Modeling rope with contact for a prosthetic hand 2

[rstupplebeen]

I am working on a project where joints are actuated by rope and bungee cords. Both of these are modeled with 3D splines because they follow a complex path. Can someone provide some best practices for this type of system?

Ideally neither the the rope or bungee would support bending or compression loads. These should also be pretensioned.

Thank you very much for your help.
Rob

Rob Stupplebeen
OptimalDevice.com/blog
Rob's LinkedIn

 

[rstupplebeen]

I have made some progress:

https://polymerfem.com/showthread.php?195-Cable-element

suggests using the *NO COMPRESSION keyword

http://www.eng-tips.com/viewthread.cfm?qid=215151

recommends a truss element and hints at using different material properties to reduce the bending stiffness.

I am still happy to hear of any suggestions. Thank you.
Rob

Rob Stupplebeen
OptimalDevice.com/blog
Rob's LinkedIn

 

[rstupplebeen]

Well I now have second order hybrid truss elements (T3D3H) for a single rope. Convergence seems very unstable even with Automatic Stabilization on. Because I'm working with dumb solids, on the CAD side, the spline isn't laying on the hand and is floating in space near it's resting location. I expect that since the truss elements don't have any bending stiffness that they are causing the convergence issues. Once the rope is laying on the surfaces of the hand this should be alleviated. I'm running Abaqus 2016 so many of you probably can't open the CAE file but I would be happy to post it.

Any suggestions are welcome. Thank you.
Rob

Rob Stupplebeen
OptimalDevice.com/blog
Rob's LinkedIn

 

[cooken]

Rob,

I think you will struggle to use truss elements, as you have already experienced it seems, as they are rather unstable especially in contact simulations. I'm not sure your convergence will improve once you make contact in this case. You may want to try B31H instead, using the generalized section properties to reach an acceptable axial/bending stiffness ratio. I'm not sure the compression trick would work in that case though...although I have never had much luck with the *NO COMPRESSION option anyway. I wonder if you can use cleverly placed connector elements instead to prevent any compression into the cables? If you can try to prevent any compression through pre-load though, and just double check that none occurs during the analysis then that would be best.

Hopefully we'll be getting Abaqus 2016 soon enough!

 

[rstupplebeen]

Cooken,

Thank you for your feedback. I just made the change to B31H skipping the *NO COMPRESSION. The dummy material properties are several orders of magnitude less in everything except axial. I assume that if compression occurs it will easily move out of the way without taking much axial load. Is there a reason you did not suggest B32H instead? With the complex spline geometry I thought that the curvature may help.

Thank you again and good luck getting Abaqus 2016. When you do this may help. [URL unfurl="true"]http://optimaldevice.com/blog/installing-abaqus-2016-tutorial/[/url]

Rob

Rob Stupplebeen
OptimalDevice.com/blog
Rob's LinkedIn

 

[cooken]

No problem! Reason I suggested B31 is because I have found linear elements to behave better in contact simulations. This will be application dependent though so you may get better results with B32 in your model. Worth looking at both as part of a mesh refinement study for sure.

As another aside, make sure you take the time to set up the model to facilitate initial contact clearance adjustments depending on which contact algorithm/enforcement you use. Trying to get the preload in the wires is likely to give you headaches if you don't smooth out the clearances. You should be able to achieve convergence without relying too much on stabilization or nonlinear contact stiffness etc...

 

[IceBreakerSours]

This system seems to be better suited for a rigid body mechanics solver, and not Abaqus (unless, of course, you are interested in more.) Solvers such as OpenSim (or, its commercial version, SIMM) and AnyBody are not only efficient at solving the system of equations but also offer features such as wrapping surfaces for line-surface contact and muscles for actuators.

In Abaqus, I would not think of anything else but connectors (in series) for each actuator. I think there is a way to define contact between an end-point of a connector and a surface so that should allow you to model linear elements wrapping over joints. Also, connectors can be assigned a wide variety of material properties including tension only linear/nonlinear elastic/plastic behavior .. .. with/without friction .. ... and much more.

*********************************************************
Are you new to this forum? If so, please read these FAQs:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083

 

[IceBreakerSours]

And convergence issues can also arise from *No compression (because of the strong discontinuity at zero stiffness).

*********************************************************
Are you new to this forum? If so, please read these FAQs:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083

 

[rstupplebeen]

IBS,

Thank you for the insight. I am going to put the incompressible desire to the side. I'm still getting acquainted with the project and I now assume that the bungees are tuned to never go slack so this seems like a good assumption.

I am thinking about going over to Explicit because of all the sliding contact.

Thanks for the help!
Rob

Rob Stupplebeen
OptimalDevice.com/blog
Rob's LinkedIn

 

[cooken]

Explicit...fine take the easy way out

Achieving an implicit solution will be a challenge but definitely doable. I would echo IBS sentiments on connectors, they are quite versatile.

 

[rstupplebeen]

IBS,
I didn't see your first message. Your right a RBD solver may be a good option here. I was hoping to model in FEA to help improve it's strength to weight ratio. It may make more sense to couple the two simulation types together. It might be involved because of the liberal use of sliding contact in the design with passive pulleys.

Cooken,
I have never seen Explicit to be the easy way out!! Problems like this I hit a wall then bounce back the other way because there couldn't possibly be as many issues on the other side.

Thank you all again for the help!
Rob

Rob Stupplebeen
OptimalDevice.com/blog
Rob's LinkedIn

 

[IceBreakerSours]

If you are not iterating (such as in an optimization of some sort), then you do not need to worry about using an FE code. If you are, then you have something to think about because the computational expense and time to solution will increase dramatically.

Those cords must bend when they come in to contact so the compressive stiffness will come in to play, wouldn't it? I also recommend studying the seat belt type problems in the documentation.

Also, about connectors, I end up having to play with them separately (say, one or two connectors) before you get a good handle and control on their functionality. There are a lot of variables that you may (or may not) have use for.

*********************************************************
Are you new to this forum? If so, please read these FAQs:

http://www.eng-tips.com/faqs.cfm?fid=376

http://www.eng-tips.com/faqs.cfm?fid=1083