Pick and Place from a pile of components

[Shaun]

Hello
This is my first post, but I have really enjoyed the discussions, questions and videos here, so thanks. They are definitely inspirational.

Firstly, I have spent a fair amount of time looking for the solution to what I am needing, without too much luck, and honestly, I land up getting distracted by other posts. Different searches I have done include "match moving", "feature tracking", "template tracking", "image processing" and "image recognition"

My situation is this : we do metal castings, and then have a huge team that "finishes" off these components by placing them individually onto a trimming tool, which trims off any flash around the edges. A VERY repititive job. They then go for further processing.

What I want to do, is incorporate some kind of pick and place system, where each component will be picked up by a gripper, and correctly placed onto the trimming tool. Unfortunately, the components are delivered to the operator in a bucket, so it would need to somehow discern which ones are on top, and their orientation, before trying to pick it up, something that is obviously very easy for the human operator.

We have already (as a company) started on the route to automation, but the system we have includes a hopper system to separate the components and align them, they then slide down a chute, and are then picked up and placed into the tool. Gives us huge headaches. Also, this kind of system is designed around this particular shape component, (physically), whereas I would think that a Pick and Place system could be programmed to recognise the different components, and so would be portable between the different tools.

Can someone please point me in the right direction, which websites would help, is this an easy system to implement, etc. I would really appreciate it.

Shaun

[Asellith]

Is it possible to adjust the way they are delivered what your trying to do is very complex. If you could predict the part that is coming out of the previous machine and place them into a crate aligned a special way that would make the next step a lot easier but of course you would need 2 pick and places.

Also know the materials in question (ferrous or not) and some rough idea of how different each part is will help us to point you in the right direction

[Shaun]

Unfortunately there is no way to pre-align them, as they come directly from the casting process. About 12 get cast in each shot, and they fall freely into the crate, in which they are then given to this operation I am talking about. The system we use now, they get dumped into a hopper, and this then attempts to align them in some order before sliding down a chute. I am thinking to eliminate the hopper system for a few of reasons, some of them below.

1. Slow (compared to what I believe might be possible)
2. A lot of rejected alignments, meaning they then go to the bottom of the pile and start the whole re-alignment journey again. A waste.
3. Space consuming
4. Current system is full of issues, done by an external company who no longer offer the support we need
5. A need to upgrade/improve our system

The material is a type of brass, so non-ferrous, the specific grade we use is a golden/yellow colour, matt finish.
We produce about 250 000 per month of this particular component. Weight is about 50g, about 70mm long, 25mm across and in depth. We have 3 shifts working this product (and others) on a full time basis, so if it has a payback period of within 18 months, it becomes very feasible for us.

I understand that this is a complex undertaking. Are there systems like this or similar existing? I have seen a few sorting videos posted, but I did notice they were planar placed and individually separated.

[Asellith]

Its possible to use video to do this. use a vacuum to pick the piece up and orient it to the camera use software to position it properly then drop it in place.

Hmm another solution if they are all the same shape or if you have extended runs of several different shapes , you could pick up the item and place it in a sensor field. Probably IR sensors on several axis at key points. Determine the necesary distances and adjust the part till you get the correct profile then drop it in place. This might require that all the parts be oriented a little first. for instance they all need to be lying flat before they can be analyzed.

just tossing ideas into the air here.

[Shaun]

The kind of system I am hoping for is one where the components are delivered, lying haphazardly over one another, and using some type of vision system, a gripper is able to orient itself to correctly pick up one component at a time, obviously the ones lying on top first, from the pile, and then deliver it to the trimming tool. By having to first separate the components and possibly orient them, i would be slowly returning to our existing system, maybe something like our hopper system.

I like your idea of using a suction though, maybe the suction can be set to only be strong enough to support the weight of one component, but then it would always have to suck at the very same place of the component each time, which could be tricky, seeing that the components are all jumbled together. How would the vision be able to discern what is on top, what is lying underneath, if not initially oriented first to lie flat?

Would some kind of neural network be needed for this?

Thank you very much for your input/ideas so far.

[Admin]

Apologies for the delay in response, I just saw this . . .

My situation is this : we do metal castings, and then have a huge team that "finishes" off these components by placing them individually onto a trimming tool, which trims off any flash around the edges. A VERY repititive job. They then go for further processing.

I know a robotics company that makes custom aluminum heat sinks in large quantities. They also had this deburring problem, but have a very unique solution to it. They use a small 'vibratory tumbler', throw in a dozen heat sinks, some type of grit, and let it tumble for an hour or so. They come out pretty clean and nice. The one they use is 93252 by Chicago Electric Power Tools. Get one and try it out. Buy 20 of these things to run simultaneously, and you can reduce your human labor to probably 10% of what you have now for the same output rate.

We have already (as a company) started on the route to automation, but the system we have includes a hopper system to separate the components and align them, they then slide down a chute, and are then picked up and placed into the tool. Gives us huge headaches. Also, this kind of system is designed around this particular shape component, (physically), whereas I would think that a Pick and Place system could be programmed to recognise the different components, and so would be portable between the different tools.

Sounds like a perfect job for robotics, the ability to quickly retool the manufacturing equipment.

We produce about 250 000 per month of this particular component.

That's quite a lot. I'd almost say dedicated machinery would cost you less per unit than robots. Robots are slow, inefficient, and need to be monitored/repaired often. Robots are better for below 10k units/month, and for parts that vary a lot. I recommend running a cost analysis of robots vs custom/dedicated machinery.

There are professional pick and place robots you can buy. I believe this is what you want:
http://www.societyofrobots.com/robotforum/index.php?topic=2727.0