Is there any one know how the PHOENICS simulate the micro fluid ? There are several conditions should be considered. One is the modified boundary condition and the modified governning equation. The other one is the caculation precision. I try to run 3-D model in micro meter scale, and I find that the mass is disappeared in the domain and is not conservative. I guess that it should be caused by the round-off error to eliminate the cell volume. Is there any good idea to resolve this problem. (I know the dimensionless G.E. is one of the solution, and is there any other way?)

You could work in a different system of units, e.g. "microns - kilograms - seconds". Just specify distances in terms of microns instead of metres, and make sure you change all the values of the fluid properties accordingly (e.g. density, viscosity etc). Provided you are careful, it should work.

David Glynn

Flowsolve

Dear Richard

Yes, I agree with David, scale the problem. It should be a very carefull operation but it is not impossible.

Regards

Kike

Thanks for all your good advice.

Could we chang all the real variable to be double precision in PHOENICS through main to subroutine program and also change the TINY variable to be smaller than current default setting ? If we can re-compile it, is this way available ?

Dear Richard

I think it is not possible. I mean, you can change the precision for user declared variables, moreover you can chage some "real" declarations inside include files (d_includ); but I think PHOENICS' F array is declared as REAL somewhere inside any compiled library, so you has no chance to solve DP variables.

Maybe you can ask to CHAM experts or maybe some of us (at this forum) have a different answer. I have made a "grep" through PHOENICS 3.2 instalation directories and I couldn't find the type declaration of F into a source file.

Regards

Kike

Richard,

Is there away scale the problem size up, using non dimensional numbers. I know that I have used this technique to scale down problems, then scale up the results. Just wondering if it could go the other way ?

Leon

I can't see how you could do this unless you had access to the complete source coding. Even if you had, and converted all the variables, I would not be at all surprised if there were other things to do as well (which we do not know about) before a double-precision Phoenics could be got to work.

I really do think it is much easier simply to work in a different system of units. The only place where units enter into Phoenics is in the property values. Apart from this you can interpret the Phoenics variables as being in any system of dimensions that you like. So, just choose a suitable one (e.g. using microns, see my previous comment) and redefine all the relevant property values. It's easy, you just have to be careful!

David Glynn

Flowsolve