Hallo! Friends,

I would like to know which coordinate system is used in Fluent in defining the governing equations! Is it Cartesian,cylindrical or spherical, or all of them? If there are different coordinate systems available, how to choose one of them?

This became important in one of my calculations, where I wanted to use a 2D reduced problem definition for flow in a 3D cylinder. I reduced the 3D geometry into a rectangular surface. But a rectangular surface is a 2D reduced form, of both a cylinder and a right parallelepiped. Only if I can implement the governing equations defined in cylindrical coordinate system, for the cylinder case, can I distinguish between them. Or does there exist some other way?

Hope the problem is clear enough!

Thanks in advance, for any useful suggestion!

varghese

dear Varghese

I hope you have an axisymmetric problem. If not, your problem is intrinsic 3D and you'll have to model it as such.

If your problem is axisymmetric, model only half your 2d-section, make sure the axis has a 'axis'-boundary condition and enable 'axisymmetric' in the solver-settings. Fluent will use cilindrical coordinates.

hope this helps,

Laika, still orbiting

Dear orbiting Laika,

Thank you for the response.

Let me clarify some points about my problem. It is not axisymmetric. But a natural convection problem, which cannot really be axisymmetric. However the geometry has a cylindrical symmetry, which I wanted to exploit, to simplify the problem setting(due to the memory constraints of my computer). Is it not allowed by Fluent to have such 2D reduction of the problem?

I am aware that some other code(which is an in-house code in our Institute) which uses cylindrical coordinate system to define the governing transport equations, for 2D geometries. My question is whether such allowance is there in Fluent, to change coordinate system?

Expecting more reflections and suggestions on the problem,

Varghese

Dear Varghese,

if you make a 2D approximation of a 3D flow, you assume symmetry. If you simulate the flow around an airofoil, you assume the flow is really 2D (independent of the coodinate along the aerofoil length), and you can't say anything about the influence of wing-tips.

If you use a 2D setup with cilindrical coordinates, the solution you obtain is the solution for an axisymmetric problem. You can't obtain intrinsically 3D data with 2D simulations.

See it this way: your cilinder in 2D is a rectangular section. You mesh it. Each cell contains the data of that location. Suppose your problem is not axisymmetric, so your variables are a function of r, theta and z. Take a cell at location r1, z1 and theta=0 (suppose theta =0 is your 2d section) You cannot expect the software to store in that cell the information of all the locations (r1, z1) and taking into account for that. If it did contain all these info, it would require as much memory as does the complete 3D-case.

You cannot calculate with information that isn't stored. Your in-house code can't do it either.

sorry

greetings,

Laika, still orbiting

I think the answer to this question is that, yes in Fluent you can have a 2D rectangular surface represent either an axisymmetric portion of a 3D cylinder or a simple rectangular domain.

What you should do is this - right down the equations you want to solve and compare them to those solved by Fluent - see manual. If different, evaluate whether there's a way to modify those in Fluent using udfs to represent those you want.

I recently solved a natural convection problem in 2D which was a slice of a 3D domain with mass addition. There was a gradient in the z-dimension, but because this was small compared to those in the other directions, such a 2D approximation was justified. However, because the problem was dependent upon the small z-gradient, I needed to add a source term to the x and y momentum equations and continuity to ensure a proper mass balance. In this case, I assumed that the mass addition was distributed uniformly over the 2D cross-section. A 3D calculation of the same problem shows this is a reasonable assumption, but still an approximation. Perhaps you can do something similar.

I often find older papers in CFD very good sources of these tricks - since people had to work harder in the early days because of severe computer restrictions. These days, people get Fluent, use gambit to mesh up something and press solve....without often thinking too much about the problem!

Greg

Dear Laika,

Thank you for the interest you have in the problem. I thought, however, that I must clarify something more about the problem.

I agree with you, dear Laika, that one cannot obtain intrinsically 3D data with 2D simulation. And, in fact, my interest is not to generate a 3D data, rather to generate a 2D data. Exploiting the inherent symmetry of the cylindrical volume, I thought, would keep me within the memory limits of my computer, if I could treat the problem as a 2D problem. I think, I am justified in doing that, since my interest is to calculate an averaged quantity on the top surface ( the area weighted average of surface Nusselt number on the top surface).

But there arises the problem I described in the earlier postings, to distinguish between a 2D reduction of a right cylindrical and a right rectangular geometry.

As you say, a 2D set up with cylindrical coordinates will give me the solution for an axisymetric problem. Conversely, it means that I can solve the problem as an axisymmetric one. This is an interesting suggestion, which I will test for myself. But I am not sure, whether that will help me to remain within the memory limit of my computer!

I am still looking for a solution.

Keep orbiting,

Varghese.

Dear Greg,

Thank you for your understanding of the problem and for the keys you suggested towards a solution.

As per your suggestion, I read Fluent manuel and revisited the fundamentals once again. I find that the transport equations in Fluent are integral equations defined on a cell basis. But the manual is not clear as to which coordinate system it uses to write the integral transport equaions. Can you please shed some light to this problem.

To my mind, there is no other way to distinguish a 2D surface, that represents a cross-section of a 3D cylinder and hat represent a simple rectangular domain, other than by changing the co-ordinate system used to write the governing equations. (Laika's suggestion to set the problem in an axisymmetric manner seems to be another way, but that does not seem to solve the memory-limit-problem I encounter!, or not?)

Will a UDF help me to change the coordinate system of basic governing equations? Please speak more on this aspect of the problem and please give me some more clues to write one, if it is possible.

I am grateful for any further suggestion.

Varghese

Fluent will use a planar co-oridinate system by default.

You can activate the 2D axisymmetric model by going to the solver (define/solver) panel.

Thus you can choose how the co-ordinate system is defined.

The difference between them is due to the swept area of the face to form a volume if one wanted a 3D domain. In 2D this is a simple extrusion and in axisymmetric problems this is a revolution about an axis...

So you can do both and I can't see any real problem - with Fluent - its just a matter of working out which is most suitable for your problem.

I don't think you'll need udfs...

If you're still unsure, why not try both for a very coarse grid and see what you get...that might help you understand a bit more about what Fluent does in each model...

Greg

Dear Greg,

It works! Thank you very much for the idea.

Varghese

Dear Laika,

Thank you for making the discussion an interesting one.

I did the problem as an axi-symmetric one and it looks pretty good.

By the way, is there a way to get a 3D impression of the results from the 2D axisymmetric data?

Keep orbitting,

Varghese