mathias
Forum Replies Created
-
Posts
-
You need to define the right density at the boundaries (defineRho or for bounce back in the constructor), i.e. 1 for attracting and 0 for repelling. Best Mathias
I use Paraview version 4.0.1 64-bit Linux.
Dear simonguo,
which paraview versiona re you using. I works fine on my comuters with not the latest version of paraview. I guess it is rather a bug in paraview than a problem of the maschine. Try with a older version and please report if you face the same problem.
Best
MathiasDear applous6,
since you created a pointer you need to use cuboidGeometry->setPeriodicity(true, true, true). We strongly recomment our spring school in March 2017.
Best
MathiasDear Ivan,
basically for multi-phase/-component flows you can revese the setting of inlet and outlet as it is done in cylinder3d, i.e. at the inlet you fix the density of both phases by a pressure bc (addPressure..) and at the outlet you fix the velovity by a velovity bc (addVelocity).
Best
MathiasThere is the CFL condition. But there is also a lot of experiance due to boundary conditions, the LBM modell, the actual flow domain, …
SuperLatticePhysDrag3D computes the drag coefficient. Best Mathias
Dear Alina,rnrnsome years ago we simulated the flow around a falling ball (cf. http://optilb.org/openlb/others 3d falling ball). We got good results for Re=10,100,1000 and 5000 with D3Q19 BGK-Boltzmann model and also for Re=10000 and 100000 with D3Q19 Smago BGK-Boltzmann model.rnrnMaybe your boundary conditions are not right or your resulotion is not fine enough.rnrnBestrnMathias
Dear Alina,rnrnyou have chosen CharL=0.2m, CharU=0.01m/s and CharNu=8.9e-07 m^2/s in the Converter. Thus, you get a Re=2.247.19. Your Re=10.000 does not enter the code anywhere.rnrnTo get stable results you need to choose much smaller grids and time steps or you can use a turbulent model.rnrnBestrnMathias
Hi,rnrnyou can look for publications of Li-Shi Lou https://www.odu.edu/directory/people/l/lluo . rnrnBestrnMathias
Hi,rnrnOpenLB is for sure a good choice using many cores. I am running simulations with more then 2000 cores and the parallel performance is great.rnrnBestrnMathias
Hi Alina,rnrnso you found a solution?rnrnBestrnMathias
Hi Ivan,rnrn1/ For details on the “”Random”” function, please, check out the source code http://optilb.com/DoxyGen/html/d8/d3a/classolb_1_1AnalyticalRandom1D.html rnrn2/ No, 0.98 is not related to gravity. I guess it comes from a benchmark paper.rnrnBestrnMathias
Hi jackyko,rnrnWe do not have a GPU version in OpenLB. However, OpenLB performs excellent using MPI on cluster machines of hundreds or thousends of CPUs. Have a look! rnrnBestrnMathias
Dear Vijay,rnrnOpenLB dont have a non-uniform mesh implementation right now.rnrnBestrnMathias