[pilou]

Hi,

I'm now facing a huge problem in a simple study.

I'm trying to simulate a very simple molecular diffusion case in a porous medium. The case consists in a rectangular 2D mesh divided into three parts

http://img825.imageshack.us/i/poctd1mesh.jpg/

The middle part is a porous medium and both sides are purely gaseous. The left wall is 500K and right is 2000K.

The initialization is made with a linear temperature gradient close to the theoretical solution and with 0.2 mass fraction of H2 on the right side and the rest is filled with H2O, there is no reaction between the species, this is a pure diffusion problem.

I try to make an unstedy simultion of this device and I noticed that the quantity of H2 in the entire model doesnot remain constant at all, which is a complete nonsense since there is no sources or sink of H2 and the domain is completely closed. The error reaches 55%.

Does someone already encountered this problem and found a solution.

I run an unsteady single precision, density based solver, with the Stiff chemistry solver on. The diffusion is Fickian (I didn't notice any changes in the results with the full multicomponent model) and the porous medium is just considered as a sink in the momentum equation. The problem is completely laminar (no macroscopic velocity at all).

Increasing the CFL seems to reduce the problem but I'd like to know if there is a way to reach a complete conservation of the elements.

Thanks
Pilou
Pierre-Louis Schmitt