Learning Forum

[collinsw]

Hello!

I am currently working on a steady-state natural convection heat transfer simulation. I have been using Boussinesq approximation for the fluid, which seems to be applicable, considering the temperature difference between the fluid domain and the heated surface is relatively small and also given that the condition referred in the ANSYS Fluent guide is satisfied.

The problem is, despite the temperature profile looks as expected, the velocity profile does not fade away to the far-side ambient velocity which is actually zero since the fluid is quiescent in the begining. Interestingly, other portions of the domain seems quite well and all of the boundry and operating conditions look reasonable. However, the velocity profile, obviously, does not seem okay. Anybody have any ideas?

The images related to domain, temperature profile and velocity profile is provided below.

Thanks for your interest.

[Amine Ben Hadj Ali]

What are you expecting and why? You have buoyancy effects due to heating and density temperature differences. This will drive the flow. n

[collinsw]

What are you expecting and why? You have buoyancy effects due to heating and density temperature differences. This will drive the flow./forum/discussion/comment/100651#Comment_100651

I expect the fluid velocity to decrease gradually. The flow is driven by the buoyancy force which emerges due to the density gradient and the density gradient arises from the temperature gradient across the fluid domain. Fluid temperature decreases as it moves away from the heat source thanks to the heat being transferred to the colder ambient fluid and hence, the velocity must decrese as well.n

[Rob]

You may need a larger domain to see that. How well resolved is the near wall mesh? n

[Amine Ben Hadj Ali]

You might increase your domain to see that re-laminarisation. Can you share plot of your mesh in the region near the wall and far from it? n

[Amine Ben Hadj Ali]

Also plot velocity vectors on that plane!n

[collinsw]

You may need a larger domain to see that. How well resolved is the near wall mesh?/forum/discussion/comment/100668#Comment_100668

I have already tried to use a larger domain (the one I posted here is 2500x1500 mm and I had tried the same analysis with 4500x1500 mm and 21500x1500 mm domains, didn't work). nThis particular domain has a resolution of 300x300. I tried to obtain firmly rectangular cells so that I could have a good orthagonal quality. Thus, I used face meshing with bias to increase the resolution near the wall while having a properly structured mesh. I chose the bias factor to be 10 which, I think, worked well (no errors or warnings when I checked the mesh in Fluent).nImage 4: MeshnnImage 5: Mesh (next to the wall)n

[Amine Ben Hadj Ali]

What about velocity vectors? n

[collinsw]

What about velocity vectors?/forum/discussion/comment/100688#Comment_100688

Sorry for the wait. The vector display is a little bit odd, to be honest. Here they are:nImage 6: Velocity VectorsnnImage 7: Veclocity Vectors (the upper part of the domain only)nnImage 8: Velocity Vectors (zoomed in)n