LES Near-Wall Velocity Profile
 

Dear all,

Has anyone previously investigated the mean velocity profile in the near wall region of a flow using the Fluent LES turbulence model?

I have investigated various flows in square cross-section ducts at Reynolds numbers ranging from around 16,000 to 83,000, in each case with meshes meeting best practice guidelines (y+ order 1 etc). In each case I have found that the laminar sublayer law of the wall (u+ = y+) is not obeyed in the region of y+ from 1 to 10 apart from at the first node from the wall - as a result although the skin friction turns out to be reasonable accurate, the velocity profile is incorrect.

Has anyone observed the same phenomenon or better still actually got correct near wall velocity profile results from Fluent LES code?

Many thanks in advance,

Alex May

Re: LES Near-Wall Velocity Profile
 

I'm actually starting my investigation of the Fluent LES module.

Even if i'm experiencing a lot of strange issues related to the solver (accuracy and convergence in the fractional step), the mean velocity profile seems to be good, at least qualitatively. That is maximum velocity around 18 m/s, viscous sublayer up to y+ = 18 and log law up y+ = 180.

My case is the turbulent channel flow with Ret = 180 with bi-periodic boundary conditions and nita fractional step solver.

I think you should first try to refine the near wall grid up to 4-5 cells inside the y+ = 1 layer because the LES module has no wall model so you have to fully resolve the viscous sublayer (i actually used only about 2 and half grid cells inside the layer). The time step should be refined accordingly (i'm using dt = 0.005).

It may also depends on boundary conditions; in my case there is full periodicity in the streamwise direction so after a while the flow will become fully turbulent everywhre in the cannel. If this is not your case, maybe there are problems with the position where you are monitoring the mean value.

Hope this helps

Re: LES Near-Wall Velocity Profile
 

Thanks for the response!

I will try increasing the number of cells inside the y+=1 layer and see what happens.

Interestingly the 'log law' seems fairly accurately reproduced currently, it is just the laminar sublayer that isn't...

Thanks,

Alex