-based, interdisciplinary program integrating research and education. It will be part of a new program in Computational Science and Engineering, serving as the archetype of how applications, computational, computer, and mathematical scientists can work together to tackle societal problems that cannot be solved solely by any one discipline.