Fully Automated Manufacturing

[JakeL]

I am curious to get others opinion on this. How long will it be before job shops are fully automated? I mean like automatic quoting, programming, machining, shipping, and all the handling. It seems like a complete daydream but it also seems like all the necessary tools to make this a reality are currently available.

We already have the CNC machine which runs automatically. Some companies claim to have automatic quoting. I hear rumors of some companies with automatic programming (only with part families). These robot arms seem to be getting cheaper and more popular as well. It just seems like a matter of time before someone figures out how to put all of these components together into a single company.

I understand this is an obscene simplification of "fully automated manufacturing" but I'd love to hear everyone's thoughts on this.

[JParis]

Even those places with "fully automatic" quoting and machining have limitations.

There are features they can't/won't produce....we deal with this alot on prototypes...

There is also  a hang up on more complicated parts, the system cannot handle them and a real person must intervene...

The technology is still propriatary I believe, so that limits it..

I know I am starting to beat the Cobot drum here, with robot fed machining lines, there is no reason we can't get to robot fed loading...

Decades unless some breakthrough makes it easier is my guess

[neurosis]

programming,

I know that this gets joked about a lot. You hear a lot of the "so easy a 5 year old can do it" thrown around from time to time. 

When I saw how my dentist makes crowns it was a real eye opener. 

They 3d scan the tooth and import it in to software that automatically programs what I assume is some sort of 5axis router type machine. 

When they're done the tooth always fits perfectly and the entire process only takes a couple of hours so you get your crown complete in one visit. 

[JParis]

a tooth isn't an Aerospace part....

Your point is well made though, a lot of stuff can and is being done right now.

[neurosis]

a tooth isn't an Aerospace part....

Your point is well made though, a lot of stuff can and is being done right now.

I don't know if we'll ever see anything like this in our field in my lifetime but it certainly appears that the technology is there.  I'm talking straight out of my bunghole of course as I don't know if they could ever apply this to the types of machines that we run. 

[neurosis]

However, it relied on the dentist graphically on screen adding some clearance etc....ie experience.

That was part of the design process.  The scanner scans your teeth individually, then they scan your bite.  The software generates a tooth in the place that the crown goes and then the dentist can massage areas that look like there will be interference.    When my dentist got one of these in, the first crown I got they let me watch the entire process start to finish.  It was pretty cool.

[gcode]

They 3d scan the tooth and import it in to software that automatically programs what I assume is some sort of 5axis router type machine. 

Yes.. my dentist used to use a system like this
I don't know what he used to program it, but ModuleWorks has a dental module
The machine itself was about the size of a small microwave oven and from scan to finished crown took about an hour
He doesn't do crowns this was anymore. They have new materials for dental crowns now that are much tougher and more durable,,, too tough to machine...
He buys them cast exactly to shape which takes 2 weeks

[JakeL]

But you could always buy different machines....so ultimately, as always, it comes down to the parts.
The issue comes with things like distortion - millturn in one and as soon as you release chuck pressure the part goes out of tolerance.... or a 2 tenths bore that's tucked away deep inside a blind narrow pocket on a sidewall....
A gauge block looks easy to make - it's only a rectangle.... experience is the obvious thing that these aids lack IMHO

I'm going to play devils advocate and go off the deep end... Couldn't you avoid this if every part of the manufacturing process could talk to one another?

For example, a part that springs after you remove it from the machine. If an inspection machine automatically checks the part and sees distortion, then it could communicate the issue back to the programming software where the part could be reprogrammed and sent through the whole system again. And this process could just repeat until the software produces a way to make a good part.

A lot of the guys I work with are in their 50's or 60's and I pitched the whole "automated manufacturing" idea to some of them. One of the responses I got was "Have you ever seen Terminator, or IRobot?" I think this was their way of telling me my dreams may be a little too big (or something like that)

[YoDoug]

Many of the technologies to go full iRobot automation exist, but some are still lacking. Computing power is one. For the machine learning and AI needed to accomplish this, currently it would have to be done in centralized processing. The shear amount of data transfer would be difficult. 5G could possibly speed things up but there would surely still be bottle necks in the process. There is also the speed/latency of sensor input. For an iRobot humanoid robot to rival a humans dexterity and sensory capabilities there are still a lot of limitations. Boston Dynamics dancing robots are fun to watch but they still have balls for hands because there is a limit to the sensory input and processing needed to mimic human hands.

Now as far as CNC machining/programming. I think current machine learning and AI are more than capable of programming even the most complex parts. A machine learning computer given access to the entire body of documented machining knowledge could learn pretty quickly how machine complex parts. The AI and processing power could even learn to calculate things that make machining a skilled trade that takes years to learn, like tool deflection, part warp, material properties, tool wear, etc. A skilled machinist knows how to keep the floor or walls of a thin aluminum aerospace part flat and straight from experience. The AI could calculate this in seconds given the right time to "learn" aerospace machining. However, until computers and robots can design and build other computers and robots, I don't think there would ever be a cost justification to spend the time and resources to teach AI to program, it is a relatively easy task for humans to learn and there is so many higher needs for AI.

[Here's Johnny!]

AI works 24/7 crunching data/ learning. Eventually it will take over a lot of things, possibly within 20 years.

[Matthew Hajicek]

If you want to see the state of the art in automated job shopping, look at Protolabs.  They have their own proprietary software, standardized procedures, etc. to cut every possible bit of human labor out of the process.  They have certain limitations, though they keep improving over the years.

The danger I see is that once all the easy to medium work is fully automated, you stop teaching humans how to do it.  Then you don't have any humans capable of figuring out the more complicated jobs that the automation can't yet do, and you're stuck.

[CNCAppsJames]

How long until we're there? I'm not certain that I will see it as you describe @JakeL but my son who is 18 will moat definitely at least the lino's share of it. 

What is truly going to be the Achilles heel is the programing of problem materials that for one reason or another won't be able to be additive manufactured. Make no mistake, those parts and material will ALWAYS exist. Their numbers will continue to decrease but they will always exist.

Additive will have a bigger say than all these other things. It will get faster. It will get more accurate. More materials will become available.The software will get smarter about thermal issues with materials. 

JM2CFWIW 

[JParis]

How long until we're there? I'm not certain that I will see it as you describe @JakeL but my son who is 18 will moat definitely at least the lino's share of it.

What is truly going to be the Achilles heel is the programing of problem materials that for one reason or another won't be able to be additive manufactured. Make no mistake, those parts and material will ALWAYS exist. Their numbers will continue to decrease but they will always exist.

Additive will have a bigger say than all these other things. It will get faster. It will get more accurate. More materials will become available.The software will get smarter about thermal issues with materials.

JM2CFWIW

The machine we have, we are already 3D printing titanium....this machine can also print inconel 718 and other things as well.

The current issue is printer resolution, currently it is difficult with those "high stress" and pressure body type items...but IT IS getting better every year.

The big question that's still open is "how good can it get?"

[gcode]

.this machine can also print inconel 718

We are currently doing some work on a 3D printed inconel part

It is a 1/4 scale rocket engine nozzle and it took our customer 7 months running 24/7 to print it

[JParis]

We are currently doing some work on a 3D printed inconel part

It is a 1/4 scale rocket engine nozzle and it took our customer 7 months running 24/7 to print it

Dayum....

We have a 400mm x 400mm x 400mm print volume....decent size but not that large...wow