[Talerbla]

Hi everyone,

I am currently working on a incompressible laminar pipe flow simulation combined with a passive scalar transport. Therefore I chose the pimpleFoam solver (max Courant Number of about 20) and the scalarTransport function object. So the scalar transport is solved after the outer-corrector loop of pimpleFoam. Everything works fine so far but I fear I have a little misunderstanding / problem in my setup. The reason why I am suspicious is that in my results I see unrealistic concentration osciallations (of the passive scalar) close to the walls whereas the velocity field looks normal. Ignoring numerics for a second, my understanding is:

I have a underlying velocity field with a certain viscosity (momentum diffusion). When I solve the passive transport (with a diffusion coefficient of 0) I would not expect these oscillations as the advection is basically "limited" by the viscosity of the fluid. I am most likely mistaken here!?

My other thought is that because the velocity close to the wall is very small, the scalar can "stay" there without being advected and then form these weird osciallations, as there is no diffusion that allows to smooth out the concentration gradients.


As soon as I set my diffusion coeffcient to something like 1e-6, these osciallations vanish as the additional artifical diffusion improves the numerical stability.


Best
Tobias

[tomf]

Hi,

Without diffusion you indeed run into stability problems that can cause these oscillations. There are some differences in the schemes to use, but using a Courant number of 20 may not help in this sense.

I guess some explanation, even if it is not exactly the same problem is found here.

I would expect to see oscillations damp out with lower Courant number and indeed by choosing a small value of D. I would also expect some diffusion to be present. The choice of discretisation schemes also influences the result of course.

Cheers,
Tom