Problem with fallingDrop3d

[Danial.Khazaeipoul]

Dear developers,

I am experimenting with the free surface implementation in OpenLB and I noticed that in the “fallingDrop3d” example, the volume fraction values becomes negative during the simulation, i.e., it must remain between 0 and 1.0. To reproduce the problem, you just need to run the mentioned example from version 1.7r0 on either cpu or gpu.

Is this a known bug? Because it is not limited to this specific example and I am also seeing negative values in the verification example I am trying to develop and share with the community.

Would appreciate your comments on this issue.

Regards,
Danial

[mathias]

Are the value below zero far from zero or close to zero? Does that course problems? If needed you can cut off them by setting values smaller than zero to zero.

[Danial.Khazaeipoul]

“If needed you can cut off them by setting values smaller than zero to zero.”

I agree that if these negative values are close to 0, a cutoff can be applied to rectify the problem. However, these negative values are not small compared to the physical limits of the volume fraction field. In the “fallingDrop3d” example, you can observe values of -4.6 and -5.3 for the second and third written steps when viewing the results in Paraview. Later in the simulation, the negative values start getting smaller and closer to zero.

In the verification example I am working on, rising of a single bubble in a column of liquid, the negative values remain far from zero, -8 to -12, for the entire simulation.

” Does that course problems?”

Unfortunately, I did not understand this part of your answer.

[mathias]

”Does that course problems?” -> is the simulation getting unstable? A cutoff aproach might help..

[Danial.Khazaeipoul]

Thank you for the explanation. I need to say that the simulation runs fine without diverging or unexpected behavior. However, I am not sure if the obtained results are reliable when the volume fraction field is wrong by having very large negative values. I am working on a verification case for this purpose, but I am not sure if these large negative values are causing the discrepancies which I observe between OpenLB and experimental results.

[mathias]

Lets see, keep us updated!

[Author]

Posts