[fjtb]

Hi guys, I'm modeling a boundary layer flow from a wind tunnel test. It is stated in the CFX documentation that "if the MAX residual is more than one order of magnitude larger than your RMS residual, it usually indicates that the problem is concentrated to a local region." In my case this is true since I used a hybrid tet and hex mesh, with more dense tet meshes near my structure. I tried decreasing and increasing the timescale but still won't reach the 1e-5 RMS residual criteria. However, when I followed the advice in the documentation (for convergence problems due to global effects) by changing the advection scheme originally from High Res to Specified Blend Factor starting with 0.75 and gradually increasing to 1, it reached the convergence criteria, although there are jumps in the residual plot immediately after I increased the blend factor.

So my question is, can I now accept this result ending with a specified blend factor of 1 or should I continue to run the simulation but this time, using a High Res advection scheme? Thanks in advance to those who will answer.

[ghorrocks]

First of all please note there are many, many things to consider beyond residual and advection scheme if you want to show your result is accurate. I assume you are aware of this.

If you got blend factor=1 to converge then you were using purely second order advection. This is pretty good. High res uses second order differencing but scales down to first order in areas where it judges it is required - so a high res result should be no better than a blend factor = 1 result which has converged.