[8-13]An Algorithmic Framework for Solving Systems of PDEs from Multi-physics Applications on Parallel Computers

Date:2009-08-10

Title：An Algorithmic Framework for Solving Systems of PDEs from Multi-physics Applications on Parallel Computers
Speaker：Xiao-Chuan Cai
Time：10:15-10:55 am, Thursday, August 13
Venue：Room 334, Level 3 Building #5

Abstract：

Mature technologies are available for solving many types of single physics problems, but for coupled multi-physics problems, robust and scalable techniques are badly needed, especially for large scale parallel computers. The focus of the talk is on some new domain decomposition based nonlinear preconditioning techniques for the numerical solution of some highly nonlinear, coupled systems of partial differential equations (PDEs) arising from multi-physics applications. These PDEs often represent multiple interacting fields (for example, fluid and solid), each is modeled by a certain type of equations. Current approaches usually involve a careful splitting of the fields and the use of field-by-field iterations to obtain a solution of the coupled problem. Such approaches have many advantages such as ease of implementation since only single field solvers are needed, but also exhibit disadvantages. For example, certain nonlinear interactions between the fields may not be fully captured, and for unsteady problems, stable time integration schemes are difficult to design. In addition, when implemented on large scale parallel computers, the sequential nature of the field-by-field iterations substantially reduces the parallel efficiency.
To overcome the disadvantages, fully coupled approaches are investigated in order to obtain full physics simulations. The success of such a fully coupled approach depends heavily on a nonlinear algebraic system solver that is robust and scalable. Unfortunately, traditional nonlinear iterative methods do not work well, for example, Newton-like methods often converge very slowly because of the existence of local non-smooth components in the solution and the lack of good initial guess. The new algorithms are motivated by the nonlinear preconditioning methods recently introduced and the scalability is obtained by incorporating multigrid methods into the algorithms. Several important applications will be discussed.

Biography：
Xiao-Chuan Cai received his PhD from Courant Institute of Mathematical Sciences, New York University, in 1989. His research interests include parallel algorithms and software for numerical solution of partial differential equations. He is currently a Professor in the Department of Computer Science, University of Colorado at Boulder.