26.12.2013 Views

Adaptivity with moving grids

Adaptivity with moving grids

Adaptivity with moving grids

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Adaptivity</strong> <strong>with</strong> <strong>moving</strong> <strong>grids</strong> 9<br />

methods. The first group is referred to as velocity-based since the methods<br />

directly target the mesh velocity and obtain mesh point locations by integrating<br />

the velocity field. Methods in this group are more or less motivated<br />

by the Lagrange method in fluid dynamics, where the mesh coordinates,<br />

defined to follow fluid particles, are obtained by integrating flow velocity. A<br />

major effort in the development of these methods has been to avoid mesh<br />

tangling, an undesired property of the Lagrange method. This type of<br />

method includes those developed in Anderson and Rai (1983), Cao, Huang<br />

and Russell (2002), Liao and Anderson (1992), Miller and Miller (1981),<br />

Miller (1981), Petzold (1987) and Yanenko et al. (1976). The method of<br />

Yanenko et al. (1976) is of Lagrange type. In the work of Anderson and Rai<br />

(1983), mesh movement is based on attraction and repulsion pseudo-forces<br />

between nodes motivated by a spring model in mechanics. The <strong>moving</strong> finite<br />

element method (MFE) of Miller and Miller (1981) and Miller (1981)<br />

has aroused considerable interest. It computes the solution and the mesh<br />

simultaneously by minimizing the residual of the PDEs written in a finite<br />

element form. Penalty terms are added to avoid possible singularities in the<br />

mesh movement equations; see Carlson and Miller (1998a, 1998b). A way<br />

of treating the singularities but <strong>with</strong>out using penalty functions has been<br />

proposed by Wathen and Baines (1985). Liao and Anderson (1992) and<br />

Cai, Fleitas, Jiang and Liao (2004) use a deformation map approach. Cao<br />

et al. (2002) develop the GCL method based on the geometric conservation<br />

law (see Section 4). Similar ideas have been used by Baines, Hubbard<br />

and Jimack (2005) and Baines, Hubbard, Jimack and Jones (2006) for fluid<br />

flow problems.<br />

The second group of <strong>moving</strong> mesh methods is referred to as location-based<br />

because the methods directly control the location of mesh points. Methods<br />

in this group typically employ an adaptation functional and determine the<br />

mesh or the coordinate transformation as a minimizer of the functional. For<br />

example, the method of Dorfi and Drury (1987) can be linked to a functional<br />

associated <strong>with</strong> equidistribution principle (Huang et al. 1994). The<br />

<strong>moving</strong> mesh PDE (MMPDE) method developed in Cao, Huang and Russell<br />

(1999b), Huang et al. (1994) and Huang and Russell (1997a, 1999) moves<br />

the mesh through the gradient flow equation of an adaptation functional,<br />

which includes the energy of a harmonic mapping (Dvinsky 1991) as a special<br />

example. A combination of the MMPDE method <strong>with</strong> local refinement<br />

is studied in Lang, Cao, Huang and Russell (2003). Li, Tang and Zhang<br />

(2002) and Tang and Tang (2003) also use the energy of a harmonic mapping<br />

as their adaptation functional, but discretize the physical PDE in the<br />

rezoning approach.<br />

So far a number of <strong>moving</strong> mesh methods and a variety of variants have<br />

been developed and successfully applied to practical problems; see the review<br />

articles of Cao et al. (2003), Eisman (1985, 1987), Hawken, Gottlieb

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!