Excessive displacement

I try to simulate a step of pushing ball into cavity. Here, 1/8 of the entire geometry is used. For BCs, I prescribed a motion of 1.85 mm. Simulation run well, but during the post-processing, displacements seems too big. Max value, shown on the legend is according to BCs.
Is this purely graphical issue or truly physical error? There is also a warning message when loading the results.



The compressed *.pmx file is too big to share it here (i.e. 12MB). Anyway, here is the link to the file.

Deformation is scaled by default. Change it from Automatic to True scale to see real deformation.

Ah, yes! :grinning: My mistake … I really did not think about scale factor.
Many thanks @FEAnalyst .

The warning message informs the user that there are some differences between the initial nodal positions of the model and results. This usually happens if you use the contact setting Adjust = Yes and CalculiX moves the nodes before running the analysis.

@Matej
thanks for additional explanation. Indeed, I used “Adjust” for contact condition.

I setup the simulation model (1/8 of entire geometry is considered) and run the analysis. Several things could be observed:

  • in the 2nd step (i.e. step 2), 1/8 model behaves as there are no symmetry BCs … although fixed BCs are prescribed in local cylindrical coordinate system (defined by *Transform)
  • consequently, simulation in the step 2 takes lots of time (with numerous increments) → analysis failed
  • then, I tried to run simulation in different way: I create new simulation and import *.inp file (generated by the previous failed simulation) … in this case, both steps are OK and simulation finish successfully
  • I noticed that in the new *.inp file there is no “op=new” statement and also “Fixed” statement is omitted next to Boundary (*Boundary, Fixed does not appear) … I do not know if this contributes to the simulation, but at least this is what I observed.

With respect to the first three lines, did I miss something?
Model is available on this link.

Thanks in advance for support.

Better use U2 = 0 mm instead of Fixed. The latter has quite specific application in multistep analyses:

the FIXED parameter freezes the deformation from the previous step, or, if there is no previous step, sets it to zero

Yes, it works nice with U2=0. Thank you @FEAnalyst !

2 Likes

Hi Celikan,

This kind of models are very helpful for the users. Thanks for sharing.

ÂżMay I ask you why is there a section excluded from the Transform Card?

ÂżIs that intentionally? ÂżIs there a Slot/groove on the punch?

Regards and Thanks again

Hi @ANYS !
This section does not belongs to symmetry plane and should therefore not be included in BCs for tangential movement restriction.

Hope it helps.

¿And how do that “partial symmetry” translates in terms of geometry?

As mentioned above, 1/8 of entire geometry is considered.
Hope it helps.

Hi Celikan

This is my interpretation of your BC in terms of geometry.

There is a small offset in the main plane YZ (0.1mm) which would produce 4 cuts and 4 overlaps in the model. Seems like using small slices require an extra care in the geometry preparation. Another option is to offset the transform coordinates deltaX=-0.1

By other hand I would say removing a recessed face from the transform represents 4 big grooves in the socket.


Hi @ANYS !
Yes, this is the correct interpretation. I have also noticed the overlap in YZ plane; however this does not impact simulation itself because all BCs are given with respect to local cylindrical CS.

Sure.?Âż

Check the following pmx file. The origin affects the result. I have always understood it like this, and it seems to behave like this. Is this a bug?. I’, have been very lucky. :sweat_smile:

Hi @ANYS !
I have checked you pmx file. The difference you get is logical because you have loads defined in the same cylindrical coordinate system (CSS), in theta direction in particular.
So, as both models rely on the same CSS, it is normal that theta coordinate for model 1 is not the same as for model 2 → force direction is not the same for both models → results are not the same.

I have modified your transformation definition in order that both models have their local CCS, located in the most bottom left node, for each model respectively. Then, FEA results are the same.


I have deleted *Transform command, so it means global CS is used. In this case, results are the same (because load is applied in the similar direction).

See also your modified pmx file (analysis-5 refers to the 2 local CSS).
Transform_modCelikan.pmx (50.2 KB)

Hope it is clear for you.

Now imagine that instead of plates you have two copies of your model.
Of course, yours is not as offset as mine, it is just to be more noticeable.

ÂżWhere should the transformation axis be placed if your model is offset?

Âż0.0, 24.5, 0.0, 0.0, 25.5, 0.0 or -0.1, 24.5, 0.0, -0.1, 25.5, 0.0?

Good morning @ANYS !
Yes, you are right concerning definition of my local CCS. The correct definition should be this one:
-0.1, 24.5, 0.0, -0.1, 25.5, 0.0 (so, offset of X component for -0.1).

Actualy, I had these coordinates in the beginning, but then I have no idea why do I changed X component to 0.0.
So, thanks again.
Have a good day.

Hello @celikan,

May I also have access to the model file?

Many thanks,
JW

Sure, I grant the access.