hi,
i’m observing results of gap element with NLgeom deactivated shown larger than activated. Any reason and explanation why the support stiffness shown too soft when Nlgeom deactivated?
below an example models, left plate using *Spring element and right plate is *Gap element with the same value in stiffness.
model
result (Nlgeom deactivated)
result (Nlgeom activated)
The cause is an inappropriate linearization of the Pressure-Overclosure curve around d=0.
Keeping just the linear term leads to a Stiffness which is 1/2 the user input.
If you increase the stiffness of the Compression only support x2 you will obtain the right comparison between Surface Spring and Only compression when NLGEOM is non active. They both become the same.
Not really a bug but Compression only support without NLGEOM definitely needs some adjustment on the Stiffness.
thanks for some hints, i do simple test by multiplied stiffness of two and results shown almost identical but not exactly the same in numbers.
i’m not really know how it becomes inappropriateness in linearization. Is this a limitation or bug and possible to fix in the solver codes?
Mine is exactly the same. If you share I could take a look.
The issue is easy to understand if one looks at the Pressure Overclousure curve. This element behaviour is nonlinear, especially arround zero.
If one linearize the response, one takes the original behavior, develops it in Taylor series and keeps up to the first linear term. In this case, the linear approximation around zero is completely distorting the behavior of the element because it reduces its stiffness by half. The slope of the linearized function is 1/2 the slope of the original function in zero. NLGEOM OFF change stiffnes too much.
I think it can be easily fixed. If the user forget to activate the non-linear behavior and there is a Only compression defined, just multiply the user input Kx2 before sending the inp to the solver. The result will recover the original stiffness again. (Probably a warning will be useful in the monitor)
What I don’t know is if you can find a way to disable the springs that are in tension. I mean that NLGEOM=OFF Only compression Support will not be an only compression. It would behave like a Surface Spring Support where all the springs work no matter the force direction.
i found some explanation related to this problem, even in coupling sections not gap element.
when linearized mean as described above and lead to soften the spring by half, probably this approach is unreliable. Maybe tension stiffness zone need to ignore and use only compression stiffness.
indeed, an option of Nlgeom deactivated will ignore tension limit. User expected the same result as spring element, but not to be softened.
I don’t follow you. I am not considering Coupling at all. I am describing a problem that the gap element presents when it is linearized according to the title of the post…
i’m only try to seek by keyword in official documentation but did not find “linearized” and “Nlgeom” related to gap element
some description available at coupling sections for general condition in related to nonlinearity and Nlgeom deactivated. So, my understanding is the gap treat as linear spring, the problem is in softened by its approach.
oh sorry, now I get it.
Yep, Compression only is intrinsically made with nonlinear springs. Nlgeom Off Linearizes its behavior around an overclosure d=0. Therefore, the Springs softens and becomes linear, like a regular spring.
if this meaning of limitation in solver approach and it can not be improved, some notified in documentation is need to be available to clarify.
quadratic type also problem in node to surface contact due to opposite in sign. Abaqus treat this condition for mid-side node values to zero, however CalculiX use small value instead.
something that i’m not really know, is this actually zero value of stiffness or blank (skipped) i,e being ignored?
Well, I would not call it limitation. This is like if one wants to use plasticity as linear. If you want you can but , to capture a nonlinear behavior, you need a nonlinear approach.
not something like these, only expected similar behavior as linear spring element. It seems Nastran used linear spring also when Nlgeom is deactivated, but probably not to be softened by half in values.
maybe this required new treads to discus further. Negative sign at mid side nodes reported to be working for spring stiffness. However, i’ll stick with linear element first due to reason previously notified.
Don’t you mean corner nodes?. Midside nodes seem the right place to assign the springs.
sorry i missed at this, it seems the sign is reversed of loading scheme in solid face.
it seems NLGeom still can be deactivated with some additional definition of dummy nonlinear materials, it’s known to be working as expected and been implemented in PrePoMax.
The definition of a nonlinear material triggers silently the nonlinear solver. Even if the material is not assigned to any element.
it seems not exactly the same since number of iteration significantly different. These ratios about 1:5 in my case, i guessed discrepancy related to tangent stiffness in NLGeom activated to become longer in computational times. Result also have a little bit different even almost the same for normal case, but i did not know in details how CalculiX actually does.
add some example:
NLGeom deactivated, dummy NL materials (fast and stable)
NLGeom activated (longer iteration, stability problem)
