Robot Maintenance & Replacement in Manufacturing Plants

[StatGuy2000]

Hi everyone. I wasn't sure where to include this thread, so thought I'd include this under General Engineering (mods: please feel free to move this thread to a more appropriate subforum).

One thing that came into my mind was with respect to robots used in manufacturing (e.g. auto assembly plants). These robots, if I'm not mistaken, are run for multiple hours as part of the manufacturing process.

As with any piece of machinery, robots are subject to "wear and tear" through repeated physical motions over a period of time. Which would mean that these robots would need to be fixed, upgraded or replaced after a certain number of years of operation.

I have the following questions:

1. What do engineering staff (or others) do to determine whether a robot needs replacement, apart from discovering any major defects?

2. What processes do companies or organizations do to replace or upgrade robots, without causing any major disruptions on the manufacturing process? How do they, for example, schedule down times, and then test any new robots to ensure they are working as according to specs? I can only imagine how costly it would be to discover that a new robot may itself have issues.

Perhaps these questions may be silly, but I appreciate any feedback that any of you could provide on these matters.

 

[Vanadium 50]

What is a robot? Is a CNC drill a robot? The thing that fails most in a CNC drill is the drill bit, and that's replaced when it wears just like a bit for a hand drill.

It's a machine like any other: you have a PM schedule, nut every so often something breaks outside of that, so you replace it when that happens. It's not unlike, say, a car.

 

[Averagesupernova]

I would say before most 'robots' are put into service it is already known when the equipment will need service.

 

[berkeman]

1. What do engineering staff (or others) do to determine whether a robot needs replacement, apart from discovering any major defects?

In addition to regularly-scheduled periodic maintenance (PM), and repairing any obviously failed parts, there is also an effort under way in some parts of the industry to be able to try to detect the onset of a failure before it happens. For some electric motor technologies, you can monitor the current and voltage characteristics (especially at startup and under heavy loads), and watch for the early signs of failure. This can be reported to the humans who work with the machines, and they can decide when they want to take the machine down for maintenance.

Here is one reference for such monitoring. The company that I work for has also looked at using this technique to alarm early failure indications in electrically-powered machines and appliances...

https://www.911metallurgist.com/predicting-failure-electric-motors/

 

[essenmein]

I have the following questions:

1. What do engineering staff (or others) do to determine whether a robot needs replacement, apart from discovering any major defects?

2. What processes do companies or organizations do to replace or upgrade robots, without causing any major disruptions on the manufacturing process? How do they, for example, schedule down times, and then test any new robots to ensure they are working as according to specs? I can only imagine how costly it would be to discover that a new robot may itself have issues.

Perhaps these questions may be silly, but I appreciate any feedback that any of you could provide on these matters.

I can only speak to my experience (high volume electronics manufacturing), our lines have a very high level of automation, not sure if this fits the "robot" category, is a pick and place machine a robot? What about a screen printer?

1. What do engineering staff (or others) do to determine whether a robot needs replacement, apart from discovering any major defects?

There are several different aspects to this.
- tool wear, most cutting/mold tooling etc have a finite life, as the tool wears tolerances start to drift, finish suffers etc. Generally they are just replaced at a fixed interval. If they for some reason wear faster than expected, this is (hopefully) caught by AOI or periodic CMM checks. For us its 100% AOI, and we do some periodic (say every 1k units) one gets sent off to CMM to make sure its dimensionally correct. So say a transfer mold press, the tool lasts about 100k shots, and needs constant replacing, but the press itself would just undergo regular maintenance (eg hydraulic oil changes, cleaning etc).
- equipment wear, eg smt machine starting to miss place components, maybe due to linear bearing wear. This would be detected by the machine itself since it is self checking. Or subsequent AOI would detect it. Worst case you start seeing an uptick in board failures during ICT.
- outright failure would bring the line to a halt and then the pressure is on to fix ASAFP.

2. What processes do companies or organizations do to replace or upgrade robots, without causing any major disruptions on the manufacturing process? How do they, for example, schedule down times, and then test any new robots to ensure they are working as according to specs? I can only imagine how costly it would be to discover that a new robot may itself have issues.

Generally people that plan these things know that you need to factor in down time. If the plant is not running full capacity (ie 3 shifts a day, 7 days a week), then the period where the line is not running would be used for maintenance. If it is at capacity during the week, then weekends for major work. In our case, once a line is validated there is no "changing equipment" to upgrade etc. Once its running thou shalt not do anything other than maintenance, or the customer will make you re validate, PPAP is not fun. Part of the initial step is to validate the line and make sure its all working together before series production is even started.

A note on down time, part shortages are usually the largest driver of line stoppages, and our customers charge us handsomely if we happen to hold their line up by us not delivering required number of parts on time. So a lot of planning goes into making sure we do NOT do that!

 

[DEvens]

The specifics will depend very strongly on the details of the machine, it's operation, the conditions, the potential failure modes, and the potential harm of failure. However, the principles at work are fairly common across many engineering systems.

Consider just one joint as an example. This joint allows one part to move relative to another. Say it's an "elbow" joint similar to the one in this picture.

From a variety of sources, the potential for degradation will be estimated. This will involve such things as observing similar joints under similar conditions and for similar mission times. This will guide the designers and maintainers on what to look for.

Then there will be tests. For example, the almost-vertical portion might be locked in place, and the almost-horizontal portion given a calibrated force at a calibrated angle. The "play" in the joint would then be measured. Meaning, you push this hard at the wrist and the arm moves that much. There will be specified acceptable limits for this play. There will be specialized equipment specifically meant to measure it. And there will be regular schedules to come along and test it, and record the result.

By the way, this play will have an upper and lower limit. If the play is too small it may mean the joint is too tight, or full of debris, or some such. Too large may indicate wear.

Each portion and function of the device will have similar tests defined and acceptance limits specified. In some cases, the device may have a test program included in its software, and the measurement process might be automatic. Maybe once a week (or whatever is appropriate depending on how hard the device is working) it will stop and do a self-check. It would lock all joints but one and push that joint against a provided stable object. Then it would report what its sensors said about the motion of that joint. If the values were out of range it would raise an alarm. Otherwise it just records the value, then goes back to work.

So if there is some kind of gripper, for example, the strength of the grip would get tested. If it's supposed to get from here to there in some time, that time is measured. If it's supposed to use a certain amount of electric power during a specific operation, that would be measured. If it's location sensors are supposed to be accurate to some particular level, that accuracy is tested. And so on.

Then the device will be monitored during the expected operation time between replacement. If it is degrading faster than expected it may get scheduled for diagnostic tests outside normal. Maybe it gets pulled "out of the line" and a replacement put in. This unexpected failure is then carefully examined to try to determine why it happened and if it can be prevented.

At a higher level, there will be periodic reviews. The maintenance schedules and policies will be reviewed against data regarding downtime and service life. Modifications will be tried. Maybe servicing every 9 days instead of every 10 will mean less down time. Or maybe servicing every 12 days instead of every 8 will make no difference to down time, but will save 50 percent on service costs. Or maybe the wear surfaces in these joints need to be replaced every 3 days while in those joints it only needs to be done every 20 days. And so on. The goal is overall optimization.

These reviews will also consider things like changing the material for replacement parts. Or changing the duty cycle. Or changing the lubrication material or schedule. Or improving the "clean room" standards. Or various other things.

Such maintenance changes might be tested on a population of machines. Maybe out of 20 robots, there will be four groups of 5 each. And each group gets a different schedule of maintenance and testing. Then the group that shows the best overall performance will be the candidate for how maintenance is performed int he future.

 

[trainman2001]

What seems to make the difference between an industrial robot and a single purpose automated machine is simply that the robot is reprogrammable to do a task differently by "training it".

A single purpose machine, although automated, would require a physical rebuild to do a different task. For example: a bottling line. There are all kinds of automated devices on such a system such as actuators, sensors, limit switches, motors, drive systems, etc. To run a different sized bottle on the line would require a lot of mechanical changes, but it obviously is done.

The industrial robot would be taken through a new set of movements to accommodate the new product. That said, both instances, in a modern plant, would be under some form of statistical quality control that would measure all kinds of key parameters. In either case, when drift would be noted, a root cause analysis would be performed to determine if it was the machine/robot that was changing or some other environmental or feedstock related problem and the fix would be made accordingly.

I'm sure, since industrial robots are machines in their own right, they would be under a preventive maintenance schedule too. In large scale process industries (autos included) there are scheduled turnarounds where the plant is shut down and major maintenance is performed. Robots would fall under this schedule also. Since robots are easily trainable, those in critical positions on the line would have spares in waiting that could quickly be programmed to do the job. It's one of the advantages of having them.

 

[John Archer]

Generally, we just keep track of the hours on the servo motors, and grease the joints based on that.
We change out wiring harnesses every 18 months as we know they won't last more than 2 years in our application. (Larger robots, with a fair amount of movement).

You can see the hours on the robot, and schedule a time to do maintenance during what would be "normal" down time (i.e. on a Sunday for us).

Other issues will often give you advanced notice before failure. Like a leaking gear box, or one with noticeable play in it. They will keep running for a while, but you want to schedule time to replace ASAP.
A dead motor is harder to predict, but with a little forethought, they can be changed in a relatively short amount of time. Yes, this is unexpected downtime, but knowing how to get it down quickly and efficiently, and how to re-master the robot properly are key.

This is something our maintenance techs go to school for, instead of calling theobot mfg, and having them do it for us.

 

[Windadct]

I worked for ABB and worked on their S2,S3 and S4 level systems.

Mechanically they are insanely reliable - from a pure repetitive motion standpoint. They have to be - in automotive lines, a down robot could idle a complete production line 250 robots ~ 250 workers = no cars... very bad...

In these cases the controller is effectively running a kinematic model of the robot and the controller is constantly comparing the position and motor current feedback with the model, an actual emulator. Any deviation or force ( motor torque aka motor current) that is outside of the kinematic model causes a fault. Primarily this is set up for collision detection - and can be very effective. If a bearing, for example, is failing the robot will detect excessive load and begin to have collision faults.

Most failures were more due to poor application ( an early lifetime failure) or operator error, severe collisions - or improperly conducted maintenance ( most major gearboxes are lifetime sealed now). Environment / maintenance is the top issue in my book... for example, not replacing filters ( latest designs have external - unfiltered heatsinks, and sealed electronics). Humans are the problem...

Oh - for your reading pleasure...