[Tobi]

Hi all,

attached is a conjugated heat transfer problem (simple one in 2D). A simple air channel heats up a plate. The analysis I was performing was the energy conservation of the interface. Actually, I expect that the heat flux from the fluid to the solid region is identical to the heat flux which is moving through the solid body itself (simple conduction). However, if you run the case, you will get the heat flux of the following walls:

- fluid_to_platte
- platte_to_fluid

As both patches are the same (even topological identical), I would expect to have the same heat flux values on both sides. However, the result tells an unequal energy balance that I cannot understand. On the other hand, if I check out the heat flux from:

- platte_to_fluid
- ambient

it is almost identical. The ambient wall is a boundary condition that models a heat transfer to the ambient air as:




Any suggestions why the energy balance between the fluid and solid region is not equal?

[Tobi]

Hi all,

I solved the discrepancy of energy conservation. The issue is related to the difference in faces that are connected. The fluid region does have much more faces concerning the solid part. If one changes the face number of the solid to be equal to the fluid one, the energy transport is identical (except for some rounding errors).

[julieng]

Hello Tobi,

You mean that a non conformal mesh was the cause of the wrong energy balance ?
What is your y+ value?

Best regards

[julieng]

Hello Tobi,


I check your initial mesh and I see that the first issue was related to non conformal mesh. Can you update your case with the new solid mesh please?



y+ is < than 1 I think, I don't know the air density (hot air, I take the hypothesis 0.5 kg/m3). I measure 1,1 mm for the first cell thickness. The Reynolds number (Re= velocity x rho x D / visco = 0.5 x 0.5 x 0.05 / 3e-5 < 500) seems to be low for using kEpsilon model, still laminar no?



I was thinking that kEpsilon turbulence model doesn't need low y+ values.


What is your experience with this?


Best regards


Julien

[Tobi]

Hi Julien,



So the velocity is 20 m/s, the kinetmatic viscosity 5e-5 m²/s and the hydraulic diameter is 1.2 m. Thus we get:



And this is definitely highly turbulent

Maybe you mean something else? However, the kEpsilon is the standard - bad - high diffusive - model one uses as the beginning.
For heat transfer problems, the kOmega should be more appropriate as the frequency is more accurate at the wall regions. However, I am not familiar with the turbulence models so I just can give the following hints:

• kEpsilon is well for free-flows (walls are far away of the point/area of interest such as free-stream flames)
• kOmega is suited better close to walls but is less accurate in the far field
• kOmegaSST should be a mixture of both while the y+ distance is used to switch between both formulations; However, this can be checked in the source code and I do not give any guarantee of my statement here

[julieng]

Hello Tobi,


In the file that I downloaded from you, the velocity is fixed at 0.5 m/s and the width of the channel flow is 50 mm (symmetry plan at 25 mm from the solid wall).



So I don't understand your previous post.


I am using komegaSST in pipes with chtMutliRegionSimpleFoam v5 and I am not able to reach energy balance when the fluid is compressible. I try various wall conditions, y+ << 1. I am not alone to have issues with energy balance that's why your case is very interesting for me.
Can you share your updated mesh please?


Best regards


Julien

[Tobi]

Hi Julien,

I am sorry. I had a second case in use and I was referring to that one. You are totally right. The Reynolds number in the case that is attached here is small and therefore, we do not need a turbulence model.

Summing up
a) I changed the turbulence model -> laminar
b) Re-run the calculation with a non-matching interface mesh (fluid one is finer). The result is given in the picture attached
c) Re-run the calculation with a matching interface mesh (equal faces on both sides). The result is given in the picture attached

Thus, the energy conservation mismatch I observed was/is related to the interface consistency and the mapping and interpolation between the interfaces (I still have some energy discrepancy for a conformal mesh - maybe I should increase the accuracy etc... )

[julieng]

Hello Tobi,

It is conservative at the solid-liquid interface.

But what's about the power at the outlet of the fluid?
Do you calculate the same as the input power coming from the solid part ?

Best regards

Julien

[Tobi]

Hi,

if you check out the graphs in more detail, you see the mismatch at the interface. The energy should be conservative in each region (we can see that it is true for the solid one -> solid_to_fluid == solid_to_ambient)

[julieng]

Hello Tobi,


Yes but the main issue for energy balance comes from the difference between the power measured at the interface and the power transported by the flow
P=massflow x Cp x (Tout-Tin)
Can you chek this value please?


Best regards