Hi !

I have a general question regarding CFD-solvers.

If I do an inviscid calculation on a geometry with a known ideal solution, are there any reasons why the numerical solution would differ from the ideal solution based on stream function theory?

My case is a 2D NACA 0015 hydrofoil, and I'm doing some preliminary calculations to get to know the solver (Fluent 5). The numerically calculated velocity ratio (v/V) tends to be 10% smaller than the ideal solutions based on equations from "theory of wing sections".

Here v is local velocity at the surface, and V is free stream velocity.

I have also done the same inviscid calculation to a 2D cylindrical geometry, and the same tendency appears here. (while maximum value of (v/V) should be 2, I get as low as 1.8)

Gotfred Berntsen

(1). I was asking question about whether CFD is science or art. (2). In old days, to be qualified as an engineering answer, the error must be less than 5%, because people were using slide rules. (3). I can only say that, try other codes to find out whether they all have the same characteristics. (4). If you start fooling around with the code and answer, then I guess the answer will be on the art side. (5). Last Friday, an engineer came to me and show me that he was getting the same wrong answer from the code. Earlier, I sent him a few pages of notes showing the wrong answer I got from the 3-D commercial code. (It is not too bad for you, it is only 10% off. ) (6). I would strongly suggest that you run a couple of more test cases to check out the code. There is not much one can do with a black box code. In industries, normally, there are many other codes available to double check the results. In most cases, test data are also available.

Your inviscid solution should tend to the analytical ideal solution as you increase the size of your grid.

It is hard to say whether you have wrong input files, wrong settings or wrong theory, but the statement above is a basic concept in CFD. Thin wing theory does not apply to an airfoil that is 15 % thick.

I suggest that you do this:

Take some analytical airfoils (ex. Van de Vooren), compute the analytical solution (your benchmark), change your Fluent input file, run the code with increasing grid size and compare (make sure you have no viscosity).