Adaptive Grid Generation


Adaptive Grid Generation by the Deformation Method
Dr. Guojun Liao and Dr. Gary dela Pena

Home
Outline
Numerical Examples
Moving Finite Difference
References
UTA MATH Home Page

Copyright © 1999 by Dr. Guojun Liao and Dr. Gary dela Pena

Simulate u, where u is a scalar or vector function, satisfying

where L is a differential operator, on a domain D in .

An Initial (uniform) grid on D is deformed in real time to an adaptive grid according to a monitor function f ( x, t ).

Equidistribution Principle is used to construct the monitor function. A posteriori error estimates, residues and truncation errors, etc. are redistributed evenly over the whole domain. In most cases, we want to put refined grids in the regions where u changes rapidly.

For flow patterns with shock waves, we can take

where p is the pressure. In general, terms involving the values of u and the second derivatives of u (or curvatures) can also be included.

For interface resolution, we could construct f by using the level set function d as follows: Let f be piecewise linear such that

Required properties of grids: Non-folding, volume (cell size), smoothness, orthogonality and aspect ratios.

Evaluation of grid generation methods: Higher dimensions? Complex Geometry? Easy to implement? Work well with various numerical methods such as finite difference/volume, finite elements, the method of lines, particle methods, etc.? Work well for general equation(s), not problem dependent?