Brain storming and mind picking on Robots.

[MJuric]

We have somewhat of a push and interest in automating some of our processes here. Some of this has fallen on my desk since it generally lies in the realm of "Continuous improvement"

I've been walking around looking for areas where automation and robotics would be a good fit in our company. We have the standard areas, machine loading, deburring, welding and assembly that could be done with standard automation and or robots.

The one area that I think would be a perfect fit, not terribly expensive and a great way to get our feet wet without interrupting a "mission critical" system is an assembly process for one of our products.

We have a tombstone product, many of which, have lots of holes that require a thread insert, bushing and cap. The general gist of the system, robot, feeders, endeffectors etc I have a pretty good idea of how to do and what is needed. The area that I think is most open is the robot control.

These tombstones come in many different sizes and shapes(number and width of sides). The hole patterns are also relatively non standard despite the fact we do have a standard grid pattern. Probably 80% of the units we do have non standard hole patterns.

So I'm wondering if any of you have any experience or have heard of something for programming bots that would fit the bill. Things I have tossed around so far.

Whole face Vision – Scan the face locate the holes>Tells the robot where to go (Adds considerable expense, especially since some of our tombstones are large. Creates a constant space consumption issue as the vision system would have to be able to see the entire face)

Robot Vision – Some sort of vision system that would be connected to the end effector. (Robot scans face of robot with the camera leading and identifies holes as it moves. Forced to scan entire face which may or may not be a time consumer depending on scan rates)

Automatic programming based on model – This makes the most sense to me as you could place the model and not only does the robot know where the holes are but the system also knows when to rotate to the next side, number of sides, sides that may not have holes in them, different size holes etc etc.

I just don't have the experience or exposure to robots to know what the best approach is here or if something exists or there is "Something new" that is what I'm looking for. I've done and have used all of the above in different applications, just not with robots. For instance programming from a model is essentially the standard for CNC programming these days. Does something like that exist for robots....if not maybe that's what I should be working on instead of this project I've used vision for locating, inspecting, manufacturing etc in automation before....just not with robots.

I've sent something similar to the above to a couple of our vendors that also sell bots but thought I'd see what others here may have seen or been exposed to as well....1000 heads better than one an all.

Any thoughts appreciated

[Frederick_Law]

We can "teach" our welding robot. It only got laser to check if the parts are there.
If your product is one off, it won't help at all.

[Austin Schukar]

...

Automatic programming based on model – This makes the most sense to me as you could place the model and not only does the robot know where the holes are but the system also knows when to rotate to the next side, number of sides, sides that may not have holes in them, different size holes etc etc.

...

As a layman to both manufacturing automation & robot programming capabilities, I would think that this would yield good results. Couldn't there be a bar-code on the workpiece that is scanned, and would tell the robot everything it needs to know? Hole sizes, faces, components, etc.

[bnemec]

We have somewhat of a push and interest in automating some of our processes here. Some of this has fallen on my desk since it generally lies in the realm of "Continuous improvement"

I've been walking around looking for areas where automation and robotics would be a good fit in our company. We have the standard areas, machine loading, deburring, welding and assembly that could be done with standard automation and or robots.

The one area that I think would be a perfect fit, not terribly expensive and a great way to get our feet wet without interrupting a "mission critical" system is an assembly process for one of our products.

We have a tombstone product, many of which, have lots of holes that require a thread insert, bushing and cap. The general gist of the system, robot, feeders, endeffectors etc I have a pretty good idea of how to do and what is needed. The area that I think is most open is the robot control.

These tombstones come in many different sizes and shapes(number and width of sides). The hole patterns are also relatively non standard despite the fact we do have a standard grid pattern. Probably 80% of the units we do have non standard hole patterns.

So I'm wondering if any of you have any experience or have heard of something for programming bots that would fit the bill. Things I have tossed around so far.

Whole face Vision – Scan the face locate the holes>Tells the robot where to go (Adds considerable expense, especially since some of our tombstones are large. Creates a constant space consumption issue as the vision system would have to be able to see the entire face)

Robot Vision – Some sort of vision system that would be connected to the end effector. (Robot scans face of robot with the camera leading and identifies holes as it moves. Forced to scan entire face which may or may not be a time consumer depending on scan rates)

Automatic programming based on model – This makes the most sense to me as you could place the model and not only does the robot know where the holes are but the system also knows when to rotate to the next side, number of sides, sides that may not have holes in them, different size holes etc etc.

I just don't have the experience or exposure to robots to know what the best approach is here or if something exists or there is "Something new" that is what I'm looking for. I've done and have used all of the above in different applications, just not with robots. For instance programming from a model is essentially the standard for CNC programming these days. Does something like that exist for robots....if not maybe that's what I should be working on instead of this project I've used vision for locating, inspecting, manufacturing etc in automation before....just not with robots.

I've sent something similar to the above to a couple of our vendors that also sell bots but thought I'd see what others here may have seen or been exposed to as well....1000 heads better than one an all.

Any thoughts appreciated

When robots are programmed "automatically" from the model, they may also use vision or some sensors to confirm that the part geometry somewhat matches the model it has loaded. But that checking needs programmed too. At least where you're talking about needing the robot program made "on the fly". This makes me think of product configurators they show and tell at conventions, where the cad models and prints are all generated at the click of the Sales Persons' mouse. In your case you'd need the configurator set up to also generate the robot code.

Our case is a bit different for utilizing CAD models to program the robot, we would like it to help design the weld locations (in conjunction with engineered strength of course), and fixture design before the part and fixture design are finalized.

We've been using robotic welding since the 80s. They are still programmed manually through the pennant. We have been making our own weld fixtures longer than that and in 3D for over 20 years. So we have 3D models of the weldment and the fixture but still don't even consider robot program until the first runs of the parts in the shiny new fixture. We, like Matt, are trying to increase automation and integration with separate systems by using the 3D robot simulation tools in combination with the 3D models of part and fixture before toolroom starts cutting metal.

[MJuric]

When robots are programmed "automatically" from the model, they may also use vision or some sensors to confirm that the part geometry somewhat matches the model it has loaded. But that checking needs programmed too. At least where you're talking about needing the robot program made "on the fly". This makes me think of product configurators they show and tell at conventions, where the cad models and prints are all generated at the click of the Sales Persons' mouse. In your case you'd need the configurator set up to also generate the robot code.

Our case is a bit different for utilizing CAD models to program the robot, we would like it to help design the weld locations (in conjunction with engineered strength of course), and fixture design before the part and fixture design are finalized.

We've been using robotic welding since the 80s. They are still programmed manually through the pennant. We have been making our own weld fixtures longer than that and in 3D for over 20 years. So we have 3D models of the weldment and the fixture but still don't even consider robot program until the first runs of the parts in the shiny new fixture. We, like Matt, are trying to increase automation and integration with separate systems by using the 3D robot simulation tools in combination with the 3D models of part and fixture before toolroom starts cutting metal.

So I've been digging into this a bit further and here's some more info.

1) Typical CAM software CAMWorks, MasterCAM etc can not "Completely" program robots due to axis limitations. IE take a six axis robot and place it on a 7th axis and try to program it with a program that can only do 5axis and you have issues.
2) These programs, CAN get you 90% of the way there though. In short drop everything you need into your CAM software, run paths and export the APT.
3) Programs like Vericut, Eureka etc CAN program robots because they essentially are designed to take as many axis as are needed.
4) Programming motions in G-Code simulators is significantly more difficult than in a CAM system.
5) SOLUTION. Do the majority of the program in your CAM system. Export the APT file and import that into your G-Code simulator. Run the simulation and look for Axis that rotate the wrong way, singularities etc. Make necessary changes in G-Code simulator and export the G-code.

Eventually I will be working on this as we have both CAMworks and Vericut. *if* and that is a big if, this process is relatively painless, this would allow us to program "One off" products and use the robot to do the work.

There apparently are a couple programs out there specifically for robot programming but I don't have much information on them yet.

[MJuric]

We can "teach" our welding robot. It only got laser to check if the parts are there.
If your product is one off, it won't help at all.

Yes exactly. What I'm attempting to do is to work the robot program into the process of programming the part for manufacturing....which I'm also working on attempting to automate, more or less.

If the program for the robot can be created while the CAM for the manufacturing and verification is being created then we can do the work even on one offs.

But for the most part all of this is extremely preliminary.

[MJuric]

As a layman to both manufacturing automation & robot programming capabilities, I would think that this would yield good results. Couldn't there be a bar-code on the workpiece that is scanned, and would tell the robot everything it needs to know? Hole sizes, faces, components, etc.

I'm sure you could but the approach I'm looking at is similar to programming a CNC. Program off the model, drop the G-code into the controller...push button...come back when the part is done :-)

One of the benefits of this approach, at least for us, is that it uses the same software packages we use in manufacturing so we have people already familiar with the software and process.

Every Bot manufacturer has it's own "Programming method" and for the most part none of them are "Quick". Not quick enough for low production runs anyway.

There are few companies that have custom software that does exactly what I'm talking about but only for limited use, IE loading a specific machine from a specific grid etc.

[Frederick_Law]

Our robot programmer use this:
https://octopuz.com/

I think ABB has robot software also.
You might be able to find open source too.

[MJuric]

Our robot programmer use this:
https://octopuz.com/

I think ABB has robot software also.
You might be able to find open source too.

I looked at a couple options like that and contacted a couple different companies but haven't heard anything back.

I wanted to get quotes and see what the cost of the software is. The only problem is now you have another software program to learn. However if the cost is low enough you would be able to skip the CNC interface on each unit as this exports to specific Robot's.

[MJuric]

Our robot programmer use this:
https://octopuz.com/

I think ABB has robot software also.
You might be able to find open source too.

I think every company has there own programming software and methods. I know Mits has their own as do a couple others I've looked at.

[Frederick_Law]

As you know CNC/CAM can't handle all the DoF on robot. Also can't handle collision.
We have a cell with 2 robots on XY gantry welding on both sides of a rotatable fixture.
I think its 20'x20'x20' biggest 3D printer here LOL
We have lots of similar single robot bolt to the ground.

[MJuric]

As you know CNC/CAM can't handle all the DoF on robot. Also can't handle collision.
We have a cell with 2 robots on XY gantry welding on both sides of a rotatable fixture.
I think its 20'x20'x20' biggest 3D printer here LOL
We have lots of similar single robot bolt to the ground.

What I find odd is that CAM has not moved into this area. Our CAM software can't handle the DOF but our G-Code simulation software can. Seems like the CAM software would have, a long time ago, jumped on this.

Instead what we seem to have are a bunch of companies moving into "CAM for Robots" while the "CAM for machining" companies aren't. That's weird.

[Frederick_Law]

Probably they don't think robot can replace CNC.
It is true for subtractive. Robot can't get that high cut rate.
Now if you give the robot laser or plasma cutter.

The big boom is additive on robot. It removed a few limitation.

[MJuric]

Probably they don't think robot can replace CNC.
It is true for subtractive. Robot can't get that high cut rate.
Now if you give the robot laser or plasma cutter.

The big boom is additive on robot. It removed a few limitation.

For the most part Robots can't replace your typically CNC. Takes one hell of a robot to move around a 50HP spindle motor that can hold position with minimal vibration under full load. :-)

That being said they are already dealing with 5-Axis, for them it's just adding, 6,7 and 8 or more. I wonder if this is an issue where their entire software is written around only 5 axis and adding six would be a huge re-write.

On the other hand our G-Code simulation software essentially allows unlimited axis, you just have to have a model and define the control and kinematics of the movements.

Edit to add: I sent a message to the company you linked and they called me back very quickly. I have a phone conference with them at 2 today. I sent another company a request early last week and still haven't heard from them.

[mike miller]

Speaking of robots and lasers.....check this out: https://www.blmgroup.com/en-us/6-axis-l ... ines/lt360

We have one of their plate lasers but nothing like that!

[Frederick_Law]

Finally, a laser killing robot!
The end is here!

[GrahamK]

Take a look at Cenit's Fastsuite it's available standalone or as a bolt on to Catia V5.
Cenit would be able to customise to suit the tasks being performed with robot if required. Will give you both programming and simulation capability.
The downside is that it could be expensive, but it is really good software.

[bnemec]

What I find odd is that CAM has not moved into this area. Our CAM software can't handle the DOF but our G-Code simulation software can. Seems like the CAM software would have, a long time ago, jumped on this.

Instead what we seem to have are a bunch of companies moving into "CAM for Robots" while the "CAM for machining" companies aren't. That's weird.

The company I work at has been using welding robots since the mid 80s, ( I was not working here then, I was in grade school) I do not know the software they use but I know it's not any CNC/CAM software for mills/lathes. Totally different systems as I understand it. There may be some common ground but milling CAM software vs robot software have a lot of things the other doesn't do. MotoSim is what they are using as a 3D kinematic simulation environment.

[MJuric]

The company I work at has been using welding robots since the mid 80s, ( I was not working here then, I was in grade school) I do not know the software they use but I know it's not any CNC/CAM software for mills/lathes. Totally different systems as I understand it. There may be some common ground but milling CAM software vs robot software have a lot of things the other doesn't do. MotoSim is what they are using as a 3D kinematic simulation environment.

When you start talking about 5 and 6 Axis milling machines and robots there are actually a whole lot of similarities. I would not think it would be a "Big" stretch to move to programming robots. But maybe it is.

[Frederick_Law]

If you think about it.
To get to a x, y, z position, there are limited solutions on how to move the tool on 3 axis and even 5 axis.
Consider how many way a robot can move the tool to same location.
Also how many way its joins can move keeping the tool at same location.

[MJuric]

If you think about it.
To get to a x, y, z position, there are limited solutions on how to move the tool on 3 axis and even 5 axis.
Consider how many way a robot can move the tool to same location.
Also how many way its joins can move keeping the tool at same location.

I'm not sure. I suspect that regardless of CNC or robot there are similarly limit positions of joints to achieve any particular Vector XYZ position. The path to get there may be more complex on a robot than a CNC for no other reason that all the axis are stacked unlike with a CNC where the axis are usually split between two stacks.

However in both cases one would think it would be a matter of simply hierarchy as to which axis takes president for which movements.

No idea really, I'd have to play with it.

[Frederick_Law]

On a xyz gantry, any xyz point has only 1 solution for all 3 axis.
On a robot, depend on resolution on each join.

[bnemec]

On a xyz gantry, any xyz point has only 1 solution for all 3 axis.
On a robot, depend on resolution on each join.

Not just the point, but also the angle of the end effector to that point, in case of MIG weld the push/pull angle as well as side to side angle can have bigger impact than the welder settings.

[MJuric]

On a xyz gantry, any xyz point has only 1 solution for all 3 axis.
On a robot, depend on resolution on each join.

I actually don't think so. Again, assuming an XYZ and vector, which is the same for a CNC except for the fact that in a CNC the "Vector" is limited to a single plane, there are very few solutions to any position.

So for instance, the situation below. The green rod shown in space is the vector and the end of the rod is the XYZ position. There are only two solutions.
This...

And this

The only other axis with any degree of freedom is the sixth axis which is essentially the same as a spindle on any machine. So in reality a 3-Axis CNC is really a four axis robot.

The joints in a robot are not ball joints with two degrees of freedom, they are axis with a single degree of freedom. This is pretty easy to prove in a SW model. With the sixth axis locked concentric and the end mated to the end the robot can't move even though all the other joints still have a degree of freedom left unlocked.

In fact I believe that robots don't operate well when they run into situations with "Multiple possible solutions" and they are called "Singularities" and those positions are to be avoided if possible.

Unless I'm missing something, which is entirely possible, there typically are not multiple solutions for robot positions. Even in this case more than likely the first position shown is not possible due to over travel conditions.

[Frederick_Law]

If you got the model, fix the center point at the end or the face.
See if any of the join move.
Unless you're pretty far out, I believe it'll.
The problem is, every moving part need collision check.

[MJuric]

If you got the model, fix the center point at the end or the face.
See if any of the join move.
Unless you're pretty far out, I believe it'll.
The problem is, every moving part need collision check.

That's what I'm saying. I have mated the face of axis 6 to the face of the cylinder. I then mated the axis 6 concentric to the cylinder. Cylinder in space is fixed. Robot no move, it can't. This is typical of most positions.

That however does not mean that there are no other solutions. As I've shown sometimes, some joints could be positive an angle or negative and angle. Many, in fact in most cases, that is even limited by travel.

You can even do it visually at each joint. The bottom joint1 is a rotary at the XY plane. Can it rotate even if the other joints move some how? No, it can't because the only joints that would allow it are J4 or J5. If you move J4 or J5 you then come off axis of the cylinder.

Now certainly there are multiple solutions if you remove the vector aspect, but I've not seen many applications where a certain vector is not necessary. Most end effectors require mating with what it needs to at a certain vector, think drilling a hole. You've got to come in perpendicular to however you want the hole to be. Even applications where you can have multiple different vectors typically you pick one or some combination of vectors.

As stated where I actually think robots have issues are where they DO run into situations where you have an "Unlimited" possible solutions "Singularities". Situations like this where J4 is free to rotate.

[bnemec]

I realize this is a Kindergarten level question and picture, but trying to make sure I'm following correctly. The arrows and labels are correct and what you're speaking of @MJuric ?

[MJuric]

I realize this is a Kindergarten level question and picture, but trying to make sure I'm following correctly. The arrows and labels are correct and what you're speaking of @MJuric ?

image.png

Yep, correct. If you look at J4 in this case it can spin around it's axis. This is actually a case where the possible positions available to the robot are only limited to the resolution of the axis. These cases, at least in my understanding, are rare and are to be avoided if possible.

There are a few of these positions possible with a robot, however, every other position only has a couple possible solutions and in most cases, only two where one joint can be either positive or negative angle.

Keep in mind this is all coming from a complete Novice with robots so could be completely wrong

But from what I can tell from the model the possible solutions for any position are pretty limited. I think the possibilities for getting to those positions are larger with a robot than a three axis CNC, but final possible ending positions solutions are very similar.