k-epsilon coupled with heat transfer

[lweb]

Dear elmer users,

I am trying to couple the k-epsilon with the heat transfer model using a simple 2d test case.
In the test case, a fluid flows through a rectangular channel (body 1) and on one longitudinal side of this channel a reactangular heater (body 2, a body force heat source) is placed.

This is my solving strategie:
1. Solve the flow (NS/k-epsilon) for body 1 and save the result in the output file "flow.result".
2. Solve the HT for body 1 and 2 using the restart file "flow.result".

Please find attached the the test case for both steps ofthe solving strategie.

The solving strategie works fine when in step 1 only the NS equation is solved.
Coupling NS with k-epsilon the boundary between the flow channel and the heater seems to smear.
That is, flow occurs in the heater.
Elmer seems to include the heater mesh (body 2) in calculting the flow (body 1).

Using this solution to solve the HT Elmer crashes.

I changed some parameters (boundary layer thickness etc.) but had no success.

Any help or suggestions would be greatly apprciated!

Best regards
Lars

[lweb]

OK, I had the unique idea to read the tutorials and thereafter solved the problem myself.

Solving strategy:
1. Define NS + KE + HT equation for the fluid (body 1)
2. Define HT equation for the solid (hetar, body 2)
3. Solve the two defined equations simultaneously

Compare Tutorial 7: Thermal flow in curved pipe

Please find an example of the case attached.
The solver settings might not be the best - but it works.
The parameter settings (kappa, epsilon, boundary layer thickness etc.) are for test purposes only.
For comparsion, I also attached the LAMINAR solution.

Note: The default "Turbulent Prandtl Numer" is 0.85 (Model --> Material -->Heat Equation).
This leads to an error. Deleting the Turbulent Prandtl Number solves the problem.
I don´t know if this procedere is correct, but it has been suggested earlier:
viewtopic.php?f=3&t=1957&p=5893&hilit=t ... 078f#p5893

Best regards
Lars

[cemg]

Hi Lars,

just some comments on the turbulent problem:

- The K and epsilon values must be specified only in the inflow boundary and in the initial condition. The wall law (and its laminar counterpart) sets internally proper values for their boundary conditions on walls.

- If the simulation is steady, you don't need to include the time integration related keywords (Timestepping Method, BDF Order).

- I'm not sure (Elmer developers could tell you better), but to enable Neumann boundary conditions (Heat Flux, i.e.) you must set the logical flag to True (Heat Flux BC = Logical True) in the affected boundaries.

- Take into account that Elmer doesn't have a thermal wall law implementation, so, in your simulation, the heat flux trough the boundary layer it's being computed as if the layer was laminar, which is wrong.

Regards,

Cesar

[julien givernaud]

Hello,

How can be coupled turbulent flows and heat equation on boudaries in Elmer according to the Cesar post ?

Is there some features permitting to take into account the turbulent heat flux close to the boundaries?

Best regards

Julien

[julien givernaud]

Hello,

Nobody for answer to this question?

Regards

Julien