hi, does anyone has experience with solving creeping flow on unstructured triangular mesh with variables in CV centers? my question is: on a grid of reasonable quality (yet not perfect) - is it enough to take cell face centers values as avergage (ev. more clever weighted than 0.5) of corresponding CV centers values? is it enough for not neccesarily very accurate but *stable* calculation? proper 2nd order would be a lot of geometrical calculations... (intersections, distances etc.) regards dominik

Really the *stability* of a calculation depends on the behavior of the equations being solved. Fortunately creeping flow are diffusion dominant and the equations (I assuming Navier-Stokes) behave elliptically. By averaging the face-centers (using neighbouring cells) and using those face-centers to construct your cell-centered solutions -- this would be a central difference scheme, which is of course 2nd order accurate. Central difference schemes are unstable for hyperbolic equations because the information propagates predominantly in a single direction ... but since creeping flows behave elliptically, ie information spreads in *all* directions, a central difference *should* work.