Parallel Scalable Domain Decomposition Methods for Blood Flow Simulation

AbstractIn this work, we develop a class of parallel domain decomposition method forsimulating blood flow in three-dimensional compliant arteries. We model thefluid-structure interaction problem by using a monolithically coupled systemof linear elasticity equation and incompressible Navier-Stokes equations inan arbitrary Lagrangian-Eulerian framework. The monolithic fluid-structuresystem is discretized with a fully-implicit finite element method on unstructuredmoving meshes, and solved by a Newton-Krylov algorithm preconditionedwith restricted additive Schwarz methods. The investigations focuseson the performance of the one-level and two-level overlapping Schwarz preconditionerfor Newton-Krylov methods as well as the efficiency and parallelscalability of the algorithm for solving the complicated coupled system.Simulations based on the patient-specific pulmonary artery geometries areperformed on a supercomputer. Our algorithm is shown to be scalable withthousands processors and for problems with millions of unknowns. Yuqi Wu is an Acting Assistant Professor in Applied Mathematics at theUniversity of Washington. He received his Ph.D in Applied Mathematicsfrom the University of Colorado. Dr. Wu¿s research focuses on numericalalgorithms for partial di¿erential equations, computational fluid dynamics,parallel computing and reduced-order models.