[Gerry Kan]

Howdy Foamers:

I have the following question. In a compressible solver I am modifying the FV formation of YEqn.H to include a source term based on a uniform but transient injection rate for species "i":

fvScalarMatrix YiEqn  (
    fvm::ddt(rho, Yi) == fvOptions(rho, Yi) + rhoSp
);

The term rhoSp is a rate of mass fraction based on the injection rate for cell k, that is

rhoSp[k] = rho[k] * mdot[k] / ( rho[k]*V[k] + mdot[k]*dt )

where

k is the cell ID
rho[k] is the density of cell k
mdot[k] is the injection rate of species i into cell k
V[k] is the volume of cell k
dt is the current time step

The total mass of species i injected into the domain can be recovered by summing the amount of it over all cells (it does in my case), and the mass fraction of species i is expected to be very small ( < 1e-7 ).

As a test case I have the injection take place in a closed chamber, where upon the amount of species i introduced into the domain is known a priori, and verification is made by summing up the amount (not mass fraction) of species i at each time step. In other words, at each time step, the difference between the computed and expected mass of species i at each time step should be very, very small.

What I have noticed, however, is that, as time goes on, this difference creeps up geometrically and at some point early on in the simulation, it could double or triple the expected injection mass. My hunch is that it might have something to do with implicit vs explicit source terms. But unfortunately I don't really know how to proceed. Could someone shed some light on this matter?

Thanks in advance, Gerry.

[Gerry Kan]

It turned out the implementation was correct from the get go. My original boundary conditions were simply non-conserving and therefore it appeared as if I was losing mass. In the end, I had all wall in my test domain (except for the boundary where the source term is implemented) into cyclic and the problem was solved.

Gerry.

[SHUBHAM9595]

Hi Gerry,

While looking for an explicit source addition in the governing equation, I came across your post. I've minor query regarding the implementation....

Is there any specific reason why u opt to divide by cell volume ?

rhoSp[k] = rho[k] * mdot[k] / ( rho[k]*V[k] + mdot[k]*dt )