Fully Developed Flow in a Square Duct

Dear friends

I have written a D3Q19 LBM code for simulating fully developed flow in a squara duct. In order to see the interance region, I implement constant velocity at the inlet of the channel. for velocity-inlet microscopic BC, I used those presented in “Laboratory validation of lattice Boltzmann method for modeling pore-scale flow in granular materials” . But my code diverges from inlet and outlet. Does anyone have experience in simulating 3D duct flow?
Your help is greatly appreciated.


what do you mean by “my code diverges from inlet and outlet”? Please give more details. Does the simulation become unstable? How does you velocity field look like?


Hello Timm,

Thank you for your respond.
I mean that after some iterations, my flow field starts giving unexpected values from nodes that are located near inlet. And finally my flow field completely become disturbed.Have you ever done a simulation like this in 3D?


Take a look here. I think it should contain everything you need.


I am trying to write a D3Q19 LBM code for simulating fully developed turbulant flow in a squara duct. My code has Smagorinski sub-grid model, but I’m having hard time to generate fully developed turbulent flow. What initial conditions do I need to apply in oder to generate perturbated flow? My gid size is 4327474 and maximum velocity is 0.085 and Re_tau = 300.

What does it mean that you have a hard time? What exactly does not work? Do you introduce some kind of distortio or fluctuation in order to trigger turbulence?

If I just let it run for very long time without any fluctuation, I still have laminar flow. I also tried random perturbation velocity u = u+u’, v=v+v’,and w = w+w’, but it converged back to the laminar flow near the outlet. Basically, I guess I do not understand how to introduce fluctuation without any objects in oder to generate turbulence flow.

First, without fluctuation, you will not observe turbulent flow. This does not surprise me.
Convergence back to laminar flow could be explained by the fact that the Reynolds number still is quite small. I think that Re = 300 is not sufficient for turbulent duct flow. You should increase it.

Dear Timm

Thank you for your help.