This article will walk through some of the common issues that can arise when modeling wall roughness within Simcenter STAR-CCM+ under certain meshing and multi-phase conditions.
The inclusion of wall roughness in a Simcenter STAR-CCM+ analysis is something that happens quite commonly, but without special considerations during setup, it is possible to cause solver instabilities or obtain unphysical results.
Issue 1: Geometrical roughness versus sand grain roughness
The geometrical roughness height we have for materials may not be the correct value to input to the simulation. Simcenter STAR-CCM+ takes as the roughness input the equivalent sand grain roughness height, not the geometrical height which is normally what we have on hand. The equivalent sand grain roughness height (ks) is the sand grain height at which the friction factor is same as that of given roughness topography. There are correlations that relate geometric roughness height (k) to ks available in published literature, or the values can be determined by separate simulations with reference to Moody chart. So one of the first things we need to make sure of is that we are not specifying geometric height (like maximum roughness height or the root mean square of the topographical/geometrical values).
It is worth noting that this is potential culprit if you have a converged and stable solution, but the answer (such as pressure drop for instance) is not matching experiment or known data.
Issue 2: Roughness height versus near wall cell thickness
Sand grain wall roughness heights that are larger than the near wall cell thickness can cause instabilities or incorrect answers. This can be a problem when we are running cases that need or significantly benefit from low y+ values, thus needing a very small near wall cell thickness. We do have a way to place limits on the roughness height calculations in smaller cells with the "rough displaced origin" model now available for wall roughness specification. But this option is really a limiter designed to keep the run stable. Depending on the simulation this may lead to solution inaccuracies.
This leaves a gap in trying to establish both a pressure change through a flow path (which can depend on roughness) and a correct wall heat transfer (which is going to be most accurate with low y+ values and potentially very thin prism layers).
This is an area of active research to improve.
Issue 3: Roughness and near wall cell thickness in film applications
Given that wall roughness can have an effect on dispersed phase impingement behavior on film assigned surfaces, then we are likely to need to be careful with near wall cell heights for these applications as well. And that can call into question the relationship between the film thickness and near wall cell/roughness heights. If the film thickness grows larger than these items results can be questionable, and can even cause divergence.
Aside from mesh changes, which can help stabilize things if the film thickness grows too large, we can also apply the film to VOF transition upon hitting certain volume fractions within cells, which will allow any cell that is "filled up" with film to transition to VOF physics, which does not have the restrictions film does when reaching a certain amount within a cell.