[arsenis]

Hi to all,

I use OpenFOAM 2.3.1 to simulate with LES the pollutant dispersion inside street canyons. A modified pisoFoam solver (to include the convection-diffusion equation) is used.

At the beginning I used cyclic boundaries, but changed them to mapped, because I don't want to recycle and thus capture the pollutants inside the domain. I know that there are tools in swak4Foam and in newer versions (e.g. fixedValueConstraint in 4.1), to use cyclic and don't recycle the pollutants. Unfortunately, I am not familiar with the first (do I need to install it or just compile it) and any newer OpenFOAM versions won't be installed in our cluster for the near future.

I have a question about the mapped boundary conditions in OpenFOAM:

Should all the used pisoFoam properties (i.e. U, p and nuSgs) be recycled through mapping, or just the velocity U?

Reading in the forum and searching in Google didn't give a specific answer, so I present some of my results, so anyone may observe something that I have not.

In the pitzDailyMapped tutorial in the OpenFOAM installations, the U, k and nuTilda are recycled, while p and nuSgs are not. So, I have tested both cases:
a) recycle only U and
b) recycle U, p and nuSgs
For some geometries both cases have worked, with nearly identical results and both can be accepted as validated.

However, for the case b, an irregular flow develops at the inlet. The following images present this irregularity. The pressure is shown, as it is more visible than velocity, from random time steps of the studied cases.

Street canyon with aspect ratio 1/1. The irregularity is barely visible and no problems are caused in the simulation (over 2000 seconds).


Street canyon with aspect ratio 1/2. The irregularity is fully visible, but no problems have been caused in the simulation (after 25 seconds).


Street canyon with aspect ratio 1/3. The irregularity is fully visible and now the simulation crashes after ~6 seconds. A weird recycling is also observed at the inlet patch and the velocity increases constantly in the same location.
the pressure p

and the velocity U at the same time step

Another image of the same problem, in some of my attempts.


All the above flow fields have been obtained with cyclic boundaries in the beginning and then the solution was mapped (mapFields utility) in a domain with mapped boundaries. In all the above cases, the setAverage option was set to false. I have also tested variations on the offset of the mapped patches, with no change in the results (e.g. for the width of 40m, the 39, 39.5 and 41m were tested).

For the case of recycling only the velocity U, when the initial flow field is mapped to a domain with the setAverage option set to off, a significant drop in U is observed with no other problems. When mapped to domain with the setAverage option set to on, the solution crashed. I suspect that for this option, the mapped patch expects a certain velocity profile.

On the other hand, I am currently obtaining a new solution (recycling only U and the setAverage option on) from the beginning and I don't have any problems.
I must clarify, that I used this method, when started to use mapped boundaries with no problems. I temporarily abandoned it, because I cannot fully understand from the code what is the role of setAverage (not enough programming knowledge!). Which of course will be one the uncertainties again, but in any case the results obtained in this way are also validated.

Thank you,
Arsenis Chatzimichailidis

[kerim]

Dear Arsenis,
I know it's been a long time, but I would like to know about your progress on this issue. Could you share with your successes? Any kind of assistance is highly appreciated.
Cheers,
Kerim