Should I be able to complete a loop of jointed components?

[lemelman]

I've spent a considerable amount of time trying to assemble a structure consisting of 4 equal length round rods and 4 identical right-angled connectors. Each connector containing two sockets at right-angles to each other, each having the same depth and the exact diameter of the rods. In the real world, such a structure would be self levelling when assembled. Three rods could be assembled with 2 connectors, the fourth, with a connector at each end, could then be attached to complete the structure.

In Fusion I can perform the same assembly with revolute joints, until the very last joint, to which Fusion issues the message that the computation has failed. I'm surprised that Fusion cannot create this structure "automatically" with the available joints. Are my expectations too high? It seems to be a rather severe restriction.

[lichtzeichenanlage]

Please File -> Export and share the project. 

[etfrench]

Have you tried using a rigid joint?

Which of the components are grounded?

[TrippyLighting]

---

@lemelman wrote:

Are my expectations too high? It seems to be a rather severe restriction.

---

No, but your understanding of how to get your work done with Fusion 360 is not quite yet where it needs to be.

That was clear to me after seeing you last thread with the pipe an pipe connectors. You are actually making you life much more difficult than it needs to be.

As usual, please share your model.

[lichtzeichenanlage]

@TrippyLighting: Thank you for reminding me of this thread. 

[lichtzeichenanlage]

It's more or less the same task as in your last pipe/joint thread. This time I've simplified the positioning of the parts by using the Create -> Pattern -> Circular Pattern feature.

[chrisplyler]

If Fusion cannot calculate the last joint, then something is wrong. Either all four of your rods are not truly equal, all four of your right-angle connectors are not truly equal, or you've put one or more of the prior joints at the wrong position. Perhaps you made one of the joints such that a rod fits into a connector at a slightly different depth, or something like that. As you can understand, such an error would throw off the square. Or maybe your connectors aren't exactly symmetrical. That would through off the square if you got one of them turned around the wrong way compared to the others.

If you do it right, it might look something like this:

[lichtzeichenanlage]

Using revolution joints is just a wast of computing power, isn't it? Especially if he is still building a wheel chair... 

[lemelman]

Thank you for your interest and comments.

You are correct in stating that this is related to my problems about designing a wheelchair. Thanks to your suggestions and solutions, that particular project is continuing fine.

This thread was created because I felt that revolute joints should straighten the structure by themselves. (Incidently, I really cannot understand how rigid joints could do it).

In the Screencaste I show a simple, but twisted and incomplete, structure of rods and right-angled connectors, which cannot be completed. In the physical world, such a finished structure would be rigid and stable - assuming perfect fits, of course. Fusion is quite happy to move components in order to satisfy a particular joint, but although the last joint looks like it's going work it fails. It could work if Fusion adjusted the other joints accordingly. The purpose of this thread was to determine if my expectations are too high, or if it really should have worked.

Screencast will be displayed here after you click Post.

23960616-2844-4553-8180-114f7a58bdaf

[lemelman]

I agree that it works fine when all the rods are on the same plane, my query is about Fusion's apparent inability to adjust them to make them fit.

[TrippyLighting]

That has nothing to do with Fusion 360 abilities but with your incorrect assumption that such a structure is inherently self adjusting, meaning that there is enough restriction in degrees of freedom for the joint solver to find only one solution.

From a practical perspective if you'd actually take such elbows and pipes and assemble them you'll find that they need a flat surface for assembly if you want that assembly to be flat. The frame would otherwise glue quite perfectly and happily with all mating surfaces touching but not flat, meaning there is not enough limitation in degrees off freedom that would be self-adjusting.

[chrisplyler]

---

@lemelman wrote:

I agree that it works fine when all the rods are on the same plane, my query is about Fusion's apparent inability to adjust them to make them fit.

---

@TrippyLighting, I believe such an assembly IS self flattening.

In fact I have no problem assembling my rods and ells using only revolute joints, and having it snap into place upon the last joint, despite starting with components that aren't all in the same plane.

I maintain that something else must be wrong, as per my previous post in this thread.

[TrippyLighting]

---

@chrisplyler wrote:

---

 

@TrippyLighting, I believe such an assembly IS self flattening.

---

I've assembled a number of such things on the computer, as well as in real live. I can assure you it isn't 😉

[chrisplyler]

I have assembled such a square using PVC pipe and 90-deg ells. This was while I was making a support structure for mosquito netting, to be used at Boy Scout camps. Also, I was a plumber in my early adulthood.

If everything is put together without glue, and assuming no flex and tight fitting ells, it can only go together one way: flat. As my video shows, the Fusion solver figures this out also.

[TrippyLighting]

@chrisplyler Forget that rubbish I wrote. This is definitely self adjusting!

If you stick two 90-deg elbows  onto a straight piece of pipe and pivot them the distance between the elbows end points becomes bigger. If you have 4 identical tube and 4 identical elbows than this will properly join even with 8 revolute joints.

BUT using revolute joints for this is utter nonsense and not a good practice, because the joint sover has to crunch numbers when that is not necessary.

[chrisplyler]

It occurs to me that perhaps when making your final joint, you are picking the rod first that would require the grounded one to move, instead of picking the ell first that would leave the grounded rod alone and move the other stuff into place. Any action that attempts to move a grounded component will fail.

I also wonder why you start with all eight components NOT planar anyway?

[chrisplyler]

---

@TrippyLighting wrote:

@chrisplyler Forget that rubbish I wrote. This is definitely self adjusting!

If you stick two 90-deg elbows  onto a straight piece of pipe and pivot them the distance between the elbows end points becomes bigger. If you have 4 identical tube and 4 identical elbows than this will properly join even with 8 revolute joints.

BUT using revolute joints for this is utter nonsense and not a good practice, because the joint sover has to crunch numbers when that is not necessary.

---

I agree that the distance between the ENDPOINTS will change as you twist two opposing ells, but the distance between their PLANES OF ROTATION will not.

Sure, using all revolute joints for this is silly, but it does make it easy to assemble (and snap into a square) if the parts aren't already in the correct orientation. And you could always make a rigid group of all the parts after the fact. I don't KNOW that doing so would relieve Fusion's solver of having to account for all the revolute joints (because I have no idea how it is coded), but it logically makes sense that this could be the case. If his larger assembly isn't too complicated, these eight revolute joints alone aren't pushing the computational limits of Fusion too much anyway.

[lemelman]

I tried to reproduce chrisplyler's structure, but failed. I've attached my failing version if anybody would like to see where I've gone wrong.

Something else is rather strange - the connectors have turned out to be transparent and I've no idea why, or how to change them to be opaque. Any ideas?

[TrippyLighting · Accepted solution]

I've attached the model resulting from the screencast below that shows how to do this properly.

You can of course experiment with these revolute until until the cows come home but revolute joints are not the proper way to join this.

You might notice that I ground as early as possible. In the screencast I don't ground until after the first compost has been modeled, but in the attached model I've grounded the component before geometry was modeled in it and that is perfectly valid also.

Then I restructured you assembly or rather the order. I assemble each component after it has been inserted. I don't always do it that way myself in my own work, but then I don't run into these problems.

For assembling structures with repetitive components I also show how explicit joint origins can substantially speed up the assembly workflow.

Also notice if you Asse,ble this with rigid joints, you need one less joint than when you use revolute joints.

The reason this does work for @chrisplyler and not for you that the parts have a more advantageous pre-orientation before joining them. In an assembly that is not yet fully joined as in your cad the pieces have an almost random orientation, which leaves a lot of space for the joint solver to evaluate for mathematical maximums/minimums.

I can for example replace al the joints in the attached model easily with revolute joints and it self-adjusts fine when slightly miss aligned before I apply the last revolute joint.

[chrisplyler · Accepted solution]

In the last file you provided, your Rev6 joint is located at the outer end of the ell instead of down inside of it. So the rod isn't seated down in the hole. So it is too long to fit into the last ell. I fixed Rev6 and was then able to make Rev8 without problem. I told you things to check for already, and you obviously haven't checked for them.

I have no idea why the ells are transparent.