Before you do any extra work, that is not a problem, and is how it should be, the modal only uses the undeformed mesh, but the [K] matrix from preload is taken into account.
There is no sensible way that you can add a deformed shape from statics with loads onto a linear modal analysis which has no loads, and hence no absolute values on deformed shapes.
You’ll be mixing apples & pears, I believe.
maybe it still possible to extract relative displacement magnitude of all modes from Frd available since displacement of previous steps of perturbation also know, create new deformed mesh as base and replot back. However, a lot of work may need and i did not know in detail.
If you do that, you’ve lost the original model setup, and you’ll be solving a different project.
If the deformed mesh was able (somehow, you can’t really) to be passed onto a modal with same constraints, then you have already a different [K] matrix on that model vs original CAD.
If then, on top of that you pass the [K] matrix effects from static → modal ==> you have a mess that makes no sense; i.e. both the geometry changed and on top of that the [K] matrix too.
Can you pls explain with a real case scenario what you guys are trying to model here? That may shed some light onto whether it is possible from pure maths of modal solve.
i’m not sure is displacement magnitude in mode shape plotted by CalculiX Frd files are relative or it has been improperly mixed with absolute displacement from previous steps when perturbation activated. But it seems relative not mixed, extracting is needed for last condition. Create new mesh based on deformed position only for plotting result purpose if needed.
this related to how Abaqus does in plotting since previously shown at deformed position as base?
The debate is becoming very interesting. 
Let’s use the string problem as example.
I have computed the first mode . In both cases there is a perturbation parameter but in one of them the NLGEOM is active in the First Step.
These are the results. NLGEOM=ON option seems to be taking the updated length of the string (deformation) into consideration . The string is larger after the load process is completed and the resulting frequency is reduced.
Love this, now we are talking, hand calcs!
You’ve done a complex string problem, you are comparing 2 different engineering cases, both the length and mass changes. Now, spanner in the works, just to makes us all think, see if we arrive to the same bit of paper as a concept on what linear dynamics is:
Guitar string is simpler:
FEA check:
We are not modeling the same, right.
Guitar String is not a good example in this case. My string is not a guitar string. My lenght change, my mass is constant and my load increases. It is intentional.
A Guitar strings has a fixed lenght determined by the distance between nut and the saddle. Not the same.
-My string is free to increase it’s lenght in the longuitudinal direction. It has increase 88mm due to tension (T). Recall the name of the post : “Frequency step on a deformed structure”. This is what is under inspection. ¿Is Frequency + Perturbation taking deformation from previous step into consideration? . I think it is.
Yes, you’re right.
So, your test jobs replicate your hand calcs, don’t you have what you were after? Taking into account a preload K change into a modal FEA?
Not really.
I don’t understand why the solution doesn’t reflect vibrations around the deformed shape. Seems Matej is inspecting this.
By other hand, seems NLGEOM is required as in ABAQUS, not only Perturbation.
I’m still not sure how far one can push deformations.
And finally, my hand solution and Fea do not Fully agree so there is still something else (473Hz/479Hz) (Maybe Poisson Ratio, ¿?,…)
This seems it was my doubt when it wasn’t mine. Really nice observation from TKsir.
I have caught a couple of tricks in the exchange. Thank you.
I post the inp with nlgeom example in case someone sees something weird.
String_FREQUENCY_PERT_C3D20R_NLGEOM.inp (81.6 KB)
I think you have it in your nice table (?).
LHS: M=0.001kg | RHS: M=0.001028kg ==> +3% heavier.
LHS section = 10e-7 | RHS section = 9.448e-7 ==> -6% less stiff?
SQRT(0.94/1.03) = 0.955
500Hz * 0.955 = 477.5 ==> close to the FEA (?).
It’s hard to get freq hand calcs spot on with 3D geometries, but close enough I think?
True. But both displacements can be normalized and scaled. The important benefit would be that the frequency animation would oscillate around a deformed position.
