2D axisymmetric steady state flow with moving wall heat transfer

Questions and remarks about code_saturne usage
Forum rules
Please read the forum usage recommendations before posting.
lorenzo.iron
Posts: 18
Joined: Tue May 11, 2021 1:22 pm

2D axisymmetric steady state flow with moving wall heat transfer

Post by lorenzo.iron »

Hello.

I've been learning to use Code_Saturne for a few weeks for a research project related to the cooling efficiency of a water box for hot steel bars cooling. For now, I just learned how to use the GUI.

So far, I've prepared a pure fluid flow study, modelling the 2D axysimmetric fluid contained into a particular pipe (the internal shape is a sequence of convergent and divergent ducts, like this >-<>-<>-<>-<... etc...) and removing the central wedge of the fluid section representing the steel bar (I've used YACS to parametrize the geometry shape and to automatize the meshing and .med file generation).

The model is Steady State, and now I'd like to include also the thermal flux from the moving hot steel bar to the fluid, but I don't know how to do this.
I've two options in my head:
1) using the "wall 1D thermal module". Considering that the steel bar wall is moving, is it possible to apply this method? If so, can you give me some indications how to proceed?
2) the second option I see is modelling the steel bar like a very viscous laminar fluid flow and couple the two fluid domains to exchange temperature and axial speed. If feasible, how can I do this?

Do you have any other smart/feasible option to pursue this steady state CFD problem?

Thank you in advance.
Lorenzo
Yvan Fournier
Posts: 4070
Joined: Mon Feb 20, 2012 3:25 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by Yvan Fournier »

Hello,

This might require a bit of trial and error, but we can suggest patches/test small changes in the code as you progress.

- Depending on how much the steel bar moves, you might be able to use a deformable mesh (ALE) approach or not. If the deformation is strong (and using a single mesh would degrade its quality too much), there is still the possibility of using several meshes, and changing from one to another using restarts with a different mesh.

1) With a deformable mesh, either the 1d wall law or internal coupling may work, Though I recommend the second option, as it is a more complete model.

2) If you have turbulence and need wall laws, using a very viscous flow would not allow you to define wall laws at the fluid/solid interface. If you do not have turbulence or are wall resolved, using this aproach (or improving it with a porosity/head loss approach) would actually be simpler to set up than 1), so it could be a good solution.

Best regards,

Yvan
lorenzo.iron
Posts: 18
Joined: Tue May 11, 2021 1:22 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by lorenzo.iron »

Hello Yvan,

thank you for your reply.

The steel bar moves very fast, say 30 m/s, so based on your suggestions I think the ALE approach is not suitable and we should exclude your option 1). Please confirm, thank you.

Concerning option 2), I've not completely got your point (my fault, I'm novice of CFD simulations) and I need some clarifications.
I will surely have turbolence of the water in the volume around the steel bar (changing cyclically along the colling pipe path) whereas the bar can be seen as an inner laminar flow.
- Can you explain better to me why modelling the bar like a very viscous fluid (almost solid) would give me problems? Does it affect the definition of the speed and heat exchange between water and steel?
- I'd need some guidance about the steps to implement this two fluids coupling.

Thank you heartly.
Lorenzo
Yvan Fournier
Posts: 4070
Joined: Mon Feb 20, 2012 3:25 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by Yvan Fournier »

Hello Lorenzo,

For 1), the important factor is not the speed of deformation (because you will probably need to use small time steps), but mostly the amplitude: if the mesh can easily stretch between the initial and final position without too much deformation of individual elements (i.e. if the deformation can be spread over the domain), things are OK. Otherwise, mesh quality degrades, so remeshing for different positions may be needed. And remapping the initial deformation requires some special care / user coding.

For 2), the limitation is not related to theory but implementation details: since in this case faces at the fluid/solid interface are interior faces relative to the computational domain, we cannot easily apply turbulent wall layer models to them, as those are only handled on boundary faces.

Best regards,

Yvan
lorenzo.iron
Posts: 18
Joined: Tue May 11, 2021 1:22 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by lorenzo.iron »

Hello Yvan,

Is the ALE method suitable for a steady state state thermal model?
My water box is almost 4 meters long and in order to reach a thermal steady state condition the steel material shall in some way flow from the inlet to the outlet of the pipe, where I'm interested to read the outlet temperature and compare several runs with several different water box geometries to increase the efficiency. I'm also interested in calculating the shear stress and so the drag force all along the steel bar.

Here below you can see a piece of my current 2D axisym mesh representing the water part of the model, just the very first centimeters of the water box. The bottom surface is the one interfacing with the steel bar.
Mesh.png
How can I apply ALE to pursue my goal?

Thank you thousand.
Lorenzo
Yvan Fournier
Posts: 4070
Joined: Mon Feb 20, 2012 3:25 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by Yvan Fournier »

Hello,

I did not realize you were looking for a steady-state solution. If the bar is sliding along the domain, and the topology of the fluid domain is fixed, then no, ALE is not the best option. In this case, coupling code_saturne with itself is the best option.

Defining a sliding interface for the velocity part can be done simply (it is even available in the GUI), but the current internal coupling is not adapted as it assumes a fixed interface. The internal "code/code coupling" could be adapted for this, but would need a small patch. If had plans for something similar for another application, but probably won't have time to develop it before 2 weeks (in which case I could send you the patch to test on your case). Would that be an acceptable time frame ? Otherwise, I can point you to the sections of the code that would need to be adapted, but this may require hihger development skills. If this is an option, I can describe the main points you would need to adapt.

Best regards,

Yvan
lorenzo.iron
Posts: 18
Joined: Tue May 11, 2021 1:22 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by lorenzo.iron »

Hello Yvan,

it would be great if you could relize the patch and share a working test case, so that I can also learn how it works.
From my side, it would be a very hard task learning the coding from scratch (the learning curve is not compatible with my current commitments, but I want to do it in the near future).

Do you want me to share my currently applied model? (barely the water inside the cooling box with the bottom sliding wall simulating the steel bar with a fixed imposed wall temperature, done using only the GUI).

Thank you,
Lorenzo
Yvan Fournier
Posts: 4070
Joined: Mon Feb 20, 2012 3:25 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by Yvan Fournier »

Hello,

A "small-mesh" variant of the test case could be good, and possibly some illustration.

Thinking about it, using the internal coupling (available through the GUI in v7.0, for which a first RC-version should be released in a few days) might work, simply adding a sliding boundary condition on the fluid side (as you already did), and forcing a bulk velocity in the solid (using a user-subroutine, bu staying on the simple side). This would not actually require any additional development on the code_saturne side (at worse, bug-fixing for features not tested in combination so far).

Best regards,

Yvan
lorenzo.iron
Posts: 18
Joined: Tue May 11, 2021 1:22 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by lorenzo.iron »

Hello Yvan,
sorry for delay, I'm currently abroad with limited access to my email account.

For your test of the "bulk velocity of the solid" feature, please find herein attached my current implementation (a short cooling-pipe version) as a test case.

The applied boundary conditions are the following:
CoolingPipe.png
(5.7 KiB) Not downloaded yet
I applied a fixed temperature to the bottom sliding wall, representing my hot steel bar. The aim is to check the heat flux capability of the different sections of the pipe. By the way, about the "dimensionless_heat_flux" simulation result, what's the formula? (I checked the Theory Manual but I'm not sure if it is I.5.35 or I.5.37 or another one).

I used a k-epsilon linear production turbolence model (I'm not sure that it is the best choice...).
As a wall function, I used the "2-scale model" but I' not sure if the "scalable" one is more suitable for my purpose.
I had to use fixed water properties because when I tried to use the CoolProp ones, the simulation diverges (the velocity residual goes up), even reducing the wall temperature to 100°C. This is an aspect that I'd like to investigate more.

Looking forward to hearing form you soon.
Thank you very much.
Lorenzo
Attachments
coolingPipe.tar.gz
(678.9 KiB) Downloaded 108 times
Yvan Fournier
Posts: 4070
Joined: Mon Feb 20, 2012 3:25 pm

Re: 2D axisymmetric steady state flow with moving wall heat transfer

Post by Yvan Fournier »

Hello,

OK, I'll take a look in the next few days. If you have conditions which could lead to boiling, you might be outside the scope of code_saturne's standard models (and access to the neptune_cfd extension which handles those is quite restricted).

If boiling should not be reached given a good enough heat exchange, then the exiting models should do fine. Computations with CoolProp tend to be very costly, so using CoolProp simply to get a few sample points and fit an analytical formula will probably lead to better performance, though this is separate from the divergence issue, which is probably physics-related.

As for the best turbulence model and wall law, k-epsilon linear production seams reasonable for what you are doing (I do not think the anisotropic aspects of RSM are needed for a "sliding wall" type flow, so k-eplsion should do). The 2-scales law is better when you have heat exchange, but the fidelity of the exchange coefficient will also depend on the y+ values near the wall.

Best regards,

Yvan
Post Reply