displacement control and load control

[Lei Lei]

Dear STKO Team:

I wanna simulate a four-point bending test for a curvilinear beam with a displacement-time control.

I tried two methods.

The first is using the analysis integrator-'loadcontrol' command with a 'sp' constraint determining the ModelType as Displacement. However, it always can't compute the element force.

The second is using the 'displacemntcontrol' command with a 'nodeforce' constraint determining the ModelType. However, it can't converge.

Here, I want to know the use of the above two methods to add a displacement-time history to my model.

[kesavapraba]

Hi, It is not clear where the loads are applied on the structure during lab testing – please provide us with details about the loading location in the geometry.

Can you also provide the drawing of the geometry with the units (mm or cm or m)? As per the geometry from the model, it seems that the base length of your model is only 262 mm! Please also mention the unit system that you used in the model (N, mm, sec?).

EqualDOF is not necessary in your case as displacement control analysis can be set up directly by mentioning a control node on the structure.

I would suggest you to follow the same loading procedure (either load or displacement control) done during experimental campaign.

[Lei Lei]

Yes, the units are N/mm/s.

Because I'm conducting a test with a scaled curvilinear beam model, the geometry is only 267mm.

The displacement-time history is loaded on the middle two points and the maximum displacement is about 20mm.

As for the 'equalDOF' setting, I'm not sure which point should be selected in 'Displacementcontrol' part.

Here, I load two models setted with 'Loadcontrol' and 'Displacementcontrol' separately for more clear understanding.

[kesavapraba]

Hi, the displacement loading history can be applied using LoadControl integrator and single point prescibedNodalValue condition. Hence, Displacement Control Integrator is not required in your case. Following two suggestions are made for LoadControl analysis:

[1] Please note that warnings "WARNING PlainHandler::handle() - non-homogeneos constraint for node 2 homogeneous constraint assumed..." are issued here by OpenSees since Plain Constraint handler is used in the analysis settings. This is because Plain constraint handler cannot handle non-homogeneous constraints (such as SP PrescribedNodalValue). So, you should use Penalty/Transformation constraint handlers instead.

[2] Analysis Settings: I could also note some inconsistencies in duration of LoadControl setting. Please keep it same as the length of the time history if you want to apply the entire loading, i.e., 20 mm in your case. For "Time Step Type", it is better to use "Adaptive Time Step" to avoid any convergence issue. For "numIncr", you can consider using a large increment (1200 is not sufficient) to capture the load path smoothly.

[Lei Lei]

Thanks！

But if I want to use Displacemntcontrol Integrator, what changes should I make for my model?
The related model was posted in the previous post called 'Curve Beam Model Test_Disp Control.rar' .

And, I set a recorder in the analysis part while I can't chart the element-force curve even though It can be shown on the plot group. Can't STKO do this?

[kesavapraba]

Hi,

But if I want to use Displacemntcontrol Integrator, what changes should I make for my model?
The related model was posted in the previous post called 'Curve Beam Model Test_Disp Control.rar' .

This would be simple. You should apply reference nodal loads (vertical) on the two nodes and then use displacement control integrator with any one of the nodes selected as control node and the associated vertical displacement as DOF in the analysis settings.

And, I set a recorder in the analysis part while I can't chart the element-force curve even though It can be shown on the plot group. Can't STKO do this?

STKO can get you the section forces, but you should use the correct option in post-processor. In your case, the dispBeamColumn uses fiber section propoerty with 5 integration points. Please do the following to get the section forces:
[1] In the preprocessor, select section.force option in MPCO recorder. You missed this one.
[2] In the postprocessor, create Gauss Point Plot -> You can see 5 Gauss points for each element (see attached figure below) -> Select the Gauss point where you want to get section forces -> Chart Data -> On element Gauss point -> Section.force (either N or Mz).

I hope this answers your query. Please let me know if you need any further assistance.

[Lei Lei]

Thank you very much for your patient reply!

Now, I've modified my two models with different integrators.

I think I've made sense of the difference between the two methods.

However, the modified model didn't converge.

I am very sorry to consult you about the reasons for its non-convergence.

I apologize for wasting so much of your time and would appreciate your reply！

[kesavapraba]

Hi, You are welcome. I changed some of the analysis settings in your model, now it converges up to the final load increment. Attached is the revised model with LoadControl integrator - you can do the same for the model with DisplacementControl integrator. Following are the changes that I made to your model.

[1] Penalty factors were too high to get proper convergence (so, You should keep it 3 to 4 order higher than the average stiffness of the system; Please note that I have added a custom command 'OOM' to provide your details on the min, max, average stiffnesses of the system)
[2] You can use 'UmfPack' system option for better performance.
[3] In general, you might consider using "Adaptive Time Step" to avoid any convergence issue. This would at least ensure trying reduced time steps before final convergence failure.
[4] Total time from the supplied time history is only 13411.3, but you had used 14000. This means you have applied extra dispacement to the system. This could also cause issue if the system fails.

[Lei Lei]

Thank you very much for your help！

It really helped me solve my problem.

However, the DisplacementControl model still doesn't converge, even though I've made the changes you mentioned above.

Best Wishes!

[kesavapraba]

Hi, your interest is appreciated. I would like to clarify the rationale behind the request for a displacement control analysis. The load control analysis employed involves the application of displacements at specified nodes, rendering it functionally equivalent to a displacement control analysis in this particular scenario. If you require further clarification or information, please feel free to contact for any further clarification. Thank you.