[Aragon]

Hello everyone,

I am an Italian engineering student at the very beginning using Fluent CFD, and I have some questions about what I have to simulate. My task is quite particular: I have to study flows over some airships geometries in order to validate experimental tests conducted by the Germans during WWII in a small wind tunnel having an experiment chamber with these dimensions: lenght of 2 m and flow entrance cone of 1.2 m diameter. The models used at that time were 800 mm long. The airspeed was 56.5 m/s and the related Reynolds number 3.1*10^6. As I said before, I know little about Fluent (and about CFD in general, I should say) so, probably mistaking, I modeled the geometries as bidimensional profiles so that my problem, substantially, could be considered as the simulation of flows over symmetric airfoils. And this is obviously a source of errors that I don't know how to handle: I mean, if I simulate this 3D problem with a bidimensional geometry, probably I have to change the angle of attack of the experiment I wish to simulate in order to achieve likeable results. But if I know something about the correction I have to make for wings, I do not know about anything similar for airships. The other problem is that I created a farfield boundary very similar to what is usually considered in the study of flows over airfoils, so very far from the body (i.e. about 10*cord lenght in all directions) since it must have little influence on the flow, but I do not think this is correct; probably the best thing to do is using a boundary that resembles the dimensions of the experiment chamber of the wind tunnel that I have described before, in order to achieve better agreement with the conditions of the tests I want to simulate: the flow over a model 800 mm long in a chamber 2 m long will surely be influenced by this relatively little difference in lenght between the model and the boundary. My question is: are these considerations correct? Should I immediately skip to a tridimentional simulation and developing a 3D boundary which resembles the dimensions of the experiment chamber of the wind tunnel? And, generally speaking, what are the better boundary conditions and viscous models I have to use in order to simulate wind tunnel tests properly? Is it correct to use, at least in first approximation, the inviscid viscous model since the Reynolds number is so high? Sorry for the lenght of the post, for my bad English and even for my ignorance in these problems, and thanks for all the suggestions anyone of you could provide.

Bye!

[Freeman]

Originally Posted by Aragon

Hello everyone,

I am an Italian engineering student at the very beginning using Fluent CFD, and I have some questions about what I have to simulate. My task is quite particular: I have to study flows over some airships geometries in order to validate experimental tests conducted by the Germans during WWII in a small wind tunnel having an experiment chamber with these dimensions: lenght of 2 m and flow entrance cone of 1.2 m diameter. The models used at that time were 800 mm long. The airspeed was 56.5 m/s and the related Reynolds number 3.1*10^6. As I said before, I know little about Fluent (and about CFD in general, I should say) so, probably mistaking, I modeled the geometries as bidimensional profiles so that my problem, substantially, could be considered as the simulation of flows over symmetric airfoils. And this is obviously a source of errors that I don't know how to handle: I mean, if I simulate this 3D problem with a bidimensional geometry, probably I have to change the angle of attack of the experiment I wish to simulate in order to achieve likeable results. But if I know something about the correction I have to make for wings, I do not know about anything similar for airships. The other problem is that I created a farfield boundary very similar to what is usually considered in the study of flows over airfoils, so very far from the body (i.e. about 10*cord lenght in all directions) since it must have little influence on the flow, but I do not think this is correct; probably the best thing to do is using a boundary that resembles the dimensions of the experiment chamber of the wind tunnel that I have described before, in order to achieve better agreement with the conditions of the tests I want to simulate: the flow over a model 800 mm long in a chamber 2 m long will surely be influenced by this relatively little difference in lenght between the model and the boundary. My question is: are these considerations correct? Should I immediately skip to a tridimentional simulation and developing a 3D boundary which resembles the dimensions of the experiment chamber of the wind tunnel? And, generally speaking, what are the better boundary conditions and viscous models I have to use in order to simulate wind tunnel tests properly? Is it correct to use, at least in first approximation, the inviscid viscous model since the Reynolds number is so high? Sorry for the lenght of the post, for my bad English and even for my ignorance in these problems, and thanks for all the suggestions anyone of you could provide.

Bye!

Hi Aragon, and welcome to this overwhelming world of CFD

I'm not shure I understood you well: could you please enumerate again your tasks and goals in a more schematic way?

If you need to correlate experimental wind tunnel results of real geometries, you will of course become the highest fidelity with a 3D simulation.

As an airplane is a streamlined body, perhaps it is enough for you to use the Spalart-Almaras 1-eq turbulence model, but if you have to study different angles of attack perhaps with the probability that stall is present (flow separation at wings), the model which will fit you the best is k-e Realizable or k-omega SST.

The last thing I can suggest you is to really invest time in making a good mesh, specially the boundary layer. If you're going to use wall functions as your near-wall approach, take care about the first row of cells of this boundary layer, as it must fit with the Y+ restriction of being in the interval (30,300); an unstructured hybrid mesh made of tetrahedra and your prisms of the boundary layer would be fair enough for now.

And about your boundary conditions, I would recreate a wind tunnel instead of making a far-field boundary condition: it would be easier for you to mesh, to the solver to iterate and converge, and for your time, as you may save cell size.

Hope this has enlighten you a little. Just post again if I have not answer to your question.

Good luck,

Freeman

[Aragon]

Hello Freeman,

thanks for your reply. I realize that I was quite tediously long in the previous post so I could not be very clear, but you have understood the problem perfectly: I have to make a comparison between lift, drag and moment coefficient calculated through wind tunnel tests and CFD simulations of the flow over real models of airships (not airplanes, as I believe you think, but those Zeppelins used by the Germans during the thirties - think to them essentially as ellipsoids). I will follow your suggestions about creating immediately 3D geometries and meshing them properly, expecially the boundary layer in order to satisfy the Y+ restrictions. The only problem remains this: the wind tunnel in where tests were made had the experiment chamber 2 meters long and the diameter of the inlet section of the flow of 1.2 metres; the models of the airships were all 0.8 metres long; airspeed was 56.5 m/s. So my questions are:

1) If I have to simulate these conditions in the cfd, should I have to create a boundary region which even geometrically resembles the dimensions of this chamber, i.e. a cylinder 2 metres long and with a diameter of 1.2 metres for the base circular section?

2) If it is so, what would be the most likeable boundary conditions for the inlet and outlet bases and lateral surface of this cylinder? (the only data I have are the airspeed in the chamber and the Reynolds number, and I am questioning myself about how to handle the lateral surface of the experiment chamber of a wind tunnel - i.e. if it is a wall or a velocity inlet)

3) The final and greatest doubt: such a small difference between the lenght of the models and the lenght of the boundary region will not interfere with the development of the flow and so with the calculated lift and drag coefficient? (but I realize that maybe this is a question more specifically concerning the behaviour of the flow in wind tunnels rather than cfd, and that if my goal is to simulate those conditions, i probably should not waste time questioning about why they made the tests in such a short chamber)

Thanks a lot for your help!

Aragon

[paka]

I don't really understand how come you want to test whole plane model creating 2D model. Unless you only want to test sections of the plane.

For future reference, try to use paragraphs It is easier to read and get your points.

Last point, the bigger the domain around your plane, the better. If you create small domain around your structural model, additionally with wall boundary conditions on the side, it might produce unreal results.

Following last paragraph, try to create the whole test domain as big as was used during experimental tests, for example, the same as size of the wind-tunnel room.

Best,
Krst

[paka]

Oh, sorry, somehow I missed two following posts.

First of all, Freeman, why do you recommend unstructured mesh? For example for boundary layer I would definitely go with rectangular/polyhedral mesh.

Then:
Ad. 1) Yes try to do as accurate and as similar to the original model as possible.

Ad. 2) You define boundary conditions at inlet as inlet velocity equal to the velocity in the cylinder (that's what I would do). At the end you set up outlet, on the top and bottom you define wall boundary condition.

Ad. 3) I leave that answer to others.

[Aragon]

Hello Paka,

thank you very much for your help. I will follow your suggestions and those of Freeman and I'm sure I will do a good job.

Best regards,

Aragon

[Freeman]

Originally Posted by paka

Oh, sorry, somehow I missed two following posts.

First of all, Freeman, why do you recommend unstructured mesh? For example for boundary layer I would definitely go with rectangular/polyhedral mesh.

Then:
Ad. 1) Yes try to do as accurate and as similar to the original model as possible.

Ad. 2) You define boundary conditions at inlet as inlet velocity equal to the velocity in the cylinder (that's what I would do). At the end you set up outlet, on the top and bottom you define wall boundary condition.

Ad. 3) I leave that answer to others.

Hi,

Yes, I suggest to use unstructured tetra mesh for all the domain not containing the boundary layer, as I said. And in the boundary layer, grow prisms from the surface of the plane, 8-10 rows of cells would be fine.

With this approach then you can switch in Fuent to polyhedra cells if you wish and save at least 60% of cell count of your domain without loosing precision in your solution: if you do a structured mesh or hexa/hexcore mesh you will not have this capability.

For your question #3 Aragon, I would not worry about the way they did the wind tunnel: you must do your CFD domain to contain a fully develop flow on all your boundary walls, specially at the outlet. Do not waste too much time trying to recreate the same tunnel as they used in your simulations, as it may be not suitable to run it up to the convergence, for instance. The only thing you should ideally have the same as in the experimental wind tunnel tests is the flow conditions at the inlet (i.e turbulence intensity, hidraulic diameter, flow velocity, and so on) for your correct set up of the inlet parameters at the corresponding boundary.

Regards,

Freeman

[Aragon]

Hi Freeman,

thank you again for your explanations; as I said before, I'm quite new in this kind of thing, and I'm sure all of these suggestions will be very useful to me.

Regards,

Aragon

[Freeman]

You're wellcome

ciao!