Specifying Reynolds Number in 2D simlulation

[Daniyal Ahmed Khan]

Hello,

I am simulating 2D axial compressor blade in nek5000, using 4 non-periodic and 4 periodic boundaries, respectively (see attachment A).
I am trying to simulate Re=1385. Hence, I did it in 2 separate ways: Case A and Case B, as mentioned below.

Case A: (Reynolds Number = 1385)
Density = 1, ux=cos42 & uv = sin42 (V=1), characteristic length is set in mesh as 1, viscosity= -1385. As a result, I am getting the separation of boundary layer and vortices at the end (see attachment B).

Case B: (Reynolds Number = 1385)

Density = 1, ux=cos42 & uv = sin42 (V=1), characteristic length is set in mesh as 1, Re=1/1385. As a result, I am getting the separation of boundary layer but NO vortices at the end as I got in Case A(see attachment C).

My question are:

Firstly, since Reynolds Number can be specified in both the ways; Case A and Case B and the value of Re is also same .i.e 1385. So why I am getting different results in both the cases????

Secondly, when I am trying to simulate Re=-138500 with all other values kept as Case A, my solution diverges. I have tried refining my grid but I still get divergence.

Any help in resolving the above-mentioned problems will be highly appreciated.

Regards.

 

[stek...@gmail.com]

Hi, I'm not sure if this is the problem here, but have you tried setting viscosity to 1.0/1385.0 instead of 1/1385 to make sure nek doesn't use integer arithmetic?

[stek...@gmail.com]

Some more remarks: I'm not sure if you even can specify an unevaluated expression in the par file, so I would rather write 0.00072202 instead of 1/1385 or alternatively set it in the usr file for your second test case.

As for your case with Re=138500: raising Re by a factor of 100 is a giant step, so my guess is that you either haven't refined your mesh enough or maybe your startup phase isn't smooth enough (depending on initial- and boundary conditions)

hope that helps!