[seedl]

hi every one!
I am trying to simulate the aerodynamics of the body of an insect fly.
but the problem is that the fly's speed and size is very low. and it will have a very little reynolds number.
in other hands, the Angle of attack of the body is so high!
.
and the other problem is the Flow separation which will occurs.
if it's possible. please tell me which turbulent model is the best for this case?
.
and can i use the default Coefficients of the model in software?
.
the simulated body will experiments in two methods. one that all of the wools of the insect are cleaned and other one that insect's body have more than 10000 of hairs.
thank you!

[LuckyTran]

This is somewhat my personal opinion but insect flight, especially the flapping wing kind, is generally powered by large coherent structures. Meaning this type of flow is best characterized as unsteady laminar flow (which is evidenced by the low Reynolds number). The turbulence is externally generated and comes from the environment. So this belongs to the class of turbulized laminar flow and is not turbulent in the classical sense. In either case, I think this type of flow is suitable for LES and DNS rather than a turbulence model which is tuned for other flows (wall bounded flows, free shear layers, etc).

The problem are the hairs. The hairs I think should be modeled as a pliant structure (which would involve fluid-structure-interaction). The hairs are also tiny and at that scale the surface tension forces become significant and need to be taken into account.

Basically, I think turbulence models are the lesser of your worries. If you disagree and have some references, I would be interested in reading about biological CFD since it's not a topic I specialize in.

[seedl]

dear friend "LuckyTran"!
i read your replay, thanks so much. i thought about it all over of last 24 hours.
and i agree your idea that for low reynolds numbers dns method is very good.
.
but i have limitation in my hardware, i use just one amd athlon(tm) II x2 260 processor in 3.2 ghz clock frequency to proove my idea and getting more hardware like 500 numbers of it!
.
i think to simulating a fly just with rigid wools, i need more than 1000 of this cpu and ram at the same capacity!
i don't know about les model. and hardware requirement of it.
.
what's your guidance about it?
thanks so much!
.
.
one other thing, you talked about references to study!
i don't have access to good references in english language, but if you like, i can give you some cad files to simulating and research about it!

[LuckyTran]

I think your estimates are probably pessimistic. The grid resolution needed is on the order of the size of the fly.

Large Eddy Simulation is in-between URANS and DNS. It is much much cheaper than DNS and a bit more expensive than URANS. The grid size for a well-done LES is a fairly fine RANS grid, but not nearly as fine as a DNS grid. Most of the cost savings from running LES is from using a coarser grid compared to DNS.