What's new
What's new

Maybe we need a different approach to automation

garyhlucas

Stainless
Joined
Oct 17, 2013
Location
New Jersey
It occurs to me that we are trying to automate machining by using robots to do things the way a human would. VMCs have a moving table which works well with a human loading. However if you want simple low cost automation we really need machines with a stationary table. If the table doesn’t move you can use the spindle with a gripper or an auxiliary gripper on the side or front of the spindle to pick a part from tray or conveyor and place the finished part in a second tray or conveyor.

This same concept applies to chucker style vertical apindle lathes. Parts get picked, machined and discharged by simply moving the spindle.

I am currently applying this concept to a 16 spindle pipe drilling machine. It was intended to have 2 operators and instead we simply added 4 grippers on slides. A pipe gets picked up by a gripper on one side of the spindles and the finished pipe is picked by a second gripper on the opposite side. The new pipe is placed in the clamps while the previous part drops in a cart. No robots, and way faster than a robot or human.
 
Fixed table machines of the "Mill" variety are not so common any more.

I agree the spindle can be a useful robot-ish.

MAZAK have been using grippers in spindles for quite a while - not sure how popular they are vs robots arms and pallet systems.

Completely agree quasi anthropomorphic robots / arms being unnecessarily complicated and slow versus machines designed to be automated by simpler more efficient machines - but that's a more dedicated factory mindset I guess vs. broad spectrum flexibility of supposed mixing and matching of different robots and machines in the future for different jobs.

I think Dennis of @dstryr fame posted something a while ago using a brother and grippers and a 5 axis trunnion for automation of small parts processing. (If I remembered that right ?).
 
This has been done for at least fifty years ..... not exactly a new idea, gary :)

Yes I am aware of that. I am only saying that maybe we don't need robots that work like humans but should instead focus on how to make the whole task easier by using different machines. Wheelie King has a robot with a fast mill. What if the mill was optimized to work with part feeding instead of humans? He is space constrained, eliminate the robot and maybe you would need less space. Using the robot you still need to present parts to the robot to pick them up, as we are still a bit away from vision systems that would allow random picking from a box of parts. So why not just present them right to the mill? I was involved with packaging machinery and virtually all of the effort was getting the parts to be packaged oriented correctly. Once you had that it was pretty easy to package things very quickly.
 
Couldn't agree more, Gary.

The answer is 5-axis. On most trunnion style 5-axis verticals, the spindle moves in X-Y-Z.

And multiaxis is itself a form of automation by eliminating extra setup labor.

Two birds with one stone.

When we bought our Haas UMC-500s, we spec'd them with thru-spindle air and extra M-codes with relays, for this exact application in mind - eliminating the redundancy of having a standalone robot - by self loading with the spindle using a CAT40 based gripper. The M-codes can activate retractable trays that come through the side. Coming "soon"...
 
It's hard to beat the flexibility of a 6 axis robot. While I have a fixed table mill just sitting, I haven't turned it on in 7 months since its dedicated product line died from Covid, I am going to try the robot route, I think.
 
Depending on the part and robot, exchange times for a 6-axis robot can be way quicker than even dedicated pick n place automation. I have machines running that exchange parts via robot with a "machine out of cycle" time of 1.13 seconds. Most are in the 2.65 - 3.3 second range.
 
This whole thread arises from trying to apply automation to "standard" machines (because that's what's at hand.) But of course, if you have a pallet changer machine (could be a vertical, though most are horizontal) it wouldnt' be so difficult to have some kind of "over the top" belt thing that loaded/unloaded it. Then again you can buy standard robots at "industrial commodity" prices to do that same work.

Oh, and I saw an Okuma VMC using a coolant powered schunk gripper to load parts from a rack on the table into a collet (and out) most of a decade ago.... So various "not a 6 axis robot" schemes CAN work with verticals...
 
Which if you really think about it is fast. Many machines can not even open and close the doors in this amount of time.
From the program end to the next cycle start what is a bar feeder?
Bob

I'd estimate my bar feeders at about 12 to 15 seconds to exchange a 12 foot bar. Amortize that over the 700ish parts per bar, it's not bad. Granted, not all my parts are sub .200".
 
It occurs to me that we are trying to automate machining by using robots to do things the way a human would. VMCs have a moving table which works well with a human loading. However if you want simple low cost automation we really need machines with a stationary table. If the table doesn’t move you can use the spindle with a gripper or an auxiliary gripper on the side or front of the spindle to pick a part from tray or conveyor and place the finished part in a second tray or conveyor.

There have been a few examples of standard VMC's with spindle mounted grippers loading/unloading onto a table mounted tray. (Yamazen Brother Spindle Gripper for unattended operation - YouTube). It seems like a great way to automate mid volume applications or ones where load/unload is a small percent of the cycle time.

It would also be relatively straight forward to have a part load/unload tray or conveyor come in from the side of the machine as well. Bring the unloading apparatus to the spindle instead of the spindle to the unloading device. I don't see much benefit to adding to the overall axis travel of a machine for the sole purpose of getting a part off the table. Machine tool axis need to be stiff and are consequently expensive to make larger. The longer axis also has a negative impact on machine dynamics.

Spindle grippers have a somewhat narrow window where they make sense. As your spindle up time becomes more critical, it makes sense to move load/unload to a separate system. As Tonytn36 points out, this can be done in ways that are VERY fast. A spindle gripper also is more difficult to implement in process QC that can be accomplished with sensors on the end of an arm. We definitely see a lot of suboptimal robotic arm implementations (like using the arm to open the damn door!).

More linear, gantry like arms can make sense for some applications, but by the time you end up making them adaptable, you end up with 4-6 axis and might as well go with an arm.
 
I see flexibility as probably the largest influence on which automation solution makes the most sense. If you are only ever going to run 1 part in very high volume applications you can design dedicated machines with custom tooling. Families of parts can be run on similar equipment with some planning in the beginning of design, but the price and complexity goes up.

The 6 axis robots are really just the most fine tuned option out there at the moment that can accomplish most tasks. The fanucs we have require virtually no maintenance compared to how much they run, they are accurate, relatively cheap in the world of CNC machines, and most importantly they are flexible. You don't have to build in adjustments to each step of your process when the robot can just be programmed to move over another 0.050".

I work on gantry loaded lathes daily and they are amazing and relatively quick, but they are pretty finicky to setup. probably easier when you are just loading slugs instead of complex castings. They work for us though since we run millions of 1 part per year.
 
I was looking at an old VMC the other day, and got wondering...
What if I took the spindle off, and mounted a 4th axis with a collet closer in it..
The material bars could poke up into the attic.
I could cover the table with simple spindles..
A gang tooled, bar feed 4 axis production machine...
Maybe some multi tool blocks from a swiss machine.
I have made Many, many parts that would be Perfect! for that setup.

Always wanted to bar feed a machining center, gang tooling it would be another notch.

Back to reality.
 
I'd estimate my bar feeders at about 12 to 15 seconds to exchange a 12 foot bar. Amortize that over the 700ish parts per bar, it's not bad. Granted, not all my parts are sub .200".


12 seconds total out of cycle for 700 parts ? .017 seconds per part? Now that is flying.
12 seconds per part out of cycle is different and would be like watching paint dry so I'm confused.
Bob
 
12 seconds total out of cycle for 700 parts ? .017 seconds per part? Now that is flying.
12 seconds per part out of cycle is different and would be like watching paint dry so I'm confused.
Bob

I think that's to pull out the leftover stub, and put in a new bar.

My friends swiss shop has these feeders, the clamshell opens, new bar get's flopped in, clamshell closes.
 
12 seconds total out of cycle for 700 parts ? .017 seconds per part? Now that is flying.
12 seconds per part out of cycle is different and would be like watching paint dry so I'm confused.
Bob

12 seconds total for 700 parts.

Load bar, run 700pcs, 12 second bar change, run another 700pcs.

That's an extreme example, for making .200" long parts out of a 12ft bar. Lots of parts are in the 3/4" to 1.5" range.
 
12 seconds total for 700 parts.
Load bar, run 700pcs, 12 second bar change, run another 700pcs.
.

If you time it that way mine is zero seconds as I can put in a new magazine full of blanks while the machine grinding. I'm sure Tony is this way also.
To me the part change is non-run time which averages 3-7 seconds for auto load and 8-12 seconds for hand load.
A bigger time killer on the autos is doing in machine part measuring time which can eat as much as 30 seconds once every 3-6 parts.
Hand load machines measured while the machine in cycle on the next part so no time lost but a size adjust if needed will eat 5-10 seconds.
Bob
 
To be fair, you need to add in the unchuck-move-rechuck part of the cycle into the per-part time too, though.

That's all accounted for That's pulling the bar out of the guide bushing, opening the main spindle, turning control over to the barfeed, retracting the remnant, removing the remnant from the collet, loading the new bar, pushing it forward, retracting the pusher, dropping the pusher into the channels, loading the bar in the pusher collet, feeding the bar to the insert position, turning control back over to the machine, pushing the bar back into the guide bushing, doing the top cut, and ending the bar feed program before jumping back into the main program.

I was curious, so I looked it up on youtube. This is a barfeeder like mine. They claim 35 seconds total. I think that's a little long, personally, but I've never timed mine, I was just going off of how impatient I get when I'm manually loading a bar (my attention span is very short when I'm not actively doing something at the control)

https://www.youtube.com/watch?v=tuSyK3Dl5a8
 
That's all accounted for That's pulling the bar out of the guide bushing, opening the main spindle, turning control over to the barfeed, retracting the remnant, removing the remnant from the collet, loading the new bar, pushing it forward, retracting the pusher, dropping the pusher into the channels, loading the bar in the pusher collet, feeding the bar to the insert position, turning control back over to the machine, pushing the bar back into the guide bushing, doing the top cut, and ending the bar feed program before jumping back into the main program.

I was curious, so I looked it up on youtube. This is a barfeeder like mine. They claim 35 seconds total. I think that's a little long, personally, but I've never timed mine, I was just going off of how impatient I get when I'm manually loading a bar (my attention span is very short when I'm not actively doing something at the control)

https://www.youtube.com/watch?v=tuSyK3Dl5a8

I am pretty sure what Emanual was meaning is that you have to add the chuck open / push / chuck closed to each part where the bar advances in addition to what the bar change time is. That is the true dead time. If you are pushing every time for that 0.2 long part, then that chuck open / push / chuck closed time has to be added to the initial bar load time. (And each part cycle time to get actual takt time.)

So 700 0.2 parts but you push every 5 th part and the push takes 3 seconds then you need to add 420 seconds to the 35 second bar load time. It costs 1.54 seconds per part for exchange.
 








 
Back
Top