Heat exchanger: tube/shell

[jrm_1971]

Is it possible to simulate a tube/shell heat exchanger, either parallel or counter flow, in multiphysics by assigning mass flow/volumetric flow rates and temperatures to the fluids in the tube and shell? Video tutorial somewhere?

[jrm_1971]

Hup,

What would cause the attached element count error? The model consists of two parts. The tube part (Part 1) and the fluid part (Part 2). I was running the k-epsilon because LES was giving an mpiexex error. After your recommendation of switching back to LES, I followed the wiki http://wikihelp.autodesk.com/Autodesk_Simulation/enu/2012/Help/0000-Installa0/0150-Autodesk150/0153-... on running with MPICH2. Seems like I read somewhere prior that the k-epsilon solver is the one that should be used in coupled multiphysics (transient). Besides, the k-epsilon allows for tube roughness in forced flow, which is what I have, forced flow. Please advise.

p.s. I have run the analysis with the LES processor with Part 1 & 2 consisting of all tet elements, refined to 35% and Part 1 with block elements and Part 2 tet elements. Both runs return the element count error.

[hupn]

Hi,

Not sure exactly what caused the error message. Maybe related to the sync between post-processing and solving procedure. What will happen if you click the "OK" button? You can check whether those solving procedures are still running from your task manager. If they are still running, you may just ignore the message. For turbulence model choice, it is a limitation currently that only LES turbulence model is available for coupling analysis, the turbulence model option should not sneak into the UI options when doing coupling analysis.

hup

[jrm_1971]

Hun,

I already have done that. Simulation is unresponsive before clicking OK. I have to kill it in the task manager. It is operating on monster machine. Windows Server 2008, 64-bit. Xeon 8 core processor. 8 MB of RAM. Any thoughts?

[hupn]

Hi,

That most likely the un-sync issue.  Since you observed that the solving procedures were still running in background even the UI did not response. Instead of killing those running procedures from task manager, could it possible to try to close the result environment and later switch back; or close the whole simulation environment and reload the model? Just a guess though.

hup

BTW, I thought you mean 8GB(not 8MB) machine. It is not a so powerful configuration anymore nowdays in my opinion.

[jrm_1971]

Yes, 8 GB.

OK,

Here's another solution need for this long saga.

Going back to messages 20 and 21. Regarding the true "transient coupled fluid flow and thermal" Multiphysics:

All tetrahedra mesh. 3/8 conduit (0.4 inch ID, 0.5 inch OD), fluid density 2x air (gas), force fluid flow

For straight conduits and conduit with bends, element errors or convergence are received with the following settings:

1. k-epsilon, buoyancy, turbulence 1 (element error for bended conduit), no errors for straight conduit  (25% tetrahedra)  (120 ft/s, 200 F)

2. k-epsilon, buoyancy, turbulence 0 (element error for bended conduit), no errors for straight conduit (25% tetrahedra)  (120 ft/s, 200 F)

3. k-epsilon, no buoyancy, turbulence 1 (velocity update fails, does not finish analysis), no temperature results for straight conduit  (25% tetrahedra)  (120 ft/s, 200 F)

4. k-epsilon, no buoyancy, turbulence 0 (no errors - converges for bended), no errors for straight conduit (25% tetrahedra)  (120 ft/s, 200 F)

5. LES, buoyancy, turbulence 1 (element error for bended conduit), no errors for straight conduit  (25% tetrahedra)  (120 ft/s, 200 F)

6. LES, buoyancy, turbulence 0 (element error for bended conduit), no errors for straight conduit  (25% tetrahedra)  (120 ft/s, 200 F)

7. LES, no buoyancy, turbulence 1 (no errors - converges for bended), no errors for straight conduit  (25% tetrahedra)  (120 ft/s, 200 F)

8. LES, no buoyancy, turbulence 0 (no errors - converges for bended), no errors for straight conduit  (25% tetrahedra)  (120 ft/s, 200 F)

Is there some setting for conduit with multiple bends that allows for convergence with buoyancy and turbulence set to "1"? I ran the straight conduit as a benchmark test but need conduit with multiple bends.

I did one test run (for both straight and bended conduit) with tets and wedges with "Fraction of mesh size", 0.1 total extrusion distance, 1.2 growth, 1 layer and the inlet/outlet excluded from the mesh boundary layer.

k-epsilon, buoyancy, turbulence 1 (element error for bended conduit), no errors for straight conduit  (25% tetrahedra)  (120 ft/s, 200 F)

Why is Algor hating conduit with bends for this analysis?

[tubbl]

On the note of one-way coupled analysis....you said for the 2nd design scenario you would have both the fluid and solid parts in the scenario, but what material do you make the fluid?  The analysis type makes it so that it must be a solid material.

[jrm_1971]

tubbl,

My fluid was made by

1. mesh-->fluid generation-->internal

2. with "select surfaces" toggled, select internal surface of the tube, on the "Generator Fluid Interior" dialog, hit the "select" button

3. Select the inlet/outlet surfaces of the tube, then hit "add" on the "Generator Fluid Interior" dialog.

Then from the new part that is created on the side menu, select Element Type-->Fluid Flow 3D

[jrm_1971]

Hup/AstroJohn

Any ideas on message 25? It would seem that for the required analysis (high velocity on tube, very low velocity on shell side, volume on shell much greater than tube, small diameter conduit, predictively high Reynolds No.), is ideal k-epsilon with buoyancy and turbulence set to "1". The wiki would lead one to think this is possible.

http://wikihelp.autodesk.com/Autodesk_Simulation/enu/2012/Help/0220-Setting_220/0289-Setting_289/043...

Is there some setting for conduit with multiple bends that allows for convergence with buoyancy and turbulence set to "1"?

Why is Algor hating conduit with bends (both hard and gradual bends) for this analysis? (I ran just the coil, no shell or fluid on shell side first). Same deal with a shell with fluid.

[jrm_1971]

AstroJohn,

Any ideas on messages 25,28?

[hupn]

Hi,

I think that there are several things tangling with your varies testing.

(1) As I mentioned early that k-epsilon option should not be available for coupling analysis, so try not to use it since the behavior could be un-predictable.

(2) As for buoyancy option, this is a control for one-directional/two-directional coupling. With buoyancy, the problem is much more complicated and the analysis run as full two-directional coupling analysis, therefore it is harder to converge. Without buoyancy, it is a fluid->thermal one directional coupling problem, therefore, it is easier to converge.

(3)The element error issue is most likely a un-sync issue between solver and post-processor, or should be something machine specific. Unfortunately, I do not know a good way to get rid of it.

hup

[jrm_1971]

OK,

I thought as an ANSYS user, that was the most user hostile software on the planet. Now I believe Algor/Autodesk Simulation is tied. Two months is way too much to get a solution.

1. As I mentioned, LES provides unbelievable answers in both coupled "the real multiphysics" Analysis Type and just straight steady state fluid fluid flow, then one-way coupled. The post processor results for flow are either way too small or way too big.

2. For a tube/pipe I know my Q and A and have solved for V by hand, from Q=VA. I have applied this V to the inlet and defined a predescribed outlet at the other end. For steady state flow, I have tried both LES and k-epsilon, where as directed in the wiki, defined a predescribed tubulence at the inlet for the k-epsilon scenario. (same surface where the predescribed velocity applied). This is a 3D model. I have tweaked on the predescribed turbulence paramters, penalty options, relaxation control, SSOR. There has to be something that I need to click or change to get something close to my original flow rate "Q".

Also, out of observation, for steady state flow, psuedo time should be set to "1" with let's say, 5 steps. However, for a 0.406 inch diameter pipe, the solver, for some reason, does not like velocities above 0.9 ft/s and pseudo times below 60 s. My required analysis should be, and is set to "1" on the turbulence since the Reynolds number will be well above 2300.

3. If we should not use k-epsilon in coupled "real multiphysics" why is it an option?

4. Autodesk has killed the video help that was out there. Now the whole Algor site with the wonderful videos are gone. Maybe you guys can put a whole video section on this topic  in the Wiki. Sal Khan has been able to teach millions this way. http://www.khanacademy.org/    It is definitely a value proposition for both Autodesk and its clients.

UUGGHH!!! Nothing should be this difficult!!!!

[hupn]

Hi,

I understood your frustration. I would highly recommend you to contact technique suport team. As the forum focus for providing tips, relative simple help etc., the technique team can surely work with more detail issue with you together. And thanks for the advice, your voice is definitely heard.

hupn