Learning Forum

[SpencerJordan]

Hi all,

I am trying to simulate a simple 2D situation where the walls of a cylindrical structure are cooled to 240 K and 270 K, respectively. Then, air is injected at the top of the cylinder at a certain velocity and is cooled as it flows down. My analytical calculations predict a much greater affect on the flow profile of the air from the temperature of the walls than is being shown in the simulation results - almost no change in flow profile is seen in the simulation.

Currently, I'm using the Realizable k-epsilon model with enhanced wall treatment and thermal effects. I have the energy equation enabled and I am setting my walls to constant temperatures. Should I be using a different model/parameters to model the expected thermal turbulence? Thanks!

[Surya Deb]

Hello,
Do you have sufficient length to see the development of thermal boundary layer?
Also check your mesh resolution near the walls. What is the Y+ value near the walls?
Do you impose a transient boundary condition where the wall temperature cools with time?
Regards Surya

[SpencerJordan]

Hi Surya What length do you think would be required for this situation? The chamber I'm simulating is about 1m long. I'll attach a photo of my mesh resolution at the walls, but I have refined it several times at those boundaries. The Y+ value does seem low, it is 1.0 or lower along both of the walls. And the boundary condition is not transient, I would like to simulate with the walls being held at constant temperature.

[Surya Deb]

Hello Spencer,
Generally the Prandtl number of air is around 0.69, which indicates that thermal diffusion is faster and more dominant than the momentum diffusion. So thermal boundary layer formation should occur faster.
Are you trying to model a 2D and compare the analytical results for a 3D cylindrical case?
I would recommend you to model a 2D axi-symmetric case instead of 2D planar.
In 2D planar, you will be missing out on the additional terms.
Regards Surya

[SpencerJordan]

Okay, thank you. You are correct, I am trying to compare to a 3D case so I'll give axi-symmetric a try. And the Prandtl Number I'm seeing (in the original planar simulation) is about 0.38

[YasserSelima]

Make an inflating boundary ... try to decrease the mesh size of the first layer close to the boundary and let it increase as you go away from the wall.
One thing I have seen in an earlier post in this forum; someone had a completely weird flow behaviour in natural convection and the reason was selecting large timestep. So, you might take this into consideration.
Regarding Prandtl, it is almost 0.7 for air between 150 K to 1500 k ... so probably something is set wrong in the material properties if you are simulating dry air

[SpencerJordan]

It seems that my issue was that I had forgotten to enable gravity in the model, therefore no thermal buoyancy effects could be calculated. I turned gravity on and made sure air's density was set to a temperature dependent model (incompressible ideal gas) and I am now seeing my predicted results.

[DrAmine]

Thanks for feedback