Predictive Maintenance and the Industrial Internet of Things (IIoT)

Very excited for this webinar!

I'm giving away my webinar question but isn't Predictive Maintenance just a relabelling of Preventative Maintenance?

I did a seventeen year stint doing preventative and corrective maintenance on ground based aviation electronic systems. I'm reading this predictive maintenance concept and thinking wait a minute, this sound a lot like preventative maintenance.

The word predictive would suggest you are trying to better understand your mean time between failure and avoid downtime. That was why we did preventative maintenance, to avoid downtime.

When vendor relabel stuff it of course better, right?

colporteur    Thanks for asking.    Let me start with "This is my opinion."     But you are so right about the relabeling. 

There are so many "philosophies" of maintenance that it's ridiculous.  Most of it is marketing.  It's the same old stuff repackaged with new phrases and words.  BUT...  if it gets people to take care of their stuff, I do struggle trashing it completely.  I still subscribe to the core belief that maintenance programs fall into three categories: breakdown, planned and predictive.  Breakdown is pretty clear and it's the only one that we don't discuss at length.

"Preventative Maintenance", in my world, falls into the planned maintenance category.  We have a schedule and we do services and may replace wear items at prescribed intervals.  Here is where the lines get gray.  If you inspect the replaced parts, determine if they have useable service life and adjust your maintenance schedule accordingly, you are dabbling in predictive maintenance.  If you just chuck them in the trash, you are doing planned maintenance.  The same goes for testing.  If you perform condition tests but no one analyzes the results, you are doing planned maintenance.  Go/No Go testing is planned maintenance.  Electrically speaking, if you are checking the Dielectric Absorption Ratio or the Polarization Index, that is planned maintenance.  If you chart the DAR and PI across multiple years, that is predictive maintenance.

Computers and automated data collection have greatly enhanced our ability to do the trending and analytics.  We can now "see" small increases in current draw on a motor.  If it accompanies a corresponding increase in bearing temperature, we know that we have a bearing problem.  If the heating is less localized but still increasing, we can suspect an insulation problem.  If there is not corresponding temperature rise, we can look for issues in the driven load, transmission circuit, etc.

As you can tell, I like this stuff.  Preventive Maintenance is "Fix before failure."  Predictive Maintenance is "Fix just before failure."  The key is that someone has to look at the data OR you need a maintenance management system that will alert you to changes in the operating parameters.

Now... with you working in aviation... I'm really good with the Preventative Maintenance plan.  In the paratroopers, we used to say, "The sky, much more than the sea, is terribly unforgiving of even the slightest mistake."  Gravity is a constant.

Mike

After making my initial post I realize, maybe I'm full of sh^& and need to do a little more research.

An example of predictive maintenance I thought was around a discussion in a machine learning session E14 held a little while ago. Be damned if I can find the link! The IoT sensor that sends back data on motor vibration. The ML is tuned to only send abnormal readings. This could be considered a warning that a failure is eminent.

During preventative maintenance on a VOR( VHF_Omni Range) the tech took measurements from a series of sensor around the antenna array. Fourier analysis was applied to the reading to determine the error. If the error was outside acceptable range adjustment were required. This confirmation enabled you to set up he monitor to alarm if the signal went outside range.

The preventative maintenance to me is the exact same as predictive. If the VOR monitor went into alarm you had eminent failure.

I see an ML IoT module as the lowest form of preventative maintenance. If I get readings that indicate failure then I do something. I could say the same for the VOR but the preventative maintenance established the parameters of the acceptable range. If I could put up a pre-calibrated monitor on the VOR and say if it alarms I have eminent failure then I eliminate preventative maintenance and rely on the monitor to always be right.

Let say there is an accurate intelligent IoT sensor. You install equipment and install sensors to monitor. You now say there is no preventative maintenance. If the sensor alarms we are at corrective maintenance. Isn't the IoT sensor predictive maintenance just a trim down version of preventative maintenance.

I think I may have said the same thing twice after reading this over. I never grew up in this model, mine aligned with the three you indicated. Our were two just under different terms. Preventative and corrective maintenance.  This conversation would be so much easier with a coffee over a napkin for diagrams.

After making my initial post I realize, maybe I'm full of sh^& and need to do a little more research.

An example of predictive maintenance I thought was around a discussion in a machine learning session E14 held a little while ago. Be damned if I can find the link! The IoT sensor that sends back data on motor vibration. The ML is tuned to only send abnormal readings. This could be considered a warning that a failure is eminent.

During preventative maintenance on a VOR( VHF_Omni Range) the tech took measurements from a series of sensor around the antenna array. Fourier analysis was applied to the reading to determine the error. If the error was outside acceptable range adjustment were required. This confirmation enabled you to set up he monitor to alarm if the signal went outside range.

The preventative maintenance to me is the exact same as predictive. If the VOR monitor went into alarm you had eminent failure.

I see an ML IoT module as the lowest form of preventative maintenance. If I get readings that indicate failure then I do something. I could say the same for the VOR but the preventative maintenance established the parameters of the acceptable range. If I could put up a pre-calibrated monitor on the VOR and say if it alarms I have eminent failure then I eliminate preventative maintenance and rely on the monitor to always be right.

Let say there is an accurate intelligent IoT sensor. You install equipment and install sensors to monitor. You now say there is no preventative maintenance. If the sensor alarms we are at corrective maintenance. Isn't the IoT sensor predictive maintenance just a trim down version of preventative maintenance.

I think I may have said the same thing twice after reading this over. I never grew up in this model, mine aligned with the three you indicated. Our were two just under different terms. Preventative and corrective maintenance.  This conversation would be so much easier with a coffee over a napkin for diagrams.

colporteur  We have a business consultant that we use for our annual planning meeting.  He says that our team is often in "violent agreement"; very impassioned versions of the same thing.  We may be drifting that way.

I think part of the struggle with the phrase "preventative maintenance" is that we've used it for so long to mean everything other than "wait for it to break."  I'm surprised that one of the new maintenance programs hasn't been named Eclectic Maintenance.  Data and trend based maintenance fits my understanding of predictive maintenance, but you still have to schedule time to look at the data.  Report by exception still qualifies as predictive, but what if a squirt of oil every quarter could prevent that from ever happening?  If the lines are really gray between your preventive and your predictive maintenance, it means that you've got an excellent maintenance program.  Every good program has both calendar based checks and a data based checks.

I remember the article you reference.  I though it was excellent because it processed the data, learned what was good and not good and only reported when something was outside the norm.  THAT is handy!

So I'm excited about the webinar because most of the maintenance that we do for our generator customers is calendar based - per regulations.  It's exciting to think of adding something like the VOR monitor you mentioned so my team could go out at 6am instead of 1am to fix it.  Or maybe even tomorrow - or the next day.  The technology already exists, but they are constantly improving.  Most of the monitoring my customers do is subscription based and many would like to own the solution.  These webinars get me one step closer each time.  Whoo-hoo!

Okay.  maybe I should have held off on that third cup of coffee.       BR, Mike

Awesome response M²

Me thinks we are both on the same page.

Our support teams looked after all sorts of gear. Most of it was defined aviation systems but there were other support systems. System that were necessary for the aviation part to run smoothly. Electrical/electronic system, mechanical system, HVAC systems, distribution systems and computer systems were all slotted into preventative and corrective maintenance.

You don't oil computers but a little squirt on the bearing of UPS diesel generator motors that sits idle for months is a not such a bad idea. Especially when the system is located at a fly in site up in the high arctic.

I like the predictive IoT. If the UPS system has sensors that collect data to evaluate performance, maybe instead of twice a year firing up the generator on a site visit, it is done four times a year and two of the times not visit is made by using predictive to determine no problems. Yeah the squirt of oil is done on a trip.

The Non-directional beacon radio signal at a site in the high arctic was monitored by a receiver two miles from the site. The receiver could detect changes in beacons transmission signal parameters. It never told you that the number two blower in the 1Kwatt power combiner of the beacon was was running at reduced RPM. During preventative maintenance site visits blower checks were required (god I hate cleaning flies from blower grates!). Can you guess why?

I'm concerned when I hear that predictive maintenance reduces cost, it is one size fits all. I think it is a tool in the toolbox of maintenance but I don't believe it is one size that fits all. That saving money promise pushes managers to take action that I don't feel is prudent.

I am hoping they will provide their definition of predictive, preventative and corrective maintenance before they start. I think clearly understand what they mean helps me not jump to conclusion about why the goal will achieve.