Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Introduction to Finite Element Methods

David ONeal oneal@psc.edu

Overview
Basic Concepts Typical Element Geometry Nonconforming Element Geometry Computational Phases Parallel Solution Methods

Basic Concepts
General technique for constructing approximate solutions to boundary value problems Rigorous theoretical foundation Handles complicated geometries and general boundary conditions Systems of linear or nonlinear equations

Typical Element Geometry


Nodes at corners Shape functions Mesh Global matrix

Nonconforming Element Geometry


Nodes at mid-edges Shape functions Mesh Global matrix

Computational Phases
Construct the mesh Compute element matrices and vectors Patch together element approximations into global system of equations Incorporate boundary conditions Solve the global system of equations Present solution

Parallel Solution Methods


ANSYS Domain Decomposition Solver

A tetrahedral mesh was applied to carrier with the volume meshing command (VMESH). Element size was controlled through the use of the SIZLVL argument to SMRTSIZE automatic mesh generator option. Ten test cases were created.

SGI/Cray O2K Elapsed Time


7,000 6,000 5,000 4,000 3,000 2,000 1,000 0 1 2 4 8 8 4 2 1 OpenMP Threads DDS Elapsed Time (seconds)

MPI Processes

16

32

64

128 128

64

32

16

Example c10: DDS Elapsed Time


248,095 Elements; 370,493 Nodes; 1,111,479 DOF

SGI/Cray O2K Memory Usage


14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 128 64 32 16 16 128 32 64

Total DDS Memory Usage (MBytes)

MPI Processes

OpenMP Threads

Example c10: Total DDS Memory Usage


248,095 Elements; 370,493 Nodes; 1,111,479 DOF

SGI/Cray O2K Total CPU Time


30,000 25,000

20,000 Total CPU Time (seconds)

15,000

10,000

5,000

0 128 64 32 16 16 32 64 128 OpenMP Threads

MPI Processes

Example c10: Total CPU Time


248,095 Elements; 370,493 Nodes; 1,111,479 DOF

SGI/Cray O2K Scalability


10000 Example c10 Example c09 1000 Example c08 Example c07 DDS Elapsed Time (seconds) Example c06 100 Example c05 Example c04 10 . Example c03 Example c02 Example c01 1 1 2 4 8 16 32 64 128

MPI Processes

Examples c01 thru c10: DDS Elapsed Time


Origin 2000

TCS Scalability

Parallel Finite Element Method

You might also like