2
$\begingroup$

I am a researcher on robotics and AI, and wonder if traditional manufacturing machines, such as lathe, CNC, or grind use any kind of machine learning technique or algorithm. This algos are widely used in mobile robots.

I.e.: in a plant where there are some lathes, do they interact somehow between them? Do they adjust on the fly their parameters when machining a new piece? Or do lathes try to adjust to the parameters to the lathe that is performing the best? Kind of collaborative learning?

I did not find this kind of examples in industry yet, but I wonder if industry 4.0 is about implementing this kind of techniques.

$\endgroup$
5
  • 1
    $\begingroup$ Not really. You could stretch a point to say that end-stops 'teach' the machine how far to move, or strain gauges 'teach' the machine to reduce the feed rate, but that's a long way from "conscious" learning. I'll also point out that any attempt to produce an exact definition of "AI" leads to massive flame wars. $\endgroup$ Dec 1, 2016 at 15:06
  • $\begingroup$ I'm not sure what they could usefully learn. Generally the part is designed to certain specifications. Any adjustments would already be programmed in by an operator. I can't see any good use for on the fly changes to the operation based off previous experience; at least in the way manufacturing machines are currently used. $\endgroup$
    – JMac
    Dec 1, 2016 at 15:35
  • $\begingroup$ Manufacturing is about producing many identical or at least well defined items, don't know where the problem is AI solves better than conventional control systems. There is/was research in using expert systems to operate biogas plants, maybe same for other process plants. $\endgroup$
    – mart
    Dec 1, 2016 at 15:47
  • $\begingroup$ OK, thanks for the replies. I thought that machine learning could help mostly in companies were many different requests are dealt. I mean, instead of having a trained operator to adjust the lathe, the machine could adjust itself. What about CNC self programming? $\endgroup$
    – galtor
    Dec 1, 2016 at 17:48
  • $\begingroup$ You don't necessarily need someone to manually adjust the lathe, it can be programmed into the system. The machine can't program itself, that's usually where operators come in for the high-tech automated machining processes. It's still not AI though, because it is being told exactly what to do. I'm not sure if we'd want manufacturing machines deciding what to do on their own. People are generally much better at interpreting requests from others than an AI is. $\endgroup$
    – JMac
    Dec 1, 2016 at 19:20

1 Answer 1

3
$\begingroup$

In general terms what you really want from CNC machines is to repeat the same actions exactly the same every time, while you may need to monitor and account for things like tool wear, temperature changes, small variations in material properties etc this is still in the realms of fairly straightforward feedback control.

Ultimately lathes aren't the best example as they really only only really have 3 degrees of freedom, cutting speed, tool feed and tool travel. A better example would be 5 or 6 axis milling which is inherently a lot more complex. Here there may well be scope for AI applications, particularly in designing and developing tool path design. This is the process of specifying the sequence of cutting moves to machine a part as quickly and efficiently as possible to a specific quality standard and may require numerous tool changes as well as avoiding collisions, nesting parts, deciding on which order to perform operations, optimising part/billet setup and orientation.

In particular there is a lot to be gained in the paths used for roughing cuts and finding the optimum tool path to move material quickly and effectively for example by taking a smoothed path around sharp corners rather than just using a fixed offset from the final geometry.

Automated tool path design has been around for a while but it is certainly conceivable that AI could have a role to play and it is certainly an important field.

$\endgroup$
2
  • $\begingroup$ Thanks. Do you know of any example for tool path calculus with this technique? Are big enterprises already implementing it? $\endgroup$
    – galtor
    Dec 2, 2016 at 7:14
  • 1
    $\begingroup$ Autodesk Powermill autodesk.com/products/powermill/overview is probably a good place to start. $\endgroup$ Dec 2, 2016 at 15:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.