CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   STAR-CCM+ (http://www.cfd-online.com/Forums/star-ccm/)
-   -   Temperature of solid/fluid interface inconsistent (http://www.cfd-online.com/Forums/star-ccm/111977-temperature-solid-fluid-interface-inconsistent.html)

rks171 January 18, 2013 19:02

Temperature of solid/fluid interface inconsistent
 
I have a fluid flowing around a solid cylinder. The cylinder has a power applied to it. I setup an arbitrary probe using a VTK file to get the temperature on the surface of this cylinder at many different point locations. I then setup an XYZ table that tracks temperature and uses the arbitrary probe as its part.

I see a very steep temperature gradient just by switching the part that the arbitrary probe tracks. If I select the cylinder wall from the fluid region, I get 576 K. If I select the in-place interface from either the fluid region or the solid region, I get 635 K. If I select the solid region, I get 682 K. The position of the arbitrary probe is set to the same location, but it is set to track the temperature of different parts, so I was surprised to see such a large difference just by changing the part. Is my method for watching the surface temperature a bad one? I want to be sure I'm watching the right numbers and not ones that are 100 K from where they should be.

mk_mard January 19, 2013 10:08

have you used prism layer in your mesh and is your mesh fine enough?

rks171 January 19, 2013 14:05

No, no prism layer. I have experimental temperatures to compare my predictions to and it seems like I'm overpricticing all of the measurements by a good 20 K everywhere. Do you think increasing the fluid mesh density near the surface would increase the heat transfer and drop the surface temperatures? I would imagine that the solid region mesh would be of less importance, since I would expect a linear temperature profile anyway.

mk_mard January 20, 2013 10:40

I would suggest using prim layer; hence, finer mesh around the interface.
Solid mesh is also important.

rks171 January 21, 2013 13:27

I mis-spoke earlier when I said I was expecting a linear temperature profile in the cylinder. Actually, it is cylindrical geometry, but it is a tube rather than a solid cylinder. I would expect a logarithmic profile of temperature reflecting heat being generated inside the material. But it doesn't look like this at all. See the plot below. The theoretical temp profile is shown by the green line, while a probe placed inside the tube region in STAR-CCM+ is shown with the red dots.

http://s19.postimage.org/al8t3ef9v/temp_prof.png

This is what my mesh looks like now:

http://s19.postimage.org/qi7l047o3/bare_mesh.png

Does this seem too coarse to properly capture the physics? I guess I will try adding two prism layers and add another two layers inside the solid.

abdul099 January 21, 2013 14:22

Indeed I would recommend a finer mesh. Even without some detailed knowledge the mesh looks pretty coarse to me. And I would also use some prism layers on the fluid side since the mesh quality is critical there at the wall.

rks171 January 21, 2013 15:27

Thanks, I'm increasing the refinement now. Also, I noticed that meshes of the solid and fluid are not conformal. Will this impact accuracy as well? If so, how would I make the meshes conformal? I use two different mesh continua for the two different regions, and the manual seems to indicate that this prohibits the meshes from being made conformal during meshing, but is there any way?

rks171 January 22, 2013 17:12

Adding the prism layer did drop the surface temperature of the cylinder by about 15 K so far, but the case is not yet converged, so I'm not sure what the final value will be. It does appear that it will be much less than before. I added one more layer of cells to the solid tube, but it has had no impact at all on the temperature drop across the tube wall yet. I figured it would at least cause a small change. The surface temperature is still off by another 20 K, so I'm planning on refining the prism layer more and adding a second one to see what effect that has. I'll also add another layer in the tube to make it 5 total, but I'm not sure that will help things with the tube temperature drop.

Pertaining to the conformal mesh, I reread the manual and it turns out that a conformal mesh will not be generated for different regions that use different mesh continua. So I reduced my model to one mesh continua and tried again. It still didn't come out conformal at the interface. I reread the manual again and it looks like a conformal mesh cannot be guaranteed at interfaces between regions where one region uses the embedded thin mesher, which is what I use for the tube (unless that region is completely enveloped by the bulk region, which it isn't in my case since the inside of the tube is blank). So my options seem to be, either use a polyhedral mesh in the solid tube, which will probably lead to poorer quality cells and a higher cell count in order to get my conformal mesh, or else deal with the mesh not being conformal. I guess I'll stick with option 2 for now.

rks171 January 26, 2013 10:42

I ended up making another, more refined mesh, by adding a second prism layer. But I switched the tube mesh to polyhedral instead of the thin mesher so that I could: 1) mesh the whole thing in parallel and speedup up the mesh time and 2) get a conformal mesh at the interface. I found out that it wasn't the mesh refinement in the tube that was causing the wrong temperature profile in the tube. Rather, I was applying the energy source at the wrong location. I was applying it to the tube face, found under the 'boundaries' node. I changed this by specifying the total energy source under the "physics values" node for the whole tube conductor. The temperature drop across the tube is now exactly what I expected it to be.

However, the interface temperature is still about 25 K too high. Adding the second prism layer didn't drop the temeprature any further. If I select the tube surface under the fluid region node, that temperature matches what the tube surface temperature should be, so there is still a 25 K discrepancy between the tube surface and the interface, which supposedly are modeling the same thing. I'm beginning to doubt that further refinement can get this temperature right. I don't know what other mistake I could have made in my modeling though.

http://s19.postimage.org/nofid54k3/b...sh_refined.png


All times are GMT -4. The time now is 02:21.