Instability of D3Q27 lattice

Palabos Bug Reports
#1

Hello,

I have tried using the D3Q27 lattice in Palabos with a geometry read from an stl file, and the results always become unstable very quickly, even if I set Nu to be about 0.8. I have tried using a pressure boundary and a fully developed velocity boundary in a straight pipe. Looking at the results in the 100 or so iterations before the instability occurs shows that the flow never looks realistic. The same case run on the D3Q19 lattice works fine and gets good results.

Regards,
Alex White

#2

Hi,

Thanks for the report. Any way we can reproduce this error? Do you observe similar problems with any of the Palabos examples? Or is the STL file sufficiently small that you could e-mail it to us, together with the code?

Thanks in advance.

#3

Hi,

Changing the lattice descriptor at the top of the file in the fullDomainFromSTL file from D3Q19 to D3Q27 causes the file to become unstable after about 150 iterations. Doing the same in the cavity3d file also causes it to become unstable, but after about 2000 iterations. All of this was done using Palabos version 0.6 release 2 and I just changed D3Q19 to D3Q27 in all cases (if I should have done more then the error is my fault).

Regards,
Alex

#4

Hello,

I’ve just run the cavity3d code on a D3Q27 lattice, with Palabos version 0.7 release 0; it is stable, and the result makes sense, after 4000 iterations. Although I don’t understand why you get instabilities with older versions, may I suggest that you simply upgrade to the current release and let us know if there are still problems?

#5

Hello,

Thanks for your advice, upgrading did fix the problems with the instability on the cavity3d case. However in version 0.7 release 1 I can’t get the readSTL tutorials (which a lot of my code is based on) to compile. The first error I get is that ‘MultiBlockDistribution3d’ is not declared in the library interface file ‘CVMLCPP_sparseMultiBlock3d.hh’. There are other errors given as well which may or may not be caused by the original error. I can give the whole terminal output if it would help.

Thanks,
Alex White

#6

Hi,

You are right, this was an oversight. A new release (v0.7r2) is now online which fixes this compilation error.

#7

Hi,
wel I m working on a readSTL examples and I want to change the aneurysme of the example by an other one one,
But it is difficult to me to set the inlet and outlet.
l didn’t get how is the mecanisme to set them when I took the coordinates of the inlet from Paraview it gives me results zero.
It will be grat if there is some bdy to help me.

Thanks

#8

Hi,

The coordinates of the inlet and outlet are the locations on the lattice that the inlet and outlet occupy, i.e. if the centre of the inlet is 1/4 of the way up the total ‘x’ length of the geometry and 1/5 of the way along the ‘y’ direction and you discretise the domain into 100 lattice units in each direction then the coordinates of the centre of the inlet will be [25,20]. The inlet specified needs to cover the entire area of the inlet, but it is fine to exceed it. You also need to be careful that the inlet is facing the right was and the ‘beforeInlet’ cells are actually before the inlet and not after, which will be the case if you change the direction of the inlet from what it originally was.

Hope that helps.
Alex