[bgp723]

Hello,

For a project that I am working on, I am interested in doing some analysis with the local Nusselt number values over a rectangular duct.

I am simulating a simple single phase duct flow in 3D and my boundary conditions are as follows:
- uniform velocity inlet, 300K
- pressure outlet
- walls: 400K

The problem occurs when the flow is very slow, or if there are some stagnation points in the flow domain, at which the fluid temperature gets very close to the wall temperature.

In some cases, the temperature values in the fluid domain nodes are actually higher than the temperature values of the wall, though not by much 1e-5 (T_wall - T_fluid, both local values), which I assume is due to small computational errors. Of course, physically, fluid cannot be higher than the wall temperature.

These data points will noticeably throw off my local Nusselt number calculations because Nu is proportional to 1/dT. Basically, I expect to see the Nusselt number approach the fully developed value but, instead, I see Nusselt number approach the fully developed asymptotic value and then blow up after a while due to dT error. Heat flux (q'') isn't perfectly in line with the very small dT nodal points either because then the local Nu wouldn't stray off from the asymptotic value.

I was wondering if anyone have any suggestions on going around this issue, and/or methods to improve the fluid temperature accuracy.

Some additional info:

Fluid/Solid properties are temperature independent

I enabled double precision and I run the code until my convergence reaches ~ e-7 for continuity, ~e-11 for energy, and ~e-5 for other parameters of interest that I am monitoring.
I have done some mesh independence studies and I believe my current mesh is fine enough.

I extract all data I need in ASCII format and do my data analysis with the nodal point values.

For higher velocity cases, Nusselt number behaves as expected because fluid reaches the outlet before its temperature getting close to T_wall.

Thank you in advance!

Dan