problem: results analysis with bodies with huge deformations.. .

[kasak985]

Hi...:redface:

I am carrying out a modal analysis on a fairly complex structure with many bodies.. .

after various analysis, carried out with mesh of different relevance, I constantly got on 10 ways to vibrate well 6 at zero hz frequency and the other 4 ways at low frequencies. in the order of 0.01.
to better investigate the cause of the problem I checked the deformation for each of these ways of vibrating. had these zero-frequency results hz referenced bars full thickness 5mm that Sembravano esplodere!. : their deformation is about tens of meters if not more and from the graphic window it seemed just an explosion... .I would definitely say that the result is unreliable and I think these bars have some problems.. .or there is a problem of setting the analysis that I miss or ignore... .

the material is ok as well as contacts... .

the same problem that I am describing to you earlier also in a structural analysis on another structure always with many bodies, even when looking at the deformations, square-sectioned bars seemed to liquefy...while the rest of the structure remained integral essential... .
this problem was arranged barbariously eliminated by the cad these components....

in summary: have you ever seen among the results bodies that have huge deformations, orders of magnitude greater than the rest of the structure?? ? I mean bodies that seem to deform completely unreasonable? ?I'm sure with experience you can help me.. . .grazieeeeeeeeeeeeeeee e buon weekend a tutti:redface:
ps: thank you for the hand you are giving me... I promise that Grandma will collect the necessary experience I will make available to all users of the forum....:redface:

 

[stefano.garbin]

Unfortunately, I'm not very careful about modal analysis but I suspect there is something that doesn't fit with contacts.
try to take a look at this link all forum xansys and see if you find useful info.

 

[Onda]

the deformed of modal is only a graphic representation, changes the scale of representation.
the most likely thing is that your system is not properly bound.
I don't know exactly how ansys treats contacts in a modal analysis, but I think either they're "glued" or the analysis doesn't work. I do not see how you can treat contacts (nonlinear solution) in a search for autovalues.
If you get low frequency, the basic motivation is that you have some untied part. and if it is not bound its deformity is huge compared to the other pieces as the component moves
verify all links between the parties by binding or suppressing unscrupulous parts.
Let us know

 

[luigi.paiano]

you absolutely must not look at how numerically the movements in modal analysis are worth because they are defined less than a constant!!! !the modes display only the form!! you have 6 zero frequency modes because they are associated with rigid body motions, of course the model has a system of constraints that allow the 6 gdl in space.
eye that the ways depend on how the structure is bound!
in your case you probably have a much more rigid area than the other:typical bad conditioned case. there are things to do in this case: I recommend the book on the finished elements of the cook.

 

[luigi.paiano]

Could you attach an image of the system?and maybe explain how it is bound and what the modal analysis you're doing

 

[kasak985]

Thank you very much for your interest! !

after numerous and exhausting attempts I solved going to minuziosamente check the model cad...in the steps that they had delivered me there were solids inside other solids that therefore created great lability for the model.
In addition, I increased the number of elements so that there were at least 2 or 3 for the thickness of the bars that gave problems.. .
at this point I have carried out numerous analyses: the first with satisfactory results lasted approximately from 5 to 6 hours....untime unthinkable. . .
I subsequently created a new project file, here, with the same mesh settings the analyses last about 10 minutes!!!! the results are now satisfactory since they give frequencies in order 10^2 hz...lontane from the own frequencies of the moving organs of the machine (electric motors).

I will not take into account the calculated movements, but only the frequencies.

Let's say that for now I skipped her.. I went to a new analysis, and I'm getting into new things for myself, so I'll soon have a new discussion. . .
:redface:

 

[kasak985]

I would say that the main cause is mesh because in the latter case the cad and contacts are controlled, re-checked and previously gave real results in other analyses.

I think this was the problem: the concurrent of elements with coarse meshatura (1 element per thickness) with elements with fine meshatura (3 elements per thickness) should send in resonance those with big mesh.. .
remains the fact that the same problem happens with components having lability.

 

[Onda]

It seems strange to me that the problem depends on the mesh, unless you have labile elements so with the unmatched edges, you should check the free edges of your mesh. (I don't know how to do ansys).
Basically a coarse mesh should stiffen the pattern and at the limit hide ways. Of course not to make it labile. Never seen anything like that.
I therefore think the problem is elsewhere.

 

[stefano.garbin]

wave observations are very correct.
why not do a preliminary static mini-analysis to verify that there are no poorly bound components? Just apply gravity force and see immediately if something falls into the void.. .

 

[kasak985]

In fact, in fact, I had great lability in the cad that I arranged with manic care....the results at this point were ok....

then in a later analysis with a revision of the model (without lability) I had these problems of instability...in this case I noticed in the mesh that in a "body sizing" I had previewed for two series of parallel aluminum bars, missing one of the two series. that is to say that the model had a row of meshate bars with elements to 2,5mm i.e. 2 on the thickness, and the other row of meshate bars with the automatic in coarse relevance 0....the result is that the bar with mesh much grosser exploded in fact in the result of the analysis. At this point, once applied the sizing body also to the bars with coarse mesh, the results gave acceptable values and in line with those found in the previous revisions of the model.. .

Strange, isn't it? ?

Thank you!:biggrin:

 

[stefano.garbin]

Yes, very strange.