Modeling compression only springs 2

[Stazz]

Is there a way to model compression only springs in an FE analysis. I know how to model for both compression and tension but the only way I can think of how to do compression only would be either to do an iterative analysis or a step loading analysis.

Iterative:
1) Assume springs tension and compression
2) Solve Model
3) Remove all springs in tension
4) Solve model
5) Remove all springs in tension
6) etc.

I might suspect that after the first iteration that the model may be instable since the forces that were obsorbed in the tensions springs now must be obsorbed by the compression spring.

For the step loading analysis:

1) Apply fraction of load to model
2) Solve model
3) Remove springs in tension
4) Add next fraction of load to model
5) Solve model
6) Remove springs in tension
7) Etc.


It would be efficient if this could be incorporated into the matrices so that a single run can analyze this compression only case.

 

[EdR]

Depending on the software you are using you may or may not have a no tension element or constraint available...(or perhaps a tension cutoff on the material property). Usually these capabilities are present in non-linear material capable codes.....If so that will do what you need.....If not then either of your methods will work and neither should become unstable.....

Ed.R.

 

[GregLocock]

Or use a contact analsysis

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Onda]

If you use Nastran:

You can make a non linear analysis with non linear material, very stiff in compression and very soft in tension.

Or

You can use the C-Gap with paramter CDITER
see this thread

http://www.eng-tips.com/viewthread.cfm?qid=178285&page=1

The linear solution iterate until all gaps are resolved.

 

[GBor]

Any type of line element to which you can apply a stress-strain curve: Negative strain, stress changes so that the slope of the curve results in a modulus with the appropriate spring stiffness. Positive strain, slope is flat (zero modulus). You may have to give it a very small slope for positive strain...some softwares don't like zero modulus.

Garland E. Borowski, PE
Engineering Manager
Star Aviation

 

[ThomasH]

Hi

Just one comment reqarding your method.

When you solve the model and remove springs the second time. Then you are working with a new structure. How can you be sure that none of the springs previously in tension become compressed?

I have heard of a software that works something like this and from what I've heard there can be a problem. The solution can start "jumping" between compression and tension. What you do, if I understand correctly, is to linearize a nonlinear solution.

Why not use a nonlinear solution instead?

Good Luck

Thomas

 

[GregLocock]

I think removing each spring as it goes into tension is a robust approach, as it mimics the real process. ie one at a time and then re-analyse.

Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[cbrn]

Hi,
Greg Locock's point seems correct to me.
Only, you will probably need more than two iterations (which seems to be ThomasH's point).

However, if you use a non-linear element which "automates" this, be careful to avoid the oscillations mentioned by ThomasH (e.g. in Ansys with COMBIN37 elem). In this case, just set the elongation where the element stops acting NOT to zero, rather to a very small value (1E-9 for example).

Regards

 

[Stazz]

Thanks guys,

I'm just trying to brainstorm a process for coding these springs into VBA excel. The iteration seems to be the only way since you can't account for nonlinearity within the confines of a linear matrix system. I was just hoping some one had some black magic around this that I could cook into the data for a single itteration.

More specifically its going to be an easy spread footing analysis so I would suspect convergence.

But that may be a good point ThomasH that if I remove the tension springs then how can I be sure that these springs aren't needed for compression in the next model? If this is a problem then I think the step-loading itteration method will be a way around this.

 

[cbrn]

Hi,
well, the black magic could be that you already know the overall behaviour of the structure, so that you place the springs only in the side(s) where you know compression will occur.
Apart from this very special case, I fear that iteration is mandatory...

Regards

 

[IFRs]

I have been using Cadre recently. It includes tension only and compression only members as well as axial spring members, lots of standard shapes and materials and elastic foundations.

 

[ThomasH]

Hi

I'm not sure that removing one spring then reanalyze and remove another is a robust approach. Primarily because you can't be sure that when you remove the 51:st spring none of the previously 50 removed springs go back into compression.

However, If you use that approach and also check that the previously removed springs still are in tension. Then I think it should work. Just check the new node distance vs the original node distance.

Since the springs have an on/off behaviour it should not be to difficult to handle. If the springs are nonlinear, then it's another problem.

But the complexity very much depends on your geometry and loads.

Good luck anyway

Thomas

 

[cbrn]

Hi,
need not necesasrily to be very complicated: at each calculation iteration, get the differential displacement between the end-nodes of each spring element: compression? OK, this element will remain "alive" in the next run; tension? the element gets switched off. It's what most FE-programs really do when they use non-linear elements which have a "cutoff switch", i.e. for example the "compression-only supports" mentioned by IFRs.

Regards

 

[truckcab]

Your first strategy is a well known strategy to simulate contact if your code is linear and you do not have any contact capabilities. Depending on your application it can work out just fine. For force distribution within a structure it might be ok. contact stresses will of course not be very good.



Live Long and Prosper !