Trying with other geometries

[ifd-student]

trafficstars

Hello I tried this solver with geometry and BC of pitzDaily from rhoPimpleFoam, it seems it does not work with this a little bit more complex geometry. I have changed mesh complexity and yet there are points where high pressure appears some where in the geometry (before flow development, attached photo) and do not let the air to flow within. Do you have any idea? have you tried this with a high quality mesh?

[unicfdlab]

Hi, this sounds very strange -- we tried that solver with very complicated cases (such as Vega rocket lift-off) and everything was OK. Can you attach here the whole case? And which OF version do you use?

[ifd-student]

Yes, sure. Thanks in advance. pitzDaily_pimpleCentralFoam.zip

[unicfdlab]

Hi, I fixed several mistakes in your case.

I. Serious mistakes

1. LTS was not designed to work with relaxation, so I commented all relaxation settins:

relaxationFactors { // fields // { // p 0.1; // pa 0.1; // alpha 0.1; // pFinal 0.1; // } // equations // { // U 0.7; // Ua 0.7; // h 0.5; // } }

2. Boundary conditions are wrong -- you have constant area inlet and p_inlet/p_outlet > 1.5, thus the flow should be trans- or supersonic. In this case you must set p,U,T at the inlet patch and zeroGradient for all fields at the outlet patch. If you want to simulate flow under pressure difference, than you must change inlet geometry to provide contraction, as well as you need mixed type BC for outlet, for example subsonicSupersonicPressureOutlet from libcompressibleTools .

3. In LTS simulation, the time step has no physical mean - it is now the number of iteration, so I changed controlDict as follows:

endTime 5000; //0.6; deltaT 1; //0.2e-4; //now time step is the number of iteration writeControl adjustableRunTime; writeInterval 100; //0.0005; cycleWrite 0; writeFormat ascii; writePrecision 6; writeCompression off; timeFormat general; timePrecision 6; runTimeModifiable true; adjustTimeStep no; //yes; maxCo 0.5; rDeltaTSmoothingCoeff table ( ( 0 0.0001) (100 0.0010) (200 0.0050) //(300 0.0100) );

II. Minor corrections:

1. rDeltaTSmoothingCoeff is parameter which allows to control convergence of LTS. The larger rDeltaTSmoothingCoeff value, the faster convergence you will have, but the less stable process of calculation will be. It's an art to find rDeltaTSmoothingCoeff which gives you fast and stable convergence process.
2. It is better to use Minmod instead of vanLeer when you are not sure about convergence, Usually Minmod implementation in OpenFOAM is more stable than vanLeer.
3. Use PBiCGStab instead of PBiCG -- it gives better convergence for iterative solution of system of linear algebraic equations.
4. Use limitTemperature from fvOptions with caution, since sometime it only breaks energy balance instead of remedying problems with temperature boundness.

III. I would recommend to devise sketch for your dummy 2D case, which mimics basic properties of complex 3D model, including: objectives of the simulation, basic assumptions, equations, BCs, ICs, condition of simulation termination, geometry, etc

pitzDaily-pimpleCentralFoam-LTS.tar.gz

[ifd-student]

Thanks a lot for your help.