Hi,

Does somebody have experience with turbulent flow past rough walls? The roughness that i am setting is quite high (1.5 mm in a 32 mm diameter pipe) and i cannot get the standard velocity profile (when non/dimensionlized with the friction velocity). The yplus from the simulations does not seem to be sensitive to the first cell height when the roughness is so high (i have been varying the cell height from 0.4 to 1.6 mm and the yplus is around 370, varying very little).

why did not you try to set first layer height less then 0.4?

because i have set the first cell height at three different levels: 1.6, 0.8 and 0.4, which is higher and lower than the roughness height and the value of y+ did not change much (as i said around 370+-10), so i thought i need some help with that. What do you think will change if i set it below? Do you know what the problem is?

I think what you should do is to keep the y plus in the range of 1-10 for this typical fluid dynamics problem to get something match with results you are expecting, its posssible only when you use prism layer with a first cell height of 0.02 mm. I think this is a query which should be answered by CFX development team, how can we keep both high reoughness and Yplus in reasonable range

Thank you! Why do you think such a small y+ will do a better job? In the CFX mathematical formulation of wall function/roughness it is recommended not to have the first cell height thinner than roughness height (with k-omega model). I am using k-epsilon and don't know what applies to this case. I did ask the CFX support and did not get any good answer on that. What i did get was another possibility of implementation of user defined wall function. But there is a problem with that too: the coefficient must be tuned for the desired value of the wall shear stress (since it is impossible to predict it), which is a lengthy process and is not unic, since i have to do simulations with different roughness and Re numbers. Do you know the solution to that maybe?

Finally, I got to know from CFX support team that in the case of rough walls, the wall virtually is moved to half-roughness (to avoid singularities at the wall), which means that it is useless to refine the mesh (first cells from the wall) to the value below half-roughness. However, I am not sure how this (in case of essential roughness) affects the results.