[assafwei]

Hi,

I am trying to model a cyclone seperator for fluids, I will try to explain the appilcation and the issue I am having (sorry if it is too long):

I have a siphon in a fuel line in which some fuel is accumulating, there is a constant air flow trough this fuel line, and once there is enough fuel in the bottom of the siphon, the fuel gets pushed up the line and enters the cyclone.
The role of the cyclone is to divert the fuel to one outlet and the air to the other (as much as possible). I was able to model this (I assumed an initial "stationary" accumulation of the fuel in the bottom of the siphon and then introduced the air flow - this is the scenario at t=0) and I got pretty good results (compared with the real cyclone in the lab). The biggest issue currently is not with the results - but whether it is worthwhile doing the simulation in the first place - the simulation is currently running using 1 HPC pack (this is what I have at my disposal at the time being) and takes about 12 hours to complete, and I have a competition - a 3D printer... The 3D printer can print about 8 design iterations in one night, and they are tested the next day in the lab (about 10 minutes setup time and 2-3 minutes for each design) so the design engineers are more comfortable with the 3D printing solution, I am trying to explain that CFD will give them more insight but they don't really care... So I was thinking what can I do to expedite the solution time. Since I am only interested in the general behavior of the fuel / air interaction, and the interface between the two is (currently) not of an interest - what is the most coarse mesh with which I can get useable results?

I know this is a really general question but any tips or pointers will be much appreciated.

Thanks.