demona Documentation

2-31s

Introduction

demona - decomposition enhanced mechanics optimized numerical analysis is a computational framework to study multiphysics problems. Interactions between different physics can be directly applied within the structure of demona. Matrix-free Newton-Krylov techniques along with multigrid and domain decomposition is embedded in to the solver. New solution algorithms are suggested for experimentation, as well.

Features

• Newton-Krylov: GMRES and BiCGStab
• Inexact Newton algorithm with adaptive forcing factor and line search backtracking
• Matrix-free: 6 different epsilon computation options
• preconditioning: jacobi and ilu(0) only (for now)
• Color based jacobian evaluation for 2d rectangular grids (5 point stencils only)
• Finite difference, finite volume discretizations

• Real time, pseudotime and steady state solutions

Future Work

Following methods are planned to be implemented:

• freezed jacobian
• diffusion jacobian
• diffusion + freezed jacobian
• low order approximation to high order nonlinear system
• multigrid
• numerical jacobian
• continuation
• fokkema vorst: store resulting F vectors
• singular systems
• finite element - ebe

Version History

Version 0.20-31 (Dec 1, 2009)

• sets introduced

Version 0.20 (Oct 5, 2009)

• color based jacobian evaluation is generalized for fv and fd
• hole option is added and checked only for fv (bonus check with ghost cells)
• high order - low order jacobian options added in fv
• number of ghost cells option is added (color based jacobian only)

Version 0.19 (Oct 2, 2009)

• Time string length issue fixed. character length is increased from 5 to 15
• initialValues (constants only) can be set within physicsInfo .
• for the case (nx < ny), error in colorbased jacobian calculation is fixed

Version 0.18 (Sep 30, 2009)

• Transient problems are tested

Versions 0.1-0.17 (Sep 28, 2009)

0.1 based on kNaN version 1.3
Version 0.1 - 0.17. Basic implementations of preprocessors,processors and postprocessors.
Domain decomposition is tested for nonlinear problems with success.

Developer

Erhan Turan turane@gmail.com