Manual vs automated fault detection by vibration

[AX3L]

Hi everyone!

As a newly graduated control system and automation engineer I got a bit of a shock when I discovered that the automatic fault detection at the company I work at is non existing. Sure, they measure vibrations once in a while manually to detect bearing fault and imbalances but considering how cheap and powerful microchips and PLCs are today I can't see why they don't do it automatically all the time. And since I've not gotten any good answers I'm curious about how it looks at the industry today. Is automated fault detection just something they teach in school but that's to expensive to implement in real life today or is just my company 10 years behind? It's a large truck manufacturer by the way, so there are a lot of turning, milling and drilling going on and it's not a small 10 machine business.

Some of your inputs and thoughts about this subject would be deeply appreciated. Thanks in advance!

Axel

 

[IRstuff]

detection of what kind of faults?

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[AX3L]

@IRstuff: Faults that can be detected using vibration analysis. That is bearing wear and faults of different kinds, imbalance in the chuck or another rotating part, problems with the rotor in the engine, worn out cutting tool and so on. Faults that not only lead to more damage in the machine but also imperfections in the workpiece.

 

[GregLocock]

In 35 years of using lathes and milling machines I've never had one of those faults come up. Perhaps I've been lucky, or perhaps the cost benefit analysis is that it isn't worth retrofitting machines with full time instrumentation.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[AX3L]

GregLocock: Thanks for your answer! That's interesting. But I guess that if you do all the maintenance in time just as the manufacturer tells you to do and you don't run your machines close to full speed you can get along just fine without that kind of equipment (apparently). And this is how my business also does the maintenance at the moment. The thing is though, if you monitor the actual health condition and need for service you can save a lot of money not only of spare parts and work but more importantly reduced downtime (try change a bearing while run the machine if you can). But as you say, maybe the investment and cost to run such a system is so high it doesn't really pay of in real life?

 

[MikeHalloran]

As was said, cost:benefit.

Further to that, when attended machines, or machines that are near attendants, begin to fail, their sound signature changes. That's basically what you can detect with accelerometers and such, but it happens that people can detect sounds too, and will report a change in funkiness of any given machine. ... if they are allowed to, by local custom and contract.

Remote/unattended monitoring might make sense for unattended machinery, or for attended machinery in a politically hostile environment. Some union shops are like that, with a constant subliminal war being conducted by people who are ill-equipped for head games.

Unattended factories have other means of detecting mechanical failure early, by means of examining the data already being collected for quality assurance purposes.

Which is not to say that you'll never make a dime installing machinery monitoring equipment, just that it's a harder sell than educators might suggest.



Mike Halloran
Pembroke Pines, FL, USA

 

[AX3L]

Mike, thanks a lot for your great answer!

You have a good point about the machine operator being able to detect many faults just like you do on your car. But at least according to my experiences so far that is more detect when things already have gone wrong for real. And the whole point is to detect the small changes in vibrations that indicates a later failure before it happen, so that the bearing fault of one bearing doesn't take others with it because of more vibrations and so on. Sure, replacing the bearings with short intervals will decrease the risk of a large failure but my idea is to cut the maintenance costs and more importantly unexpected downtime by as much as possible.

The picture you're giving me is about the same I've gotten from the more experienced coworker at my job so I'm starting to guess that still is valid. I did check with a manufacturer of automatic fault detection equipment today though and the price per unit he gave me right away for a simple system that you connect to the PLC of the machine was a bit short of $1000 (per accelerometer channel) and that seems like no money at all to me. But the question remains, haven't the price of this technology dropped so much in the last few years (to where it is now) so that's a reasonable investment or is it still only something to consider for very few branches like the paper industry that have used it for a couple of years already?

 

[IRstuff]

You can try, but realistically, it can take you years to accumulate sufficient correlation between any vibration and an actual failure mode. For instance, say your sensor detects a 0.01" wobble in the chuck; is that OK? How do you know without a truckload of data that says, "On average, a 0.01" wobble is indicative of a failure within 2 weeks of operation?"

From what I've read in the prognostics literature, controlled experiments often result in a 50% spread, which is too large to be actionable.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[MikeHalloran]

>>>... price per unit he gave me right away for a simple system that you connect to the PLC of the machine was a bit short of $1000 (per accelerometer channel) and that seems like no money at all to me. <<<

I'm easy to find. Send me $1000.
Thanks.


Mike Halloran
Pembroke Pines, FL, USA

 

[AX3L]

IRstuff: I'm not sure if I get what you mean but that's usually not how vibration analysis is made. What you do is analyse the different frequencies with a fourier transform (or maybe wavelet transform, but that's a higher course). Any imbalance in the chuck etc will have a frequency as a multiple of the RPM, bearing fault will have a certain (much higher) frequency and so on. You can calculate what frequencies you expect, and that's how you do a more advanced fault diagnos. However, for a continuous measuring system one usually only detect if any frequency is different than usual. You can read more about this in the FAQ section in this forum.

Or did I get you wrong?

Well played Mike, well played

 

[IRstuff]

No, the issue is at what point do you declare a failure? Say you see 10% of energy in a undesired harmonic; is that automatically a failure? What about 20%? My point is that you need to have a rather extensive database of failures and symptoms to even begin to decide whether to perform a repair. This is further compounded by the fact that many symptoms don't necessarily lead to immediate failure.

For example, our company did a life test on a bunch of pumps and used a 10% increase in power draw as an indicator of a possible fail. One pump did indeed fail soon after that, but another one ran for an additional 10,000 hrs and never actually failed its other performance requirements in the duration of the test.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[AX3L]

IRstuff: That's a really good point actually, and the only answer I can think of is that we don't know that. But I'm thinking that the goal is to see trends rather than amplitude of peaks. It's far from failproof though, and I guess that at least in the beginning we'll have to do some manual inspection to validate the results of the analysis.

 

[IRstuff]

I expect that it will take you a few years to build up a sufficiently large database to even begin to do any predictions. Even after that, you may find that the uncertainty is so large that it's better to run to a hard fail for most failures, because any other approach will be too expensive and will incur too much shutdown time.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[MikeHalloran]

In addition to the cost of disassembling and inspecting the suspect bearings, keeping detailed records of their condition, and correlating that data to, well, anything, you will incur some additional expense associated with removing the bearings in a way that does not damage them, or separating the removal damage from the pre-removal damage.

If you _could_ get the bearings off and disassemble them without damage, which in an industrial environment is not likely, then it would make sense, upon finding no significant damage, to reassemble and reinstall the bearings and run them to some further indication without damaging them during installation, which again in an industrial environment is somewhat less than completely likely.


You need to find a different windmill to tilt at. ;-)



Mike Halloran
Pembroke Pines, FL, USA

 

[AX3L]

Hmm, this is a quite unexpected but somehow still interesting difficulty. Since we do vibration analysis occasionally to detect faults in bearings I hadn't thought about the problem of analyse the data at all. But you're probably right, the problem of having many false alarms in a startup period isn't something to overlook when calculating the investment payback. I'm thinking of buying a complete system for this by the way, there's no chance that they already have an implemented algorithm for this?

Could performing a shock pulse test as confirmation of the vibration analysis give more accurate data? I've seen some automatic fault detection systems that perform this test and that seems more reliable to me.

 

[Tmoose]

Periodic (but not continuous vibration analysis is common on US automotive machining "transfer" lines to schedule spindle overhaul.
Also in many power plants on mills, pumps and fans.

Back in the 1990s GM published a comprehensive VIBRATION STANDARD FOR THE PURCHASE OF NEW and REBUILT MACHINERY AND EQUIPMENT.

http://maintenanceforums.com/groupe...ul/423109863/GM_Vibration_Spec_V1.0a-1999.pdf

It was adopted almost verbatim by NASA for "Reliability based Maintenance."
Pages 42 and 43 have moderately detailed vibration limits for new or rebuilt machine tool spindles. In some instances new spindles are tested on a dedicated test stand before shipping. I would expect the vibration to be "different" when installed.

They would be a decent starting point for continuous monitoring.

 

[AX3L]

Many thanks for the pdf Tmoose! I think that maybe the company I'm working for have their own guidelines for vibrations, but they might be more for "overall vibration" of the machine and not for fault detection. Either way it's always nice to have something to compare and lean against when we're make our algorithms to the system, and if they differ much to ours try to find why and where we're differ.

 

[IRstuff]

There's a trade to be made on the cost of premature replacement compared to the potential cost incurred in a catastrophic failure. That's part of the cost benefit analysis.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[AX3L]

IRstuff: It's funny you say that, because lowered cost for not having to replace the bearings when it's not necessary is one of the main advantages of the continuous health monitoring according to the manufacturer of the equipment. And that makes sense, since the machine manufacturer probably have quite a safety margin of their intervals. Besides, I'm by no means an expert of maintenance but if you grease the bearings properly and even better than the recommendations you can probably save even more time.

But this might be after a long "startup period" where you have to do many unnecessary replacements before you have found limits that'll work?

 

[IRstuff]

I guess I might do it the other way around, assuming that a certain amount of lost product is acceptable; again, this is a cost trade vs. the lost product upon a catastrophic failure vs. cost of premature maintenance.

You need two sets of things, the instrumentation for monitoring the equipment condition and process measurements where you can quantitatively identify when the process becomes unacceptable. The latter is a matter of trending data and trying to find when the process goes our of its control range. You correlate that with the instrumentation measurements to determine which of the measurements or combinations thereof result in a tolerably acceptable criteria for stopping the machine to do maintenance. Once you have these two sets of things, how you go about it is dependent on your cost constraints.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529