CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   turbulence kinetic energy at the near-wall node (https://www.cfd-online.com/Forums/fluent/129664-turbulence-kinetic-energy-near-wall-node.html)

itsme_kit February 10, 2014 10:05
turbulence kinetic energy at the near-wall node
 
I applied standard K-epsilon model with standard wall function

I'm modelling air flow around 3d building

I only apply a single layer in the ground and building surface

Is it one thin layer enough? I only applied one because I didn't use low reynolds K-epsilon model

I set the inflation layer in 0.3m

Now I wanna find out the turbulence kinetic energy at the near-wall node

One thing I'm confused is that if the near-wall node is 0.3m above the ground

Thanks

kad February 10, 2014 15:57
You should use more than just one layer. I never do less than 3 Layers, usually more.

In Fluent k-epsilon implies a zero gradient condition normal to the wall for k. Therefore the values of k on the walls are equal to the first node values, thus k is physically zero on the wall due to no-slip condition. In general you can view the first node values of all quantities by deselecting node values under contours.

itsme_kit February 10, 2014 16:09
Quote:
Originally Posted by kad (Post 474337)
You should use more than just one layer. I never do less than 3 Layers, usually more.

In Fluent k-epsilon implies a zero gradient condition normal to the wall for k. Therefore the values of k on the walls are equal to the first node values, thus k is physically zero on the wall due to no-slip condition. In general you can view the first node values of all quantities by deselecting node values under contours.
Do you mean k in the first node is zero

How can it be

We need k for value of y plus

kad February 10, 2014 16:35
Nope, it is a zero gradient condition. This does not mean that k is zero but that it does not change in the surface normal direction. Check out http://aerojet.engr.ucdavis.edu/flue...ug/node512.htm .

itsme_kit February 10, 2014 16:57
Quote:
Originally Posted by kad (Post 474343)
Nope, it is a zero gradient condition. This does not mean that k is zero but that it does not change in the surface normal direction. Check out http://aerojet.engr.ucdavis.edu/flue...ug/node512.htm .
Hi Kad

Thanks for your reply

In the viscous sub-layer, u plus = y plus

I wonder why the error is significant if y plus is located in this region


Why we normally take y plus in the region between 30 and 300 (for example)

Is effect wall roughness excluded between 30 and 300

kad February 10, 2014 18:12
Remember that k-epsilon model is only valid for fully turbulent flow. Therefore the first node has to be far enough from the wall where flow is fully turbulent. Then wall functions are needed to bridge the cell values with the wall values. If the first node is to close too the wall, you are using the turbulence model in a non valid region.

Wall roughness is generally not excluded but its influence depends strongly on the flow itself, e.g. it might be covered by the laminar sublayer.

itsme_kit February 10, 2014 18:25
Quote:
Originally Posted by kad (Post 474357)
Remember that k-epsilon model is only valid for fully turbulent flow. Therefore the first node has to be far enough from the wall where flow is fully turbulent. Then wall functions are needed to bridge the cell values with the wall values. If the first node is to close too the wall, you are using the turbulence model in a non valid region.

Wall roughness is generally not excluded but its influence depends strongly on the flow itself, e.g. it might be covered by the laminar sublayer.
Hi Kad

Thanks very much

I'm much more clear now, and still get some doubts to be clarified

If k is zero gradient and equal to the value of first node

Does it mean the velocity magnitude is non-zero at wall

If it is, how can we explain no-slip condition at wall

Does no-slip condition only confine tangential velocity is zero, not the normal component of velocity

In case we use low reynolads K-epsilon models, is laminar-sublayer solved too cause low value of y plus required for this model.

Does it mean the numerical results from low reynolads k-epsilon model is more accurate than standard k-epsilon model with wall function

kad February 10, 2014 21:15
Quote:
Originally Posted by itsme_kit (Post 474359)
Hi Kad

Thanks very much

I'm much more clear now, and still get some doubts to be clarified

If k is zero gradient and equal to the value of first node

Does it mean the velocity magnitude is non-zero at wall

If it is, how can we explain no-slip condition at wall

Does no-slip condition only confine tangential velocity is zero, not the normal component of velocity

Normal and tangential velocity is zero on walls. Like it is stated in the link above a modified k equation is solved in the near wall cells. Transport of k can still occur over the internal faces of a boundary cell so it isn't usually zero.

In case we use low reynolads K-epsilon models, is laminar-sublayer solved too cause low value of y plus required for this model.

Yes

Does it mean the numerical results from low reynolads k-epsilon model is more accurate than standard k-epsilon model with wall function

In general low Re models should give better results, but mesh requirements (especially wall resolution) are much more strict.
My general advice would be to remesh your geometry with a good resolution of the boundary layer and at least 5+ layers. Then use k-omega-sst as it overcomes some k-epsilon (and standard k-omega) weaknesses and can be used for y+< 30.


All times are GMT -4. The time now is 04:45.