Hi, I have a few general questions here.

I am using a FE code solving RANS with standard K-E turbulence model with wall function.

Let's say I am solving an internal flow problem (flow in a duct), with an "adequate" mesh density throughout the whole domain.

Accoding to the code help manual, for low Re flows (Re < 10,000), given a fixed mesh, it is advised to only apply one extra layer next to the wall so that y+ is within (35-350). Applying more layers would reduce y+ values below 35 and that subsequently the code would under-predict the pressure loss due to shear. At this point, I assume this rule is holly for this particular code, but I would like to hear some comments from you guys out there using fluent, CFX, STAR-CD etc.

My 1st question is: is one layer really suficient in resolving BL flows (Re < 10,000)?

If the answer to the 1st question is no. then my 2nd question is: based on this logic, should I reduce the overall mesh density in the computational domain so that I can specify more layers and pretty much keep y+ the same.

My 3rd question is: based on my understanding, applying more layers will help better resolving the flow structure in the BL zone, and the penalty for that is an increase in run time. If I can tolerate the run time increase, is it true that during the setup, I should apply as many layers as possible for better accuracy at the penalty of run time?

My 4th question is: what if I have an internal flow with Re between (100,000 - 1,000,000)? What is the approach then?

You should keep one thing in mind here: the way in which the near wall region is resolved depends on the k-e turbulence model that you are using.

a) If you are using the High Reynolds number Standard k-e model, you are using a wall function where y+ is required to be between 30 and 200 (check your manual!). This way you save a lot of elements. Also, make you sure you are aware of the difference between equilibrium and non-equilibrium condition (between turbulence production and dissipation).

b) you could use a low Reynolds number k-e model where you resolve the near wall region down to a y+ value of typically 1 or 2. This approach is more accurate but requires more CPU time.

It seems as though you are using model a) so in this case one layer of nodes will suffice as long as the y+ values of this first element layer are within the required range.

Thanks for the reply. Could you elaborate further on the "difference between equilibrium and non-equilibrium condition (between turbulence production and dissipation)"?

Thanks!