How much can we automate?

How much can we automate?

Joined: July 10th, 2016, 2:02 pm

March 4th, 2017, 6:05 am #1


Is anybody interested in, or working on, trying to automate entire build processes rather than just individual operations?

In particular I'm thinking of workpiece movement.

When I kick off a keyboard build, for example, I'm going to be using the mill for several operations. I'm going to use it to cut the PCB board, then I'm going to use it to cut the switch plate from aluminum, then I'm going to use it to carve the individual keys out of plastic to mount on the switches. None of these operations are a problem for the mill - but the build will repeatedly stop as I move workpieces in and out. Move the PCB board in. Mill it. Move it out. Move the switch plate in. Change endmills. Mill it rough. Change endmills again. Mill the sharper inside corners. Move it out. Move a plastic block in. Change to a ball-nose endmill. Mill the keytops (positive) on the left side and (negative) on the right, then separate left and right sides, turn the left side over onto the right (which holds it exactly in place), change endmills again, and mill the key mounts onto the backs. Then separate the keys and move the workpiece (now a tray with the keys nestled upside down in individual sockets) out.

If I had a tool changer and a robot arm whose entire purpose was fairly versatile movement of workpieces, this process wouldn't need to stall until I was ready for final assembly. As it is, it stops at least six times waiting for the human to do something with his clumsy hands.
Quote
Like
Share

Joined: January 1st, 1970, 12:00 am

March 4th, 2017, 8:11 am #2

The problem there is you'll still always need somebody in the loop, to refill hoppers of parts, to provide stacks of materials, to replace cutters, to watch for jams, etc.

Even large professional shops with stacks of very modern equipment have difficulties with automated parts transfer, and "lights out" manufacturing.

The other problem is that you can't generally just add an arm and manipulator- that can be used to place parts into the machine and remove them from the machine, but you would still need somebody nearby to "hand" pieces in for it to grab, and to remove the pieces it transfers. Or you'd have to also add conveyors or tracks to move the parts around.

And that starts getting complex and expensive. It's worthwhile for big shops doing a large number of nearly-identical items- like auto manufacturers. If you're only making a handful of keyboard, or even just a few hundred, you'll spend twice as long designing, refining, building, troubleshooting and testing the additional feed and follow systems, than it would have taken to just run the parts manually.

A cheaper and easier solution would be to simply add mills. Assuming the job justified the cost, having three mills (or whatever number) working simultaneously obviously gets the job done a great deal faster than using one at a time.

Doc.
Quote
Like
Share

Joined: July 10th, 2015, 8:58 pm

March 4th, 2017, 1:10 pm #3

I think doing a batch jobs helps too.
Quote
Like
Share

Joined: March 10th, 2015, 2:41 am

March 4th, 2017, 2:57 pm #4

Is anybody interested in, or working on, trying to automate entire build processes rather than just individual operations?

In particular I'm thinking of workpiece movement.

When I kick off a keyboard build, for example, I'm going to be using the mill for several operations. I'm going to use it to cut the PCB board, then I'm going to use it to cut the switch plate from aluminum, then I'm going to use it to carve the individual keys out of plastic to mount on the switches. None of these operations are a problem for the mill - but the build will repeatedly stop as I move workpieces in and out. Move the PCB board in. Mill it. Move it out. Move the switch plate in. Change endmills. Mill it rough. Change endmills again. Mill the sharper inside corners. Move it out. Move a plastic block in. Change to a ball-nose endmill. Mill the keytops (positive) on the left side and (negative) on the right, then separate left and right sides, turn the left side over onto the right (which holds it exactly in place), change endmills again, and mill the key mounts onto the backs. Then separate the keys and move the workpiece (now a tray with the keys nestled upside down in individual sockets) out.

If I had a tool changer and a robot arm whose entire purpose was fairly versatile movement of workpieces, this process wouldn't need to stall until I was ready for final assembly. As it is, it stops at least six times waiting for the human to do something with his clumsy hands.
And a cnc router could likely do the pcb cuts, if not the plastic. And thus run work in parallel
Quote
Like
Share

Joined: July 10th, 2016, 2:02 pm

March 4th, 2017, 7:14 pm #5

The problem there is you'll still always need somebody in the loop, to refill hoppers of parts, to provide stacks of materials, to replace cutters, to watch for jams, etc.

Even large professional shops with stacks of very modern equipment have difficulties with automated parts transfer, and "lights out" manufacturing.

The other problem is that you can't generally just add an arm and manipulator- that can be used to place parts into the machine and remove them from the machine, but you would still need somebody nearby to "hand" pieces in for it to grab, and to remove the pieces it transfers. Or you'd have to also add conveyors or tracks to move the parts around.

And that starts getting complex and expensive. It's worthwhile for big shops doing a large number of nearly-identical items- like auto manufacturers. If you're only making a handful of keyboard, or even just a few hundred, you'll spend twice as long designing, refining, building, troubleshooting and testing the additional feed and follow systems, than it would have taken to just run the parts manually.

A cheaper and easier solution would be to simply add mills. Assuming the job justified the cost, having three mills (or whatever number) working simultaneously obviously gets the job done a great deal faster than using one at a time.

Doc.
I'm not seeing the diminishing returns here. The value of a system to move workpieces around the shop multiplies the value of every machine in the shop. And if it can reach piles of stock and the 'finish' bins, it can move stock in, and finished parts out of, every machine. And it can move stock between every machine.

So, I see a track of ceiling rails running around the shop, and a couple of different carriages sharing the track, with different kinds of robot arms. As expensive as a couple of regular machines, but it would make a dozen existing machines capable of SO much more versatile use.
Quote
Like
Share

Joined: November 19th, 2014, 12:50 pm

March 4th, 2017, 10:42 pm #6

You're vastly underestimating the difficulty of automating these processes, and vastly overestimating the cost of a pair of hands.

Think about it this way: currently, you can staff your shop with skilled robots at the cost of about $30/hr/robot ($62,000/y, fully burdened). These robots can maintain themselves and program themselves, and include a pair of exceptionally versatile end effectors on their arms which are capable of operating simultaneously on a single part (or on two different parts) and include their own adaptive collision avoidance system. They have a sensor array that can respond to light, sound, touch, and heat, and can do rudimentary chemical analysis of gasses, solids, and liquids in real time.

In order to replace those with more reliable robots, you need to replace their functionality - probably with a robot from Boston Dynamics. Last I checked, their price list doesn't include anything under 7 digits, and you'll still need to program them for every new task that comes your way.


If you're making 5,000,000 pieces of something then it's worth automating to a significant degree. If you're going to spend a day making 100 of something, you're better off just hiring a human.

Quote
Like
Share

Joined: October 28th, 2014, 1:06 am

March 5th, 2017, 12:44 am #7

Is anybody interested in, or working on, trying to automate entire build processes rather than just individual operations?

In particular I'm thinking of workpiece movement.

When I kick off a keyboard build, for example, I'm going to be using the mill for several operations. I'm going to use it to cut the PCB board, then I'm going to use it to cut the switch plate from aluminum, then I'm going to use it to carve the individual keys out of plastic to mount on the switches. None of these operations are a problem for the mill - but the build will repeatedly stop as I move workpieces in and out. Move the PCB board in. Mill it. Move it out. Move the switch plate in. Change endmills. Mill it rough. Change endmills again. Mill the sharper inside corners. Move it out. Move a plastic block in. Change to a ball-nose endmill. Mill the keytops (positive) on the left side and (negative) on the right, then separate left and right sides, turn the left side over onto the right (which holds it exactly in place), change endmills again, and mill the key mounts onto the backs. Then separate the keys and move the workpiece (now a tray with the keys nestled upside down in individual sockets) out.

If I had a tool changer and a robot arm whose entire purpose was fairly versatile movement of workpieces, this process wouldn't need to stall until I was ready for final assembly. As it is, it stops at least six times waiting for the human to do something with his clumsy hands.
If you haven't already automated your tool changes, there's no point in thinking farther. That step is as easy as buying a mill with an automatic tool changer. Start with that simple step first. If that's too hard or expensive, then you sure can't do the rest.
Quote
Like
Share

Joined: July 10th, 2016, 2:02 pm

March 5th, 2017, 4:16 am #8

You're vastly underestimating the difficulty of automating these processes, and vastly overestimating the cost of a pair of hands.

Think about it this way: currently, you can staff your shop with skilled robots at the cost of about $30/hr/robot ($62,000/y, fully burdened). These robots can maintain themselves and program themselves, and include a pair of exceptionally versatile end effectors on their arms which are capable of operating simultaneously on a single part (or on two different parts) and include their own adaptive collision avoidance system. They have a sensor array that can respond to light, sound, touch, and heat, and can do rudimentary chemical analysis of gasses, solids, and liquids in real time.

In order to replace those with more reliable robots, you need to replace their functionality - probably with a robot from Boston Dynamics. Last I checked, their price list doesn't include anything under 7 digits, and you'll still need to program them for every new task that comes your way.


If you're making 5,000,000 pieces of something then it's worth automating to a significant degree. If you're going to spend a day making 100 of something, you're better off just hiring a human.
Well, that may be true. But then again, it sounds very similar to the argument against getting CNC tools at all.

Anyway, as someone else has suggested, the clear place to start is with systems for changing endmills and for changing toolheads. (I've been swapping spindle for plastic extruder to do 3d printing; another swap does flux & solder ... I'm getting some mileage out of the 3-axis robot).
Quote
Like
Share

Joined: November 19th, 2014, 12:50 pm

March 5th, 2017, 4:43 am #9

Believe me, if it can be automated it is automated. It's just much more expensive and time consuming than people assume.
Quote
Like
Share

Joined: January 20th, 2016, 3:38 pm

March 5th, 2017, 6:55 pm #10

I'm not seeing the diminishing returns here. The value of a system to move workpieces around the shop multiplies the value of every machine in the shop. And if it can reach piles of stock and the 'finish' bins, it can move stock in, and finished parts out of, every machine. And it can move stock between every machine.

So, I see a track of ceiling rails running around the shop, and a couple of different carriages sharing the track, with different kinds of robot arms. As expensive as a couple of regular machines, but it would make a dozen existing machines capable of SO much more versatile use.
Every end-milling procedure I've ever seen has resulted in a shower of swarf all over the table, vise, and workpiece. Unless you're cutting it dry for some reason, it's also going to stick to those surfaces, meaning it'll interfere with the vise grip and sensors if they aren't cleaned off.

I suppose you could set up an automated pressure washer to hose it all off between stages, but that would mean a bevy of other problems.
Knowledge is not as simple as having the right answers. Knowledge is a way of finding them.
Quote
Like
Share