Hi,

Can anyone explain me why some research codes uses a dual grid structure - a combination of staggered and nonstaggered grid ???

The code does use dual grid structure and also allow composite grids (multiblock).

For this seems like just asking for trouble.

But perhaps someone can explain this to me???

Thanks in advance.

Regards

Jens

I have come across schemes that used mixed staggered/non-staggered layouts although I am not fully sure they are what you are interested in. These methods are generally used for a FV approach.

In these methods, the cartesian velocities (u,v,w) are computed at the cell centers and the contra-variant velocities (U,V,W) are computed on the cell faces. So you need to compute 6 velocity variables per grid cell. One set of velocity components are used to compute the source term for the pressure Poisson equation. Since the pressure term couples the momentum and the continuity equations (i.e., pressure gradient enforces the mass conservation on the momentum transport equations), this set of velocity are used for mass conservation. The other set of velocities are used to attain momentum conservation.

Some people do not like these methods since (in addition to double the cost involved in solving for 6 variables), different set of velocity variables are used to enforce mass and momentum conservation respectively. As you may know, the non-staggered grids lack strong coupling between pressure and velocity fields (unless upwind stencils are used to couple velocity and pressure). Lot of complicated (multi-dimensional) interpolations/extrapolations are needed to couple velocity and pressure on curvilinear grids. Staggered grids on the other hand do not have this problem. Staggered meshes that use cartesian velocities are not suitable for highly curved meshes (e.g. flow in bent or a U-shaped pipe) since the velocities defined on cell faces do not align with face normals (additional velocity components are required to compute fluxes). The staggered grids that use contravariant velocities have their own problems. If one has a slender (diamond) rhombus shaped grid cell, only one of velocity components (i.e., component along the longer diagonal) is not well represented. SO the mixed method was proposed.

Ref :

Zang, Koseff & Street : JCP (1994 or 1995)

Rajat Mittal (AIAA paper that can be downloaded from Dr. Mittal homepage at UF, Gainesville).

Hi again

Why is this kind of dual grid structure not used more ??

Since it has some very good properties !

I can not recall to have since in other codes, beside the Zang et al. papers.

Because of tricky programmning ?

Thanks again.

Regards

Jens