CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   Law of the wall on curved surfaces (http://www.cfd-online.com/Forums/main/268-law-wall-curved-surfaces.html)

 Rasputin October 22, 1998 02:18

Law of the wall on curved surfaces

Does anyone know (or have any pointers) as to what happens to the non-dimensionalised turbulent boundary layer profile on curved surfaces (convex and concave)?

I'm sure it deviates from the standard flat plate prescription. Do any fancy body fitted general purpose CFD commercial codes accommodate the deviation?

 John C. Chien October 26, 1998 14:37

Re: Law of the wall on curved surfaces

The law of the wall issue is always an interesting subject. When used ( with or without pressure gradient correction ) for attached flow, it can save mesh points and computing time. Sometimes, it is more accurate than some low Reynolds number models. My problem is related to the separated flow calculations. It is hard to find a working logic to implement the law of the wall in the reversed flow region ( including the point of separation.). I think , in a way, it is related to how the wall shear stress value is evaluated during the calcualtion. The other issue is how do you evaluate the actual mass flow within this wall layer region ? ( between the first mesh point and the wall point ). Obviously, it will depend on the first cell size used. I think this is very important as the flow is approaching the separation point.

 Rasputin November 2, 1998 04:39

Re: Law of the wall on curved surfaces

Why in the reversed flow region?

I find it interesting that most general purpose CFD codes purport to handle complex curved geometry under a wide range of flow conditions wihtout attempting to deal with the law of the wall problem. Surely they can't expect everyone to integrate straight to the wall. Having said that the main source of error in a CFD calcuation will most likely be innaccurate boundary condition specification by the user, not the models employed by the CFD code.

 John C. Chien November 2, 1998 10:28

Re: Law of the wall on curved surfaces

You are right about the source of error, that is the boundary condition itself is usually developed for the simple flow and simple geometry configuration. At the research level, this is fine because no one is going to make the geometry or configuration difficult for research. At the practical application level, the situation is quite different. In the structured mesh case, it is easier to see the overall picture, that is you can easily bring neighboring points into the formulation. This is not the case for the un-structured mesh case, therefore, almost all the models or concepts are completed within the cell or volume. The wall function concept is derived for the boundary layer region, not just for a cell next to the wall. At the research level, you normally have access to the source code, so you can carry out modifications to improve the solution accuracy of the particular problem you are solving. The situation is not the same for the commercial code case. Unless the code developer has included some guidelines to handle the problem, the users are going to be left in the cold without knowing what to do. There were evidence that poor results were caused by (1). coarse mesh , (2). numerical algorithms ( boundary layer equations vs Navier-Stokes equations ), instead of the model itself. I must say that the code developer is not equal to the problem solver. And if the code can not solve the users problem, the code itself is useless. There is still a big black hole in between the code and the users problem. In 3-D problems, it is expensive to use low Reynolds number model. So I think, it is still a viable approach to use wall function method. In other words, improvements in this area is still needed to cover complex geometry, non-orthogonal mesh, separating or separated flows. Since you can always use a low Reynolds number model results to check against the wall function results, some new wall function models could be developed to handle different problems. In this way, at least you can bring the 3-D results to the real world. ( most of the time, 3-D results are not even qualitatively correct at all. They are just nice picture for PR. )

 All times are GMT -4. The time now is 15:27.