Hi guys, - I think this one's a tough one, so I hope some bright Fluent user knows how to do this or a workaround!

I want to use the UDS scalar capability to solve two equations, where the first is actually the velocity of a second phase and the second is say its temperature.

A simple form of the equations is:

1. d/dy(rho*u) = S (velocity of second phase)

2. d/dy(rho*u*Cp*T) = S(T) (temperature)

now the source terms are easy to handle. My problem is with the convective terms. I have used the UDS_FLUX function, but I am pretty sure that during the discretization process, Fluent chooses the upwind cell based on the direction of the primary (gas) phase flow. Thus, if my phase is not flowing in the same direction as the gas flow, Fluent will choose the wrong upwind cell and the results are unreliable.

What I would like is some method to set the co-efficient of the unknown neighbour cell variable (say u or T) in the solver matrix directly, or to tell Fluent the appropriate upwind cell etc. There might be other options I haven't thought of yet.

if one looks at how the convective terms is discretized and put into the solver matrix, the term essentially determines the co-efficient of the unknown variable values. Have a look at eqn 22.2.2 in the Fluent6 user guide and the discretised form 22.2.11. The eqn 1 above will reduce to something like:

a_p*thi_p = sum_over_neighbour_cells_n a_n*thi_n + b

where a_n are the co-efficients of the unknown variable thi in the neighbouring cells of the cell p. (Its a bit hard to explain in text, see eqn 22.2.11!)

Anyway, I'm hoping somebody knows how to set the a_n s arbitrarily by using some undocumented Fluent function!!!!

I know this long winded, if you've written your own code, you'll probably understand what I mean. If not, its probably, gobble-de-gook.

Thanks

Greg

ps. any ideas will be appreciated... pss. for the p cell, the appropriate co-efficient can be set by modifying the suggested form of the unsteady uds flux function - but I can't see a way to set the neighbouring cell co-efficients, a_ns...

Hi Guys, as a follow up to the above I am wondering if the macros, F_COEFF_01 and F_COEFF_10 might be of some use. As a guess I think they might set the co-efficient of the neighbour cell values (from cell c0 to c1, and c1 to c0) in the amg matrix of the segragated solver.

These are defined in sg_mem.h.

Anybody know this stuff???

Greg