[radix]

Dear All,

As I understood, Spalart-Allmaras is basically a low-Reynolds turbulence model. However, in FLUENT, there's an option in the model to activate the low-Re damping. It consfuses me. Why would a low-Reynolds turbulence need an option of a low-Re damping? Any reference or book would be very helpful.

Many thanks.

[LuckyTran]

1) The low-Re damping is for extending the turbulence models into the transitional regime, for intermittently turbulent or relaminarizing boundary layers. In other words, for situations where there is less turbulence.

2) Sometimes when ppl say low-Re vs high-Re, they are referring to low-y+ and high-y+, that is the model changes behavior near and far away from walls. Some models are valid only in the high y+, or high-Re regime and some models are valid only in the low y+ or low-Re regime. The confusion exists because y+ looks just like a Reynolds number, actually y+ is the Reynolds number based on the inner coordinate (y non-dimensionalized by the friction velocity).
Note that regardless of whether you have a low-Re or high-Re model (low y+ or high y+), you still cannot apply them for transitional boundary layers because the models only work for fully turbulent flows.

[radix]

Thank you for providing some insights of the damping. I agree with your second statement regarding the low-Re and high-Re. That is why either you make the grid y+ < 5 or y+ > 30 (which is the buffer region or transitional region) to be able to use the turbulence modelling. Otherwise, it will end up in inaccurate approximation.

I have a specific question regarding the damping and y+. SA model requires a very fine grid, i.e. y+ <= 1. However, if I only have y+ = 3 (for some reasons, I am very reluctant in increasing it to y+ = 1 as it will end up in very high aspect ratio and large number of grid cells) and the damping is activated, will it give a close accuracy as y+ <= 1?

Many thanks,
Radit

[LuckyTran]

Just to be clear, the intended use for the low-Re damping is if you were simulating flow over an aircraft wing that is near the critical Reynolds number (500 000 based on chord length). The low-Re damping does not do anything to the buffer region in turbulence model. Phenomonologically they are the similar, but the low-Re option and the model does not improve the accuracy in the buffer region. The user input for the low-Re damping option is an intermittency parameter (which is like a % of time the flow is laminar versus turbulent).

As for how low is low enough. It's more important to get boundary layer resolution (i.e. 10 or 16 or 32 cells across the boundary layer) than to get an exact y+<1. The velocity uses a blending function anyway so that's it is insensitive to y+, and the model itself does not change because it is is a low-Re model.

[radix]

Many thanks for your explanation, Lucky Tran.

Regards,
Radit