[1/153]

Newbie here.

I am doing a 2D rans simulation to investigate some turbulence models. I am wondering how to make the flow turbulent?

For example in a simple Re_tau=180 channel flow, or a little bit more than that. Pressure gradient driven, using some low Re turbulence model, and keep the mesh having y+<1.

Then I set the initial nut=100*nu, k/e>100*delta_t, mass is kept constant on the solver's side (I am using an unsteady solver (URANS)).

But I found the flow is always laminar or laminarized. This is very annoying.

But I read so many papers, they all have a perfect velocity profile, How did they do that? Any secrets?

Or what did I do wrong?

Any ideas?

--------
1/153

[FMDenaro]

Originally Posted by 1/153

Newbie here.

I am doing a 2D rans simulation to investigate some turbulence models. I am wondering how to make the flow turbulent?

For example in a simple Re_tau=180 channel flow, or a little bit more than that. Pressure gradient driven, using some low Re turbulence model, and keep the mesh having y+<1.

Then I set the initial nut=100*nu, k/e>100*delta_t, mass is kept constant on the solver's side (I am using an unsteady solver (URANS)).

But I found the flow is always laminar or laminarized. This is very annoying.

But I read so many papers, they all have a perfect velocity profile, How did they do that? Any secrets?

Or what did I do wrong?

Any ideas?

--------
1/153

Could you post your velocity profile compared to the DNS at Re=180?

[lakeat]

<deleted><deleted><deleted><deleted>

[1/153]

Thanks for helping. Yes the velocity profile is here: [shared just for 24 hours]

http://f.imgtmp.com/Dcdfq.jpg

[1/153]

Is that because 2D cant sustain the turbulence structures, so to damp out all the turbulence? Or is that I should use RANS not URANS?

[FMDenaro]

Originally Posted by 1/153

Is that because 2D cant sustain the turbulence structures, so to damp out all the turbulence? Or is that I should use RANS not URANS?

No ... you are solving a 2D statistically steady solution, you never see in RANS any other variable that zero-order statistics....
But your problem is that you are solving a smaller Re number, your profile arrives at y+ about 10 at the half-heigh!

[1/153]

I will test with other LRN models to see if my new implemented model is wrong somewhere. But have you ever use steady RANS to get a correct profile before? Which turb model did you use? Thanks

[FMDenaro]

RANS is perfect in turbulent channel flow, you have some wrong in your setting/input