0
$\begingroup$

New manufacturing facility of telescopic conveyors has been established one Eastern country of European Union. The managers of the facility are very proud that human workforce is their main asset - all is being done by human hands - they are specifically stating that the following jobs are being done by the human hands: welding, painting, assembling. They are stating also that their facility is the most modern one in the world.

It is quite interesting. Welding, painting, assembling are tasks that are already automated in the car manufacturing and are routinely done by the industrial robots. Why such tasks can not be done for the manufacturing of the telescopic conveyors?

One explanation is that car manufacturing is quite defined and immutable process - robots are programmed for the specific model and few specific models. But the conveyors are adapted for each customer separately. But again - today we have so much work about reconfigurable and adaptable manufacturing. Can we establish software defined pipeline which starts with the CAD files for the conveyors designed together with the customer. Then automatic transformation process generates the specific configuration and plans for the industrial robots for this kind of model, then automatic tests and calibration procedures can be run and at the last stage the plans can be uploaded to the robots and the manufacturing executed.

My question is - what prevents the use of industrial robots in the manufacturing of telescopic conveyors and whether the reconfigurable adaptable manufacturing can be used for the automatic reconfiguration of industrial robots for one-off production of the conveyor from the CAD files?

Is this is the process in the future then how near is such future and what are the main challenges for implementation of such process?

p.s. I am not the expert in robotics in manufacturing, but my efforts are around the machine learning, symbolic computation and automated theorem proving and these are the areas which can be used for the processing of the CAD files and for the generation of the production commands from some kind of meta-rules, meta-knowledge. Though I have not tried to implement that, but may be this is the perspective research and commerce area?

$\endgroup$
1
  • $\begingroup$ you are making an assumption that something other than human choice is responsible ... contact the manufacturer and ask why they are not using robots $\endgroup$
    – jsotola
    Oct 24 '20 at 2:41
2
$\begingroup$

So much to unpack. Let's make sure we understand each other.

You said " today we have so much work about reconfigurable and adaptable manufacturing". My understanding is that you are talking about something like a baxter robot. it is easy to place the robot in a new location, and to teach it a new task. Basically a robot that is easily "trained" for a new task. Instead of a programmer coming in and writing code, a non-technical person can (for example) press a button and guide the arms, and the robot will repeat the motion. One thing to remember is that this is a tool with very limited capability.

you said "Can we establish software defined pipeline which starts with the CAD files for the conveyors designed together with the customer. Then automatic transformation process generates the specific configuration and plans for the industrial robots for this kind of model, then automatic tests and calibration procedures can be run and at the last stage the plans can be uploaded to the robots and the manufacturing executed." OK, there is one core idea here that stands in the way: cost. Regardless of the kind of robot you choose, you are talking about an automated system. With enough money, you CAN create what you suggest, it just will cost an insane amount of money to execute. If you were to take that route, you would have to define all the work the robots need to do (screw every nut, put every bolt in place, attach each component, align sensors, install the belt, lace the belt, etc...), then some company would provide the robot hardware, and program it for your application, and give you an interface to the system (ex: build model A, or model B). This would make no business sense due to cost.

Lets do your question piece by piece.

" what prevents the use of industrial robots in the manufacturing of telescopic conveyors?" in short, it's cost:

  • Telescoping conveyors are low volume production - anytime we are dealing with low volume production, it doesn't make sense to automate. There is no business case for it.
  • Manufacturing requires high dexterity & mobility - you need to go under the conveyor, and say pull a cable from here to there. You need to install a cable reel under the unit, you need to install bearings, you need to swing one part of the boom into the other, etc... There is so much "variety" of work, that the cost of automation will be crazy expensive and won't make sense.

"can reconfigurable adaptable manufacturing can be used for the automatic reconfiguration of industrial robots for one-off production of the conveyor from the CAD files?" If you have unlimited money, yes. In real practice no. The real key to automation in the real world is having a robot do a task uninterrupted. Say you need to screw a nut. A $40,000 robot that screws one nut, means you paid 40,000 to get that nut screwed. If it screws 1,000,000,000 nuts, then the price per nut is very cheap, and automation makes sense. The issue is that a robot that can screw a nut, cannot simply "change it's job" as it has to be retooled. it needs to exchange it's nut screwing hardware for some other task, and this switch is mechanically complicated, and complicated from a software perspective. Take for example the steps to building an electrical panel for an extending conveyor:

  • pick parts
  • place the panel on the work bench
  • open the door (you need a key)
  • pull out the backplate hardware (will be floating around)
  • pull out the backplate
  • lay out the parts
  • cut the din rail
  • put the terminal block on the din rail
  • cut cable
  • strip cable
  • screw cable into terminal

Just here you can see how much variety of tooling a robot would need to have to build a simple control panel. One robot cannot do it all; so you need many robots. Well how many panels you building per year? 20? not enough to justify automation.

"Adaptable automation" is not very practical. In the future as more software is available for "human" like hardware (ex: hands that can screw a nut, AND pick up tools to tighten it), this will become more of a reality, but at present it's not economically feasible.

"Is this is the process in the future" I think so. When robot hardware is on par with human capability, and the software enables the hardware to do things that we people find "easy" like holding and using tools then you will see robots doing more skilled work.

"how near is such future?" I believe about 40 years, or one giant war away.

"and what are the main challenges for implementation of such process?" the software. you can put a robotiq hand on Boston Dynamic's Atlas, but do you think you will be able to code it to use a screw driver like a human? It comes down to software.

If you can get robots to learn, then the hardware is already there. Really expensive at present, but nothing mass production can't fix. The real issue is software.

The reason why doing one-off automation is not done in the real world is because a robot with specific tooling is needed for every one task. and when you break down building a telescoping conveyor there are thousands of tasks, and the cost of hardware to do those tasks doesn't make sense when you have the option to hire people (where one electrician can build a panel on their own).

$\endgroup$

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.