Comparison between different subcases... Can I automate this?

[score33]

What I'm looking at is this:

Given a structure and a set of applied loads, I have a number of subcases where small loads are applied at different places on the structure. These "small loads" are small when compared to the given applied loads, which will be present in each subcase.

If my first subcase is {structure, applied loads} and each additional subcase is {structure, applied loads + small loads}, is there a way to analyze the results with some sort of script to determine which of the {structure, applied loads + small loads} cases has the least amount of change from the {structure, applied loads} case?

I hope that makes sense... I don't have a strong background in FEA so I am not sure where to begin looking for info on this problem, so any information to references would be very helpful.

Thanks in advance for any help.

 

[rb1957]

you'll have to think about how you'll measure "least amopunt of change".

it's simple enough to subtract (structure+applied loads+ small loads) from (structure + applied loads), 'cept you don't need to ...
if you run the "small loads" on their own you can assess whihc creates the smallest internal stresses, the small external reactions ... however you choose to measure "smallest change".

 

[score33]

That's the part I'm struggling with I guess...

how to measure the change. My initial thought is to just sum the magnitude of the difference between the elemental results or my reaction forces for {structure+applied loads} and each {structure+applied loads+small loads} and select the one with the smallest difference.

 

[rstupplebeen]

Remember that linear superposition only work for LINEAR systems. So if you have plasticity, contact, nonlinear geometry, etc than be aware of this deviation from the assumption.

To RB's point are you concerned with displacement, maximum strain, etc?

I believe these 2 questions will help guide your decision. I hope this helps.

Rob Stupplebeen

https://sites.google.com/site/robertkstupplebeen/

 

[score33]

the system should be linear.

right now I'm trying to look at strucutres that I can represent with beam elements, so am mostly worried about the moments at all the joints (beam element ends) and the moments at my reaction points.

I can easily throw out cases where the moment exceeds certain limits at a joint or reaction, but after that I still can't seem to find a good solution to pick the "best" subcase left with out looking at them.

 

[rb1957]

it's odd to want the least amount of change ...

maybe if you look in the statistics of the run you can find total strain energy, so you'd want the set of "small loads" with the least strain energy ?
that'd give yo a criteria for selecting a particular set ... what it means in real life is somewhat questionable ...

 

[score33]

yeah,

what I'd like to look at is robots that are capable of building certain types of structures autonomously, given the end configuration. So incrementally adding pieces to the structure until it is complete. I'm thinking about modeling each possible next step as a small load at the locations where the next pieces could be built.

So my thinking is that I'd have a bunch of subcases that all correspond to possible next moves, but then I need a way to pick the "best" move. Which I'm finding a little challenging.

 

[rb1957]

then "best" probably has little to do with internal strain energy, or change from previous, and more to do with closer to achiving the target.

i'd suggest that the robots think before they build (always a good idea). presumably they'll have only a few modules to use to build the structure, so the first goal should be to design a strucuture to meet the goal ... if i have to support a load 10' above teh ground, and i have only 1' boxes, then i'll need to stack at least ten of them together. maybe i've got light wt and heavy wt boxes (heavier being stronger). maybe i have to worry about side load, creating bending forces in the box (so that maybe the lowest box needs to be stronger) ? boxes can be designed for load input on a corner, or on the center of the face.

but let's think for ourselves for a minute ... if we can design a robot to build structures ... are we building the end of our careers ? will the robots care if their biulding falls down and damages some nearby (useless) carbon-based lifeforms ??

 

[score33]

I'll stay away from the questions you posted at the end.

For now I want to look at giving a final configuration that may be unstable at intermediate stages of the build, and have the robot figure out the build path so that the final assembly can be achieved. And you're right, at any stage in the build there will be a finite number of choices, I just want a metric for the best choice.

So after talking a little with a coworker, I think I'll just look at throwing out cases that violate my moment criteria at the joints and reaction points, then choose the "best" case as the one that minimizes the displacements...

 

[rb1957]

could the robot "learn" what's "best" ? ie given a strategy does it discard a (potentially/ultimately) winning design more often than not ?

 

[score33]

I could probably add some learning in there, but I'll keep it simple at first and stick to more of a rule based approach.

I haven't found too many examples out there where they try something like this. Most of the self-assembly and self-reconfiguration stuff assumes that the structure is always stable, so you never have to worry about it. My goal is to move away from the assumption of stability and (eventually) be able to look at dynamic structures.

 

[rb1957]

the easy check on is a structure stable would be "is the CG within the base footprint ?"

 

[IceBreakerSours]

Is stacking up a lego model a good way to visualize your ultimate goal?

If so, let me think out loud:

Response to small perturbation might be a way to think about stability and come up with a threshold of stability. So, while a structure with heavier boxes at the top may have the CG fall within the base footprint, it may not be a very stable structure.

Another perspective is to think of the structure's response to a vertical compressive force.

~!ce.

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

 

[score33]

I hadn't thought about a response to a small perturbation before, I'll think about it and see if it could work.

For the strucutre, I'm thinking more about truss type structues where each joint can support some number of struts connected to it. For example, a joint could have six struts attached to it, one along each axis in the plus or minus directions. So basically I want to stick with beam models for the time being...