Discussion Forum

Hello,
Well, I managed to cope with the problem with my previous post (see above)… I have solved the structural and ALE analysis simultaneously for steady state solution. Afterwards I’ve tried to solve the thermal problem which is transient on the deformed mesh. However I’ve got this error :



Number of degrees of freedom solved for: 34121
Warning: Inverted mesh element near coordinates (0.000305032, 0.000295096)
Symmetric matrices found.
Format not changed since UMFPACK uses unsymmetric storage.
Warning: Inverted mesh element near coordinates (0.000305032, 0.000295096)

Reading the comsol gouides and previous posts from the forum I’ve realized there might be a problem with the elements size in the mesh… my model is 3.2 micron width on 3.2 micron hight… I have about 16 holes in the radius of about 8 micron in the model…
What should be the element size?
What other definition can help?

I hope I’ll get some reply,
Thanks,
Yael

Hi

inverted elements is a WARNING and means that you must a closer look and be extra precautionous about the results, particularly close to these points, or even elsewhere if important BC are attached to elements close to these points. But it does not mean necesarily that your results are all wrong.

There are a couple of other dicussions around on the forum about inverted elements, try a SEARCH.

It means basically that there are some elements with poor quality and when you apply higher order shapes thereon, the topology of the elements changes, therefore, COMSOL reduces the element shape to linear, locally.

I believe that is basically all, you could try to mesh in more detail around the point, but on complex geoemtries, even localising the point and find workaround solutions are not always that easy. Perhaps you have some very small features that could be removed ?

Good luck
Ivar

Ivar thank you,

Well, I have tried to use smaller mesh (refined)… I succeeded in solving for the structural (plain strain) and ALE analysis in steady state.
Afterwards, I’ve tried to solve the heat transfer (bio heat equation module) in time dependent on the ALE frame but I received this error:


"Number of degrees of freedom solved for: 36848
Warning: Inverted mesh element near coordinates (0.000304995, 0.00029512)
Symmetric matrices found.
Format not changed since UMFPACK uses unsymmetric storage.
Warning: Inverted mesh element near coordinates (0.000304995, 0.00029512)
Step Time Res Jac Sol Order Tfail NLfail
0 0 out 1 2 0 0

Error:
Time 0:
Matrix is singular.
Singular_matrix

There are 3701 equations giving NaN/Inf in the matrix rows for the variable T
at coordinates: (1.65e-005,8e-005), (1.65e-005,0.000225), (2e-005,7.65e-005), (2e-005,8.35e-005), (2e-005,0.0002215), ...
and similarly for the degrees of freedom, NaN/Inf in the matrix columns


Last time step is not converged. "


I’ve tried to remesh and even remesh manually next to the fixed border… nothing seems to help…

What do you think my next step should be?

Best,

Yael

Hi

What concerns me most is the singuar matrix message, that has nothing to do with the mesh, but it's saying NaN for T = temperatur variable.

Could it be that you have loaded some materials data from the library that whats requires a T variable that is not defined ?
Try to set a constant T= 20[degC] or 37 or whatever is correct for your model.

Good luck
Ivar

Ivar, thank you once again….

That is concerning… because before I have added the ALE module the simulation worked fine (and I haven’t changed definitions)… so I thought this error was due to the ALE application…

My T is a constant (37 degC)… I must emphasize that before I have used the ALE module I have done the simulation by starting the structural analysis plain strain for steady state analysis and than restarted for the thermal bio heat analysis and this worked fine… but as I realized I am not worling on the actual deformed shape I added the ALE…
So now I have solved steady state for the structural analysis and the ALE application simultaneously… the solution seems to work… but when I have restarted the solver for the bioheat application I have received the error…

My T is constant and I haven’t loaded anything from the material library…

Yael

Just to make sure:

As I am interested in modeling the heat transfer analysis on the deformed shape after applying mechanical displacement using the structural mechanics plain strain module…

Is there any other way to do so besides using the ALE application?

I want the post processing to be on the deformed shape, so the data in certain points will be actually taken front points in the deformed shape and not from the original geometry….

If any one knows any other solution to this (rather than ALE) I will be most happy to try…

Best,

Yael

Hi,

I think the ALE should be the last one to be activated for recalculation
This means that you have to start with

1/steady state for the structural analysis

2/heat conduction application (steady state)

3/and then ALE application. (time dependent) in the meanwhile you can change parameters and even analysis type (time dependent) for the previous applications.

Theoretically this should work.

I hope it helps and good luck,

Cheers

Thank you very nuch, I will look into that…

In the mean time I found something interesting… in the structural analysis module I can select “create frame: in the properties. Than I can run the thermal analysis on the deformed shape created.
The weird thing is that the displacement activated seem to be much larger now… I use the displacement Rx=50 micron and Ry=-15 micron on my top boundary (of a rectangular shape in witch the base of the rectangular is fixed and the other 2 boundaries are free), and it seems as though the maximum displacement are Ry=-30 and Rx= 80 micron…

I must mention that in postprocessing I have set in the deform tab that the scale factor to be 1….

Have I done anything wrong?
I have noticed that when I am not using the “create frame” option it seems the deformation is different and more like the boundary condition I am setting…
What can be the reason for this behavior?

Yael

Hi

there is a trick with V3.5a when you activate "deformed frame" in structural (I use it often as it gives you access to the deformed normal of the boundaries nx2 ny2 nz2). That is that you SHOULD NOT select the "Postprocessing general - show deformed" as then you double the displayed "deformations" and you get non-coherent results.


Hope this helps
Ivar

Ivar,

Thank you very much! This has indeed solved my problem!

Best,

Yael