Learning Forum

[Ananth Narayan]

I had this question in my mind for the past few days. In case if I simulate the flow over the cylinder and I wish to perform 2D analysis I guess the results (coefficient of lift and drag) won't match with either case (analytical results of cylinder and sphere)

A similar thing happened when I did analysis for lumped analysis for sphere (circle in 2D) and walls of the circle are given temperature boundary conditions with heat transfer coefficient calculated from Biot formula. I found that the area-weighted average of temperature doesn't match with analytical values if I take the characteristic length as the radius. Why in this case characteristic length is radius instead of R/3. I have seen many online resources and many have taken Lc = R. So I finally got the answer with I took the 3D model and solved it (here Lc = R was taken).

I am stuck in a similar condition where I need to analyze the flow over the sphere. What are the procedures to find the coefficient of lift and drag if ANSYS cannot decide between sphere and cylinder? Is it viable to find lift and drag force then convert to lift and drag coefficient? Else do I need to perform 3D analysis.

[Rob]

In 2d it's a cylinder. If you model a semi-circle in 2d axi-symmetric it's a sphere (sort of) but you'll not see any asymmetry in the flow.
Differences in values from analytical solutions vary, but are often mesh or comparison related. Ie for the latter are you comparing the same data? The HTC in Fluent for an external wall is for the non-wetted side of the wall: the solver works out the heat transfer on the wet side.

[Ananth Narayan]

Let me explain what I did with a 2D circle for Lumped Heat Analysis. I created a circle (assuming the cross-section of the sphere will be a circle). This circle I assumed as made of silver. So from the Biot formula (Bi = 0.05- assumed), I found the value of heat transfer coefficient (h)- in this case, many web resources used R as characteristic length (Lc) but according to the conventional formula, it should be Volume/Surface Area. Question 1: Why Lc has been chosen as R instead of R/3.
After meshing and named selections, in fluent I set the outer wall as constant temperature with convective heat transfer coefficient as obtained from Biot formula and solved the system. Question 2: I found that the obtained value from ANSYS doesn't match with either case of cylinder or sphere analytical formula (attached image). But once I did a 3D analysis of the sphere, I got the correct answer- matching with analytical results.
So, Doubt came to my mind: how will ANSYS determine sphere and cylinder in 2D. But you answered it and I am satisfied- I will try and let you know the results.
If I need to simulate for flow analysis (assuming the whole circle- considering as sphere- since at Re = 200 or so, we get vortices so I assume it better to go with full circle), but based on my past experience I found that coefficient of lift and drag doesn't match with lift force and drag force- as they are linked by analytical formula. Question 3: Will I get an answer if I calculate the lift and drag coefficient from the lift and drag forces, because I know it must be sphere- so the surface area constricts the flow. Is this a good alternative?

[Rob]

In the theory what governs the heat generation in the cylinder/sphere? In Fluent you're setting an external HTC and temperature and then Fluent will pull heat from that based on near wall conditions.
The 2d surface area assumes the cylinder is 1m long.
Lift and drag are forces, the coefficients rely on reference values: read the definitions to make sure you understand what you're reporting.
Finally, check the mesh and ensure you have a very well resolved near cylinder mesh (y+ is the cell height, you need to resolve circumferentially too).

[Ananth Narayan]

I saw a few videos, on how to set reference values to calculate Lift and Drag coefficients. But I just came to the conclusion, First Ansys calculates the lift and drag force, and using reference values it calculates coefficient, isn't it? Also, can you tell me the dimensions of the 2D external flow domain for the sphere? Is the domain shown in the image is correct (D is the diameter of a sphere)

[Rob]

You are correct re the force/coefficient calculation. The force is calculated on each facet and then summed for the reports.
Not sure about the domain sizing, it's about right for a 3d sphere but in 2d it means you've got a blockage factor of around 10% which may be too high. What are the boundary conditions?

[Ananth Narayan]

My boundary conditions are simple inlet velocity (of 5 m/s) on the semi-circular and outlet (pressure outlet of 0 gauge) on the rear vertical wall. What model can I use for the laminar condition of Re = 2500 to see kan vorman vortex streets? Is spallart Allamaras model is okay for such case?

[Rob]

Check the pressure to make sure you're not trying to compress flow around the cylinder. Read up on wind tunnel best practice for an explanation on the effects of blockage factor.
Shedding is quite hard to predict. Make sure the mesh is really good (y+, aspect ratio AND resolution) and that the time step is suitable. k-w should pick up shedding, but really you may need one of the more computationally expensive models.

[Ananth Narayan]

Okay, I will look into best practices on the wind tunnels. I don't get the meaning of "computationally expensive models", can you elaborate on it, please.

[Rob]

Have a look at the various turbulence models in Fluent. S-A is one equation, k-e & k-w are two equation, LES etc require a high mesh resolution, small time steps and solve more turbulence terms so cost more cpu & RAM.

[Ananth Narayan]

Oh, I see. I will take a look at these RANS and LES equations. I feel that mesh around the sphere is hard. I tried 3 combinations but I cannot get the mesh that I am satisfied with. Can you recommend any resource for meshing 2D sphere?

Thank you.

[Rob]

Not really. Are you meshing the semi-circle or circle?

[Ananth Narayan]

I am meshing a circle, inside a domain shown in the above diagram.

[Rob]

OK, so in 2d this is a cylinder. You need to resolve the circle and near wall cells. If you're using Workbench Meshing turn on curvature sizing and set an inflation layer from the circular edge.

[Ananth Narayan]

Hello sir, I have a small doubt on periodic boundary conditions. In Fluent, how do I denote the axis of rotation for 2D object (if I model as in image). Is it that fluent will always take Z axis as rotation axis or it does take the axis normal to screen.
Or do I have to model 3D with 1 degree and apply periodic. I prefer modelling as 2D. Can you tell me the important procedure to be followed.
2. Can you tell me the solution also, what does frame motion and solid motion denotes?

[Rob]

In 2d there is no z-axis. In your case however the model doesn't look very 2d-like - think carefully how the inlet & outlet will interact.

[Ananth Narayan]

Yes sir, just now got idea. We can use axisymmetric for this case and choose the axis. I just got the idea now.

But regarding flow over sphere, I am not at all getting correct answer for drag coefficient when I compare with standard graphs. Can you please simulate a flow over cylinder, then find drag force and can you validate it. Infact many of my class students have tried many possibilities but none of them got correct answer. If possible can you record the video and post here. It will be much helpful for us.

Thank you.

[Rob]

Adapt this https://ansyshelp.ansys.com/account/secured?returnurl=/Views/Secured/corp/v212/en/flu_tg/flu_tg_oneram6_wing.html To find drag force you'll need a well resolved mesh, so y+ of one AND good resolution to pick up any shedding.

[Ananth Narayan]

Sir, it is asking to register as customer. I don't have Customer number. I am using only academic version. What shall I do now to view that document you sent me. Can you share it as an screenshot or pdf please.

[Rob]

/forum/discussion/22563/using-help-with-links