Heyho,
just wanted to know if anyone can share how you are modelling bolted connections with and without preload.
I want to try if i can utilize the machine learning part classification from cubit to semi automize this.
if i can understand properly a one set of bolt contains five parts, one bolt shank and two of ring and nut for each end. This is sufficient for non-prestressed bolt but prestressed needed to split by two for bolt shanks, it may still be possible without splitting using initial condition type stress. Ideally the bolt shanks have three parts to avoid complexity in slave-master surface definition.
tie constraint or contact type tied need to be used since each part is separated at nut to bolt shank and nut to ring. Surface of ring to flange, bolt to flange holes and flange to flange worked as penalty contact. This can be general setting in commonly contact analysis.
@xyont have you ever tried to preload with the stress initial condition? How would you check the actual preload?
@NorbertH yes, but itās not in use for real case since all solid model not effective compared to imaginary beam element. In building and bridge steel structure bolt connection did not allow mixed friction bolt with bearing type.
could it be more specific in checking since initial condition type stress are direct modeling of prestressed bolt?
full 3D modelling recommended is in this link, reproduces an Abaqus example correlated with a paper
@xyont i now tried to preload with initial condition type stress. it is working. but one would need to make a run just for checking the actual preload force with section print and adjust the stress until the wanted preload is reached. But it is working and the bolt shank wouldnāt be split for preloading similar to a thermal loading.
@JuanP74 thanks for the info, i will take a look
itās related to output report of bolt force and yes section print keyword can be use, also for all solid models of prestressed bolt. In bridge or building steel construction, prestressed bolt is type of friction bolt or slip critical connections not to be mixed with standard bolt type bearing, but pipe flange as given in picture example can be different case. Clamping by small amount of prestressed bolt is required to avoid flange gap and rubber gasket as filler probably needed also.
yes right, prestressed bolt by initial condition type stress can be similar with thermal shrinkage in early days FEA practice, bolt shank length is shortened.
I have recently compared Pretension by means of load with Pretension by means of length adjustment using as reference a model that was hand calculated.
The reference in case you are interested was suggested by Calc_em user.
My main interest wasnāt the Stresses at the thread. Iām assuming that the type and number of bolts are correctly selected for the service.
My interest was the behavior at the contact area as it is critical for flanged pressure vessels leak control. Gaskets need to keep a minimum seating pressure in service condition and do not exceed a maximum in assembly condition not to break it.
What I can say according to my results is that preload by means of load canāt properly capture the separation point nor the external load split into Bolt and Contact Area.
It keeps VM constant in the bolt during the external load increase. (External load increase between 1s and 2s in search of the contact separation point )
Length adjustment can capture the small bolt load increase (FSA with itās corresponding Stress increase).
Preload with length adjustment has deliver an unexpected deviation at the separation point below 0.02%.(FEA result 2059.6KN against analytical solution 2060KN)
Length adjustment has the drawback that you need a first run to adjust the length. Preload by force canāt capture post separation behavior. Convergence fails in ccx.
@Disla are you able to share the models? I am interested on checking some of the values since 99% of my bolted joints are via the force method.
Juan posted two good examples here:
He found a way to do it with load fixing the node which I couldnāt find the way to make it work for this particular problem.
1 Like
as i understand by testing, extra keyword generate by PrePoMax *Boundary, op=New just before user keyword shown in picture make discarding and lead to unexpected results. Result of first step did not combine by adding to second steps, the problem also has been reported and discussed previously in their forums.
but, maybe the condition of OP is different due to operating system, self-compiled or another executable distribution from somewhere.
1 Like
Theory predicts the load on the bolt increase with slope =Load factor
Iām still unabble to preload with Force and keep the preload node fixed on second step and get the expected result.
Hi again,
I suspect that @JuanP74 approach works because one side of the Bolt is fixed to the ground (In his example there is symmetry boundary condition and Bolt is fixed in longitudinal direction at itās midsection.)
One side fixed + Node fixed makes it equivalent to the option distance fixed (preload by distance) which is known to work properly.
For an arbitrary bolted connection, fully modeled and unconstrained (like when being part of a bolted connection) , I would say fixing the preload ref node is not enough and doesnāt work. Itās not a *BOUNDARY OP=NEW or OP=FIXED card issue but a kinematic unconstrained model issue. I guess this is my case/problem.
ĀæHas someone been able to model a full free bolted connection with preload by means of force?
Ideally It should be possible to model free bolts if the external loads are in balance. A system like NorbertHās picture but with ends closed + internal pressure. 3,2,1 method should be enough to remove rigid body motion but I donāt think that kind of models are possible with JuanP technique.
I hope Iām wrong and someone got it and can share his/her approach.
Regards
EDITED: Pretension when fixing one side, works properly.The main issue Iām finding is when the model is completely free to expand in both directions (I use *Equation between two oposite nodes forced to move equally but in oposite directions).
Iāll try later. But remember that FIXED means freezing the BCs selected at the end of the previous step, the preload node controls the MPCs of the pretension section so is more complex. I do not see why the method shouldnāt work, but the convergence can be more difficult since the contact conditions after separation are more complex.
1 Like
iām not yet to run the model, it seems the node of boundary with fix option is unrestrained in previous step then make it work, extra keyword of new option not affected.
Thanks for all the input so far!
Cubit got some functionality to reduce bolts to simpler geometry.
Would be glad if someone also shares examples for modelling those with calculix.
1 Like
looking illustration figure of blind bolt type, it seems clamping from bolt head may not represent well and needed to extent. Node center distance between top and bottom part also too close and almost merge with zero length, usually it has some distance for take advantage of nonlinear spring element. There will be two elements of spring, one represents normal or axial and other is transversal or shear, bending is being ignored for simplify models.
image from Coreform, 2025
HI,
Iām extending Jackub Prepomax Tutorial 34 to include thermal load. My approach is being based on Preloaded Joint Analysis Methodology NASA Technical Memorandum 106943, 1995.
Iām obtaining a deviation of a 4% of the Expected axial bolt load due to uniform thermal increase according to the manual formulation.
That is good as it means FEA model response in regards Kb and Kj is very close to the analytical aproach .
Despite the good results , at this point I canāt find the source of the small deviation which Iām almost sure could be improved.
Iām considering
-First a possible error from my side
-Limitation of the theoretical approach itself. Iām not sure if it assumes no deformation of the bearing area around the bolt head.
-Some misbehaviour of ccx as the load on the bolt doesnāt increase linearly (I know Don Guido has been changing some of the thermo mechanical coupling behaviour which is still not available). I will wait for nect release with expectation.
I have lately read that preload + temperature increase is assumed by various users as it must end up in a Preload lost or ārelaxationā. Me myself was also convinced about that as it seems the most intuitive prediction but according to my results and the manual, thatās not always true. It depends on the expansion coefficients relationship in-between other facts. I recomend the paper. Itās a nice abstract that goes to the point.
The Prepomax result graph shows bolt reaction load to preload (step 1), Thermal load of increasing DeltaT (Step 2) and final External load on the preloaded and heated connection up to the separation (Step 3).
This is my approach in case someone wants to try with their own parameters or maybe suggest some improvement.
Sorry if itās a little messy but I have end up doing many different combinations until the model has been completely free to expand during the thermal load.
Another possible source of discrepancy is that NASA memorandum assumes the stress field looks like a frustum of a hollow cone.
In this case (very thick clamping areas) the shape of the actual stress distribution looks more like a cylinder and the assumption is inappropriate.
A cylindrical Stress Field Method (Q Factor) would be more appropriate to evaluate the Stiffness of the clamped areas for comparison.
It assumes the ābarrel shapedā stress field. Thatās the one used in the endeavor bolted connection.
1 Like