[matthias.frontull]

Hello all,
I am currently working on an assignment for university. I have to perform a parameter study of the drag coefficient versus Reynolds number for a flow around a cylinder.

Basically i have to recreate the following diagram as best as possible
http://www.esru.strath.ac.uk/EandE/W...efficientC.jpg

Problem specs:
diameter = 1
density fluid = 1
VISCOSTY = 10^-5 (seems to me to be the problem)

Reynolds numberrange : [0.1 to 10^6];
I have a circular flow domain with radius 64meter (according to tutorial of cornell university, https://confluence.cornell.edu/displ...ast+a+Cylinder)

MESH
For the mesh i tried a face mesching and edge sizing on the cylinder wall and a global face sizing getting almost 100000 cells.

MODEL and BOUNDARY
I tried to simulate the creeping flow region (Re = 0.1 Re = 1, u_inlet = 0.0001), using the laminar model, and calculate the drag coefficient.

MY PROBLEM:
Instead of values around 40 as in the experiment I get basically a drag coefficient equal to zero! And I dont really undertand why?

the viscosity is very low, could that be the problem, that i get a ill conditioned system of equations or smt similar?

I would be glad for any hind or suggestion, thx in advance!!

[FMDenaro]

how do you compute the stresses around the body?

[piu58]

I worked a similar (or may be even the same) geometry without any problems. But I finished with Re al low as 2.

I made everything to 1 except the viscosity, which I use for controling the Re number: U=1m/s, d=1m. I tried a different setup with the same Re, and that gives virtual the same results.

For very low Re numbers the computed region around the cylinder should be rather large: The stream lines get widened even far away from the cylinder. I set it to 10m = 20 radii at each side.

For Re=2 I got a drag coefficient at around 10.

~

> how do you compute the stresses around the body?

There are two ways:
1) Simply use the libforces.so library which needs to be included in the controlDict
2) More complicated: Use "wallShearStress -latestTime" for calculation of the stress, use the calculated pressure for the resistance part of the drag. The results for the drag coefficient are identical, if you extrapolate the pressure to the patch centers. If you use the element centered value , they differ slightly.

[matthias.frontull]

to FMDenaro: I dont compute the stresses, i directly compute the drag force in Fluent with Report/Forces/ and than selecting the cylinderwall and the x direction! Or did you smt else

to piu58: My flow region has a diameter of 64 meter, so 32 times larger, should be enough?
Thats somehow the same for me, if i try to compute the coefficient by controlling the viscosity i get mor reasonable values for Re = 1, but on the other side if i increase Re to 20 50 or 80 the coefficient rises instead of falling as expected from the diagrams! Which somehow would even make sense, that if all the rest remains equal and i increse the viscosity also the friction and drag rises or am i wrong?

[matthias.frontull]

i just tried to simulate another time controlling the reynolds number by the velocity.

rho = 1, viscosity = 1, velocity = [0.1 1 10 100]

I get the results for cd = [22.66 70.024 156.65 8962.2], so actually completely the opposite behaviour as expected from the experimental curve!

And i mistold, by flow domain diameter is 32meter, not 64meter!

[piu58]

> if i increase Re to 20 50 or 80 the coefficient rises instead of falling

May be you calculate the force coefficient instead of the drag coefficient.

[matthias.frontull]

May be you calculate the force coefficient instead of the drag coefficient.[/QUOTE]

But the drag coefficient is calculated from the dragforce?

[piu58]

> But the drag coefficient is calculated from the dragforce?

Yes, drag coefficient divided by dynamic pressure.

[matthias.frontull]

Still could no find a solution, for small viscosities I don't get any dragforce! Could it be that due to the low viscosity I have to define very very small elements on the cylinder surface? or exactly the opposite? If i computed forces from stresses, and stresses from larger CV-surfaces I should get larger drag value for each CV but just with lower resolution?

[FMDenaro]

what is relevant in the size of the cell is the Re number, not the viscosity value alone

[piu58]

> for small viscosities I don't get any dragforce!

In such cases take a look at the flow in the near of the cylinder cyrface. Is that, what the program calculates, physical plausible? In some cases I did not get s smoth flow but an oscillationg one. Drag forces form such a solution cannot be correct, of course.

[P.Charoen-amornkitt]

If you are using Fluent, Cd can be obtained directly from the software. You should compare those values with yours that whether it is same.


Sent from my iPhone using CFD Online Forum mobile app