FEM heat trasfer analysis

I have a problem with matlab code. I am doing heat transfer analysis in Matlab. One surface is exposed to heat flux, another surface is exposed to convection. Other surfaces are also insulated. I compare this code with 3D Cad simulation. But the results are different. While geometry max temperatures are around 100C in Matlab, they are around 200C in 3D CAD. What is the reason of this?

[Torsten]

Maybe a minus sign instead of a plus sign in line 279 of your code ?

Seriously: How should anybody be able to answer this with the information given ?

[sevgi]

I shared the my cod file in attachment. It is a very easy process, but the solution seems to be wrong.

[Torsten]

And you are sure you solved the same problem with the other simulation tool ?

[John D'Errico]

@Torsten has no need for the crystal ball I have always depended upon. But just suppose he is right in this random guess? Yes, the odds are well under 1% that he guessed the correct line. We would all immediately be forced to acknowledge his supremacy, the god of MATLAB. In turn, this suggests all he needs to do is make similar wild guesses on hundreds of other questions where there is no possible way to know the answer. One of them will inevitably turn out correct, and Voila! I love it.

So yes, probable user error here. There is possibly/probably a mistake in your code. Possibly you used an incorrect boundary condition, or a wrong constant, or some other simple error of infinitely many kinds. Or possibly you made some mistake using the other tool.

If you want actual valid help on something like this, you need to show code, else nobody can ever do more than make wild, random guesses. A problem of course is your code is probably lengthy and maybe messy (one of those reasons is probably why you have not shown it) and that will make it less likely someone is going to offer timely help. That in turn suggests you want to spend some serious time debugging your code. How? Given that we see no code at all, how might you do that?

I might simplify the code. Change some of the boundary conditions. Verify that when you make the problem simpler, you get something that makes sense. If it does, then I'd spend a great deal of time to clean up the code. Simpler code will be easier for you to verify and debug. Yes, I know, this means almost rewriting your code from scratch. But if your code is incorrect and you won't show it, then you have no choice.

[sevgi]

In the simulation, I defined the surfaces specified in the cod as heat flux and convection, I defined the other surfaces adiabatically. I think my analysis in the simulation tool is correct. I thought that Matlab calculates the undefined surfaces adiabatically, is it correct.

[Torsten]

I thought that Matlab calculates the undefined surfaces adiabatically, is it correct.

Yes.

[sevgi]

The Matlab calculation resulted as if the heat conduction in the geometry was very low. However, the defined heat conduction coefficient is quite high. The heat should be spread over the geometry. I think there is a mistake in the code.

[sevgi]

The units of the values defined in the code are what matlab accepts. Could the problem I'm experiencing be due to unit differences? What units does matlab use for Heat Flux, Temperature, Thermal Conductivity, Mass Density, Specific Heat values?

[Torsten]

Units are SI units (except for degC instead of K). W/m^2, degC, W/(m*degC), kg/m^3, J/(kg*degC) for your properties from above.

Note that you set an ambient temperature of 25degC over a large area. I doubt that the heat flux over the small cylindrical radius will spread the high temperature over a large volume in this case. But these are only speculations ...

[sevgi]

I used the SI unit system. The units are the same as in the 3D simulation. Similarly in the simulation, I set the ambient temperature to 25C.

Thank you for your attention, I have given you a lot of trouble.

[sevgi]

In order to verify your idea ('Note that you set an ambient temperature of 25degC over a large area. I doubt that the heat flux over the small cylindrical radius will spread the high temperature over a large volume in this case.'), I defined adiabatic except for the surface where heat flux is defined. The temperature distribution graph I obtained is attached. There does not seem to be much heat transfer within the material.

[Torsten]

Try to set a temperature (e.g. 100 degC) at the cylinder surface and start with a temperature of - say - 25 degC in the domain where all boundaries are set adiabatic. The result should be 100 degC in the complete domain.

If you don't get this, there must be something wrong in your settings.

[sevgi]

I tried your suggestion, but I didn't get a positive result. There must be a problem with my settings.

[Torsten]

There must be a problem with my settings.

Maybe the settings, maybe in the geometry import. Was it a stationary simulation ? If it was a time-dependent one, maybe the simulated time was not long enough.

[sevgi]

Time-dependent simulation. I applied transient simulation. The simulation time is 3600 seconds, I think the time is enough.

After creating the geometry in Autocad, I converted it to STL format. The surfaces were not very clean, they were very fragmented. Creating the geometry in the software provides healthier results in Star CCm+, imported geometry can cause problems. Could the problem I am experiencing be a conversion to STL or a similar problem in Star CCM+?

My geometry is quite simple, is it possible to create this geometry in a matlab window?

[sevgi]

If the analysis is run as strasystate, I can observe a temperature distribution (I have attached an image. There is convection on the outer surfaces). Again the temperature distributions are not the same as the Star CCM solutions. In Star CCM the min temperature of the geometry is around 95. in matlab it is much lower.

What should I think in this case. Isn't my error due to geometry? If it is not due to geometry, what is the cause of the error?

[Torsten]

What about the test in steady-state mode ?

Try to set a temperature (e.g. 100 degC) at the cylinder surface and start with a temperature of - say - 25 degC in the domain where all boundaries are set adiabatic. The result should be 100 degC in the complete domain.

What should I think in this case. Isn't my error due to geometry? If it is not due to geometry, what is the cause of the error?

Again: we cannot tell because too much information is insecure or missing.

[sevgi]

https://www.youtube.com/watch?v=ehh7pnbTnYo

The first application in the video in the link is similar to my study in terms of analysis conditions. In the first application of the video, the analysis is run by creating the geometry in the matlab window. I did the same analysis in matlab and Star CCM without importing the geometry. I got similar results. For this reason, I think the cause of the problem is the import of the geometry. Can I create my geometry in the matlab window as in this application.

[Torsten]

I have no experience with the preprocessor of the PDE Toolbox - sorry.

But from other posts I got the impression that it's not easy to handle and not well-suited for complex geometries. So the import of geometries from CAD tools for the PDE Toolbox should be normal and error-resistant.

Why don't you contact Technical Support ?