Questions about Boundary Layer Thickness and Turbulence Models
 

Hey Guys,

The model I am dealing with has two parts, circular pipes and parallel plates. Now I got a few questions and any help would be appreciated.

1) I am using the two-layer All y+ wall functions. So how to decide the prism layer thickness? The distance d between the parallel plates varies from 2mm to 6mm. Based on my preliminary simulations, I concluded that the flow is laminar in the parallel plates (Re=900) and turbulent for 6mm (Re=6800). The critical Re is 1400 according to an ebook I found online (https://ecourses.ou.edu/cgi-bin/eboo....1&page=theory).

So before I input the prism layer thickness into STARCCM+, I need to estimate the boundary layer thickness. The equation I used is the one from Wiki (http://en.wikipedia.org/wiki/Boundary-layer_thickness). (The issue for the equation form Wiki is that it is for a flat plate but not for the parallel plates. So anybody knows where I can find the correlation to estimate the boundary layer thickness within the two parallel plates? I searched quite a while but no luck.)

The question is a) how to decide the x in the equation showed in the Wiki? I am thinking to calculate the entrance length (L) first and use x=1/2 L, is that appropriate? Is there a tradition? b) If I get the boundary layer thickness (BLT), how to decide the prism layer thickness? It should be thicker than the BLT to capture the boundary layer, correct? c) Since I am using the all Y+ function, do I need to calculate the thickness of the first prism layer cell to make sure it is within the sublayer region or I just need to make sure the whole prism layer thickness is thicker than the boundary layer thickness?

2) I am using multiple region meshing which means I can use different mesh sizes for the pipes and the parallel plates. So do I have to calculate the boundary layer thickness for these two regions separately in order to use the different prism layer thickness or I can just use the same prism layer thickness for both regions?

3) Since both turbulent flow (in the pipes) and laminar flow (in the 2mm parallel plates) appear in the flow domain, is the k-e turbulence model I am using cable of dealing with this situation? How about the LES model?

4) I am running a steady state simulation but what strange is the normalized residual is oscillating and the Tdr is in a level of 0.1 only. The good news is the average value is constant. I monitored the mass flow rate of the outlet for the parallel plates (I have two outlets, one for the pipe and the other for the plates) and the value kept increasing but with a very low increment as a function of the iteration numbers. I run 1000 iterations and the increment was about 0.1% but it is increasing. When I run 3000, the change was about 0.01% but still increasing. I am sure that the flow can reach steady state since there is no complex geometry to create any recirculations but just simply plates and pipes. So is there anyone who can explain this phenomenon?

Thanks in advance!

Famerfamer

I think we need to evaluate our simulation results. use for example a plane section and then look at one property (For example velocity). check some area close to the wall and see if the property get influenced by the meshing or not. if yes, good try is to add the prism layer or to reduce the growth rate or to reduce prism layer stretching rate.

Thanks. So you mean we need to do the "try and error" instead of estimating the thickness of the boundary layer?

the quality of mesh depends all on your case. When you are checking one of the properties ( that is important for your case study), if the result changes by changing number of for example prism layer, then the mesh quality is not good enough. In other words, the result should be independent of the mesh, as much as possible.
However, you need to check also, to see if the evaluated properties are realistic and reasonable.