CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Phoenics (
-   -   Which RANS model? (

HollyG July 24, 2013 00:25

Which RANS model?
Hi, I am new to Phoenics and am relatively new to CFD in urban environmental applications.

We been using the Chen-Kim modified k-e model for the modelling of wind around buildings, but it seems that the KL & MMK option is more suitable for this type of application?

I used to use the SST model in bridge aerodynamics, which is different in the nature of the problem. On top we have rather limited computational resource now, so SST is no longer an option.

I would appreciate views on if the Chen-Kim k-e model or the MMK k-e model (or even RNG k-e) is more suitable for urban environmental applications?

Many thanks in anticipation.

phoenics_cfd August 8, 2013 10:40

Kato Launder & MMK k-e models
It is known that the standard form of the k-e model can produce unrealistically high turbulence production in the stagnation region on the front of an obstacle or building, and the turbulence energy is then convected around and over the top of the body; and this in turn can compromise the prediction of the surface pressure distribution and the prediction of separation zones. The correction of this deficiency is the motivation for the Kato-Launder & Murakami (MMK) variants of the k-e model. However, before using such a model variant it is important to establish that the solution is not being compromised by numerical diffusion due to the use of too coarse a mesh near the body surface, and especially in regions of rapidly-varying flow properties.

The RNG and Chen-Kim k-e model variants are superior to the standard k-e model in respect of predicting the size of recirculation zones and also separation from surfaces and off sharp edges. The Kato-Launder and MMK variants were very popular in the Asian CFD community before LES became more practical, and then fashionable.

Launder's proposal was to reduce spurious turbulence production associated with irrotational straining. For the classical case of flow around a surface-mounted cube, I found that in the first instance it is most important to use adequate near-wall meshing around the cube. Once sufficient near-wall meshing is used, with ( y+ around 25 or 30 ) or without wall functions ( y+ the order of 1 ), then the Kato-Launder model produces better results than the k-e standard model, ie longer separation regions and no reattachment on the top of the cube.

The modification becomes more important the higher the turbulence in the flow approaching the body. I do not recall the results I got with the Chen-Kim and RNG variants of the k-e model. In fact, LES gave the best result, probably because it can account for the side-wall vortex shedding, but this makes the simulation unsteady and much more expensive, and often impractical for industrial CFD if one is considering flow through a cityscape or a complex urban environment. As I recall LES took about 20 to 30 times the computer time of the standard k-e model with wall functions.

I suspect that the Chen-Kim and RNG models would lead to improved results over the standard k-e model even though irrotational strain is subject to the same Buossinesq approximation in evaluation of the turbulence production rate. This is simply because in regions where the turbulence is removed from equilibrium, these model variants should produce less turbulence production by a different mechanism.

I do not know whether Kato Launder would be superior to these models for cityscapes and complex urban environments, but one could use PHOENICS to compare the various models for flow over the well-documented surface-mounted cube. Rodi's group did this sort of work in the 1990s.

Incidentally, one can always use the IN-FORM facility of PHOENICS to code any turbulence model one requires, including Menter's SST model, by means of formulae entered by command lines in the Q1 input file ( see the INFORM entry in the PHOENICS Encyclopaedia ).

All times are GMT -4. The time now is 00:44.