[fZoltan]

Hi everybody!


I need some help about "sub-modeling" in CFX.



Firstly, I think it's important to explain what is the meaning of sub-modeling in my point of view. Sub-modeling is a "method" or an "approach" used when full model's mesh size is too large.



Let's see an example: In situation when we want to simulate fluid dynamics in case of a fully charged steam sterilization chamber it is impossible to simulate whole model in detail. I think the only way is to make some approximations whit sub-modeling process. In that chamber there is a lot of porous bags, in this bags there are devices intended for sterilization, furthermore the chamber volume is several m^3 while dimensions of porous bag (thickness) is just a few um (micrometer).



The idea is using sub-modeling. I know well this is just an approximation, but I do not have another idea at this moment.


So, my first step was to check if calculations in this situation are correct. I have made a simple model, two fluid domain separated with a thin porous domain. Initial conditions in domains: Pressure = 3000[Pa]; Mass Fraction of Air from a Mixture Gas(Air + Water Vapor) = 1; The Inlet Boundary was one of the Surface of Fluid domain (F1 domain) parallell with Porous Domain (Mass fraction of Air on Inlet = 0 and MassFlowRate of Mixture was a given value ). The other Fluid Domain (F2 domain) was closed (like in case of chambers).
I run the simulation in case of full model saving some important variable (Mass Flow Rate, Mass Fraction ...) calculated on the interface of Porous Domain and F2.



The next step was geometry reduction to one Fluid Domain suppressing other Domains (suppressing F1 and Porous Domain).



And what I do not understand:


If I give saved variables - (Mass Flow Rate throw Interface and Mass Fraction of Vapor ) from full simulation - like initial conditions to F2 sub-model, the Result is not the same (mass concentration of vapor in F2 Domain) in two cases.


If something is not totally clear please ask.



Summarizing the problem: Results of a full model is not the same in that case, when we split the model, extract some results (mass flow, mass fraction) at splitting points (BC Profile at every time step) and giving these results like boundary condition to sub-model.


Thanks a lot,

[ghorrocks]

Have you considered that this approach is probably going to require a few iterations between the large and small scale models before it converges?

For the approach you propose to work the boundary conditions you apply on the small scale model need to accurately represent the large scale model, and the boundary in the large model where it interfaces with the small model also need to be accurate. This is a statement in the obvious, but you really need to check this.

In particular, consider that it is likely that you will need to iterate a few times between both model scales before the small to large interface converges. You then need to consider numerical issues like will my approach converge numerically? Is it stable? How do I achieve stability?