I am currently reworking a more stable model that could serve as a reference.
thatās the problems, bolt interaction is complex so it cannot be simplify. Some simplifictaion model approach is usefull and reasonable well but not too simpler.
You have basically three options to model the bolts:
- Simplified solids (cylinders) with contact
- most robust (in terms of avoiding issues and CalculiXās limitations)
- most accurate, can account for any kind of behavior
- quite easy to define - once you draw and mesh one, you can copy the others
- can be easily preloaded
- can be easily checked for forces (using Section print)
- need nonlinear contact (problematic for convergence)
- Beams attached via coupling or rigid body constraints
- most accurate 1D elements
- can be preloaded, but this will remove them from the model
- forces can be obtained with
*EL FILE, SECTION FORCES, OUTPUT=2D
S, NOE
- most tedious to define
- may not work well (or at all, especially in ccx 2.23) with rigid body constraints
- Springs attached via coupling or rigid body constraints
- quite easy to define if you know the stiffness (spring constant)
- least accurate 1D elements
- preload can be applied only indirectly as initial displacement
- you would need a few of them - one per direction
- forces canāt be obtained directly (only with some tricks based on e.g. strain energy)
I think that you should try more than one approach to see what suits you best.
I skipped even more simplified approaches with no elements at all (just rigid body or coupling constraints) because they are only used to obtain forces without actually modeling the bolts themselves.
below my schematic model of possible approach to capture all physical bolt response i.e clamping, bolt shear and bending, plate hole bearing, prestressed and axial tension only, output bolt forces also. As can be seen look complex even using beam element and named simplification, schematic model is appropiates for shell part only.
i donāt know about feasibility to implement due to complexity of mesh dependent, but capability to modeling 1D element by two nodes or multiple (master and slave nodes) can help much in this task.
.
Source of the above diagram (without the colorful lines): http://www.icmas.eu/Journal_archive_files/Vol_5_No_3_PDF/167-172_Piscan.pdf
This ccx forum topic can also be helpful regarding force distribution in a bolted joint: Pretension shear/tensile resistance - CalculiX (official versions are on www.calculix.de, the official GitHub repository is at https://github.com/Dhondtguido/CalculiX).
But I think we should now let the OP prepare some basic test model and try out the different approaches we proposed.
right, iām reading this paper when response to discussion and used his graphics as basic. Colored line is large extension to the paper references were limited to axial response only.
it seems not convenient when simplification need to represent bolt physically, writing and reading extra keywords would like spaghetti code.
Trying out a few approaches (like the 3 Iāve mentioned) on a very simple model with two plates and one bolt can be really beneficial even just to become familiar with CalculiX syntax (needed to work with more advanced models and features not supported in PrePoMax). And to be more aware of the capabilities and limitations of these different approaches. Such simplified tests are IMO always good.
1 Like
Hi,
This is an inp of my proposal.
I have defined a Distributing coupling so you can test force or moment. The inp contains a distributed moment as example.
Beams can be B32R (Circular) or B32 (Only rectangular section supported). Iām using B32 because it benefits from additional integration points needed to get an accurate Section print result .
The bolt has a surface defined at the middle of the shank. We can compare this approach with the ones presented by FEAnalyst based on the different interpretations of the term āsimplicity/accuracy ā.
You can also check if that approach is suitable for your loading conditions. No BC on the distributing REF node or be am.
Note the beam RB rotation is sustained only by the friction on the contact area. No preload on the beam in this example
The model is NLGEOM and solved with just one time step to speed up the example. Rememeber there is no magic formula when talking about Simplified models.
All contact areas only work under compression.
1D-Simplified_Bolted_connection_II.inp (582.3 KB)
EDITED: IF YOU DO NOT EXPECT SEPARATION IN THE CONTACT AREAS OF BOLT/PLATE YOU CAN USE TIED TO CONSIDERABLY SPEED UP THE ANALISYS
This wonāt work in Abaqus due to the way the surfaces are defined on beams, but hereās a visualization with beam profile rendering:
But why is so large contact stiffness and slip rate needed here ?
*SURFACE BEHAVIOR,PRESSURE-OVERCLOSURE=LINEAR
2.1E+13,1
*FRICTION
0.3,210000000000
My units are Pascals,m..It is more or less the same shape we are talking about.
Good to know the aproach is new.
Yeah, this might be one of the rare advantages of having āfakeā beam elements. The recent version of Abaqus can also handle various kinds of contact with actual beam profiles (normally, they were approximated as circular), but via the general contact algorithm.
Thus, in Abaqus, shells are typically used for the head and nut when modeling bolts with beam elements and needing more accuracy with bolt-member contact included.
in CalculiX all six-side face of beam known before to be working with contact no matter it is tied or penalty type, maybe this is new for Abaqus.
however, modeling bolt using single element of solid may lead to problem in large deformation, contact and plasticity. I still prefer coarse mesh of quadratic solid element to model complete bolt parts, so the problem and issue can be tracing due to its general solution.
Itās much easier and pretty much natural for CalculiX to offer all sides of beam elements as surfaces because they are actual faces of solid elements. Any solver with true beam elements (whose geometry is just lines and profiles are used to determine the cross-sectional properties) needs much more effort. In such solvers, beam profiles are usually ignored or approximated as circular (with the radius of a circumscribed circle around the profile). Abaqus only recently improved that: Exact beam contact
Internally, it generates a mesh of quadrilateral surface elements and those are used for contact (similar things were done manually in the past).
Beams can be refined while keeping simlicity and economy. They can also capture bending compared with springs.
Solid bolt elements are rare and expensive but indeed the easiest to set up when no other options are implemented.
The inp shows a real working aproach. Not possible, real. One can take it, test it and validate if it suits his particular problem.
This is what happens when you talk based on papers.
itās not from paper but referencing to my office software used SAP2000, rigid body reference point should transmit both translation and rotation. Also, modeling bolt like that is not my goals as can be read by many of comment related to bolt modeling.
You probably canāt because the surface used to stablish the contact is not exposed in the GUI. It is an internal face of the beam. I havenāt seen any other one doing this in calculix before. It was possible to acces internal faces in Prepomax at some point in the past but Matej removed that option.
I think we only discussed the availability of various beam faces on the CalculiX forum when talking about section print: How to print the section moment (SOM) at the section on a 1D beam (B32)? - CalculiX (official versions are on www.calculix.de, the official GitHub repository is at https://github.com/Dhondtguido/CalculiX).
When it comes to interior face handling by PrePoMax, hereās the related discussion: A new PrePoMax version v2.4.0 has been released - #6 by Matej