Toolkit to Support Parallel Adaptive Computations on Unstructured Meshes

Abstract

Report developed under STTR contract for topic ARMY02-T003. The initial development of a software toolkit to support parallel adaptive numerical analysis is described. The tools are focused on unstructured adaptive techniques which are among the most difficult to implement in parallel. To support dynamic load balancing software was developed to capture and utilize a hardware model, which represents the set of computing resources available for a given problem (analysis) at hand Mesh refinement software was extended for parallel operation, by adding mesh migration steps that enable modifications to be made to mesh entities lying on partition boundaries. The resulting capabilities were demonstrated in a parallel adaptive Navier-Stokes analysis. A new methodology for solution transfer, which combines the update of solution information with the mesh modification, was also implemented and compared to a global solution transfer procedure used in a commercial finite element code. The results showed that the new procedure leads to smaller relative difference in strain norm. Methods for reducing the storage required for adaptively evolving meshes were also investigated. An approach based on storing differences between successive mesh iterations showed that a significant reduction in file space is feasible.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 04, 2003
Accession Number
ADA413057

Entities

People

  • John Tourtellott
  • Mark Beall
  • Mark Shephard
  • Saurabh Tendulkar

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Bandwidth
  • Boundaries
  • Computational Fluid Dynamics
  • Computations
  • Construction
  • Dynamic Loads
  • Environment
  • Heuristic Methods
  • Hierarchies
  • Iterations
  • Load Monitoring
  • Mathematical Analysis
  • Migration
  • Numerical Analysis
  • Parallel Computing
  • Parallel Processing
  • Simulations

Fields of Study

  • Computer science

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Graph Algorithms and Convex Optimization.
  • Software Engineering.

Technology Areas

  • Space