Couette/Poiseulle flow over a customized surface

Dear all,

as the thread topic says, I’m trying to simulate a simple 3D Couette flow over a customized surface, that shall be imported into the simulation domain. This can be done either by importing an STL geometry file with Guo BCs or by using a bool mask with a simple bounce-back BC.

When I place the imported STL geometry at the origin of the simulation domain, some strange effects occur and the calculations seem to be wrong. I have tested different positions of the geometry and figured out that those effects occurs only if the imported geometry touches one ob the bounding box sides. So I’m looking for the source of this undesirable disturbance and how to overcome it. Could it be, that those effects result from the fact, that Guo BC needs an extended neighbor relation and it somehow affects the calculation? Does anybody have any experience with it?
When I use the bool mask for the geometry description this effect does not occur but for my purposes the bounce back BC is not sufficient.

I would appreciate any advices


Does nobody have an advice?
I still appreciate your help


Dear Hanseato,

I do not fully understand the error you are describing, but what you say is true. The geometry should not touch the sides of the simulation domain. This is why there is a parameter “margin” which needs to be passed to the geometric processors, and this should at most of the cases have a value >=1. For more on how to use the Guo off-lattice BC please look at the:




Dear Dimitris,

thank you for your reply. Actually, I used the aneurysm and the external flow around obstacle examples to became familiar with the Guo off-lattice BC within Palabos. I don’t really understand the purpose of the parameter “margin” and how it affects the simulation.
Maybe, a sketch of the domain will be helpful. Here is a picture of the simulation domain:

The red arrow represents the moving direction of the top lid ( x-direction). The boundaries a,b,c,d (y- and x-directions) assumed to be periodic.
As you said, the geometry is not allowed to touch the domain boundaries. But if I shift the bottom plate by e.g. 1 latice unit inside the domain, the step effect occurs, so the pressure distribudion as well as the velocity field does not agree with the analytical solution.
Any suggestions how to solve this problem?

Best regards