[Nicole]

Hello Foamers!

I am trying to solve for the steady state flow of slag in a furnace. I was having some issues so I returned to a simple case of a rectangular box, with one wall at 1700 degC and the opposite wall at 1500 degC, with a starting temperature of 1600 degC throughout.

I used buoyantBoussinesqSimpleFoam but it seems to generate very high velocities (1e5+) before starting to converge to a reasonable solution and even after 1000s of iterations it is still not converging well.

On the other hand, I tried buoyantBoussinesqPimpleFoam and it seems to be converging to a reasonable solution much much faster.

Has anyone had any similar experiences with the SIMPLE algorithm seeming to converge better/faster than PIMPLE? I have not had a chance to study Pimple much yet. I can post an example case, but thought I would just check if anyone has some quick ideas for me, or some literature I could look at.

Thanks!

[ssss]

SIMPLE solvers and steady solvers, and thus the time you are seeing in the output does not have a physical meaning

PIMPLE solvers are unsteady solvers.

Your case has a big difference in temperature and it might be easier to obtain a "steady" solution using PIMPLE, because it helps to create a diagonal dominant matrix and because your case might be highly unsteady

There are also other factors that might influence the case such as mesh, schemes and solution configuration etc

[Nicole]

Hi,

Thanks so much. The 'time' I was referring to is the time it took my computer to solve.

I think you are probably right about my case being highly unsteady, I am glad to hear that this is something that other people may have experienced before. That's interesting about PIMPLE generating more diagonally dominant matrices, I will look into that a bit more.

I still need to investigate various schemes etc. But thanks so much for the advice

[ssss]

Dear Nicole,

PIMPLE helps to generate a diagonally dominant matrix because of the implicit term that the time derivate operator adds to the diagonal of the matrix of the equation discretized with the finite volume method.

I'm glad that I could help you

[jhv_1729]

Hi Nicole,

In addition to above, I believe you are trying the wrong solver. buoyantBoussinesqSimpleFoam is the incompressible solver with effects of buoyancy added in. Rule of thumb for this solver recommended by OpenFOAM trainers is a deltaT of 40C. In the description, the deltaT is clearly 200C.

Why not try switching to the buoyantSimpleFoam solver? It is the compressible solver with buoyancy.

H.

[ssss]

Well I wouldn't really say the same that the OpenFOAM, trainers you are referring to.

I will be quick on it. The Boussinesq approximation can be found in the rhok term which reads

rhok=1-beta*(T-Tref)

The second term of rhok should be <<1 so that the boussinesq approximation maintains true. The order of magnitude of the second term is beta*deltaT and thus it depends on the properties of the flow.

Moreover sometimes even though beta*deltaT is "big" one could stay using the boussinesq approximation to obtain some first results. Although it is true that if the second term becomes big the convergence of the solver becomes difficult

[Nicole]

Hi Joshi,

Thanks for your inputs. ssss has pretty much answered how I would have. Except to add that the temperature differences in the actual fluid are much smaller than the boundary temperature differences due to the good mixing and heat transfer. So I think my simulation falls in the 40 degree deltaT range in any case.

I am not using the full buoyant solver because of issues that I presented here: http://www.cfd-online.com/Forums/ope...ed-volume.html