Parallel Least-Squares Finite Element Method for Large Eddy Simulation of Large Scale Environmental Flows and Transport ProcessesEPA Grant Number: R825200
Title: Parallel Least-Squares Finite Element Method for Large Eddy Simulation of Large Scale Environmental Flows and Transport Processes
Investigators: Tsang, Tate H. , Yost, Scott A. , Bai, Zhaojun
Institution: University of Kentucky
EPA Project Officer: Klieforth, Barbara I
Project Period: November 1, 1996 through October 31, 1999 (Extended to November 14, 2000)
Project Amount: $475,670
RFA: High Performance Computing (1996) RFA Text | Recipients Lists
Research Category: Human Health , Aquatic Ecosystems , Environmental Statistics
Description:Numerical modeling is one of the indispensable tools to analyze complicated and challenging environmental problems. The success of numerical modeling depends on the formulation of a reliable turbulence model and the development of a robust and efficient numerical method which can simulate large scale problems with minimum use of computer memory. In this work, the Large Eddy Simulation (LES) technique and the Least-Squares Finite Element Method (LSFEM) are combined to simulate large scale environmental flows and transport processes. Two types of challenging, computationally intense problems crucial to our environment will be tackled: 1) time-dependent, three dimensional turbulent flow and dispersion of hazardous/toxic air pollutants around buildings and industrial complex; and 2) time-dependent, three dimensional turbulent flow and dispersion of contaminants in Green Bay.
A robust and efficient matrix-free least-Squares finite element method has been used to simulate moderately large scale three dimensional fluid flows and transport processes. For example, 1.1 million unknown flow variables at 270,000 nodes have been found by the LSFEM. Parallel LSFEM has also been developed for convective transport of 11 reactive species with more than 1.9 million unknowns on 173,225 nodes. Linear speedup has been achieved for this simulation on a Convex Exemplar parallel computer. In addition, the memory requirement for the matrix-free LSFEM is at least 20 to 30 times less than other methods.
The goal of this research is to extend and modify our LSFEM/LES codes and to develop highly parallelizable algorithms to simulate flows and transport processes in domains with 1 to 10 million nodes. The results of this research should find numerous applications to large scale simulation of environmental flows and transport processes in domains with complex geometry.