Predictive Maintenance in Practice: From Prediction to Prevention

Glenn Gardner: All right, and Zach, I know you're finalizing the live streams across all the channels. Looks live on my end. Looks like we're live and broadcasting over here.
Zack Scriven: Perfect. ⁓ All right, there we go. All right, cool, perfect. All right, it's coming through. There we go. All right, and we are gonna go live. And we're gonna go live in three, two, one. And we're live, everybody. Welcome, welcome to today's webinar, Predictive Maintenance and Practice. Thank you guys for signing up. We had over 1,000 registrations for this webinar, from operations managers to quality and... ⁓ engineers, predictive maintenance specialists. We have an ⁓ expert panel for you guys today. We're going to be going over the steps from prediction all the way through prevention. So I have Glenn Gardner, who is the co-founder of Abellara and is a predictive maintenance expert. Glenn, thank you for joining.
Glenn Gardner: pleasure.
Zack Scriven: Mark Kincaid, who is a predictive maintenance expert as well on vibration monitoring, who is with WAITS, who's going to be dropping some insights on ⁓ using predictive maintenance solutions to predict failure. Mark, thank you for joining us.
Mark Kingkade: Glad to be here, thanks.
Zack Scriven: Last but not least, have Kyle Beers with MaintainX who's going to be talking about that integration, that seamless workflow from prediction through scheduling a work order inside of your CMMS. We really like MaintainX. We are biased, but I do think it is one of the best CMMSs on the platform right now. to have all these parties together for you guys is, hopefully it's going to be worth your time. So sit back, relax, take notes, and I'm going to pass it over to Glenn, and then we're going to go from there. Thank you guys. Glenn, it's all yours.
Glenn Gardner: Sounds great, sounds great. Yeah, this is gonna be a fun hour. We're gonna go through the fundamentals of predictive maintenance, the fundamental ideas behind the CMMS, how they relate together, how they've transformed recently. So we're gonna fit all that into one hour. But I wanna start with a personal story. I'm a mechanical engineer by education and I graduated in the early 2000s. My first job out of school was at a plant engineer. I did plant engineering for many years and spent maybe half my time on predictive maintenance technologies and the other half on instrumentation and controls. So part of my day was doing walk around vibration routes, analyzing frequency spectra, doing thermography scans, et cetera. That's the predictive maintenance side. And then of course, documenting all of that in a CMMS so we could identify what needed to be repaired. And then the other half of my day was inside of the control cabinet, troubleshooting IO, troubleshooting control loops, tuning controllers, things like that. I loved every minute of both of them, but I recently realized a funny thing. If you were to somehow magically fall asleep in the year 2005 and then wake up 20 years later, If you looked at the landscape of the predictive maintenance side, what you would see is that the industry has completely transformed. The physics are still the same, but the way that we go about implementing a predictive maintenance program is entirely different from what you would have experienced 20 years ago. And ironically, for better or worse, if you were to wake up 20 years later and look inside of a control cabinet, you would see roughly the same exact thing. The world of controls looks very much like it did 20 years ago. There's maybe a few more VFDs inside of the control cabinet, but that aside, it's still a PLC and sensors and lots of IO. And you would know exactly what you were looking at if you woke up after this 20 years slumber. Part of the purpose of sharing that history is we want to describe today, not only the fundamentals of how you can use CMMS and predictive maintenance today, but also why they've been so transformed over the past 20 years, sort of in contrast to the controls world, which looks very similar to what it did 20 years ago. ⁓ That is the big theme. The great news is the takeaway is when I started, it took maybe one to two years and maybe one and a half to $2 million of investment to get a world-class predictive maintenance program up and running linked to a CMMS and, you know, humming right away a world-class program. ⁓ That same feat is maybe a tenth of the cost and the time on both fronts today. That's a really big deal. And it's all due to several structural changes that we'll get into momentarily. So. Before we dive into all that fun stuff, I wanna open with a simulator. This simulator is gonna show how we think about vibration as a predictive technology, and it'll give you a gut feel for why it's so useful in identifying various machinery flaws. Before I pull this up, I wanna say a big thank you to our friend Alan. Alan Friedman is the founder of Zenco, a vibration training company. He very kindly let us use his simulator live on the call here, so. Alan, thank you. a huge fan of this simulator and let me get it live here.
Zack Scriven: We've got over 30 people joining in the LinkedIn Live. So if you guys are joining, let us know where you're joining from. And thank you for being here. And we'll go ahead and, all right, we can see your screen now, Glenn.
Glenn Gardner: Okay, fabulous. So ⁓ if you happen to be an engineer, there's no doubt that you've encountered this system before. Somewhere around sophomore or junior year, every engineer in humanity has to spend an all-nighter learning differential equations and how to model the system that you see here. The good news is I'm not going to show a single equation in what I share here, but instead my goal is to show why vibration is such a directly relevant technology for figuring out what's wrong with rotating machinery. ⁓ So first of all, I'm going to start with the basics. Whenever you see a spring mass damper system like this inside of any textbooks or training materials in vibration, I want you to picture in your head the bearing and the associated framing underneath of the bearing, including the base plate. Basically, any time you see a mass like this and a spring, and I'm going to have some fun and let that thing start vibrating around, ⁓ there's lots of mathematics behind the scenes. That's what vibration analysts do. But I want you to picture that that is representative of what the machine is doing as it's supported on a bearing and the associated frame and the base plate, et cetera. So this is not abstract. We're directly modeling what a machine sees through the forces in its bearing. ⁓ Now, the most common failure mode that we see in rotating machines is something called unbalance. It makes sense if you think about it. Take a pump, a compressor, a motor, a generator. These things are spinning really quickly, typically 1800 rotations per minute, if not 3600 rotations per minute. And so if that rotor that is spinning super quickly is not perfectly balanced, you're going to have a force directly put on each of those bearings. That's what this spring mass damper is set to represent. So I'm going to show you what happens when the thing is reasonably well balanced. I'm going to displace this spring just a touch. This represents a reasonably well-balanced rotor, meaning that there isn't a lot of unbalanced force. And it's no surprise, it's completely intuitive that when you watch the spring vibrate, it's relatively small displacements. So in your... ⁓ yeah, please go ahead, Zach.
Zack Scriven: Glenn, real quick. Yeah, could you zoom in on your Chrome tab like 10 or 20 %? Just zoom in. Yeah, perfect, perfect, perfect. Right there. That's perfect. Yeah.
Glenn Gardner: How's that? Yeah. Awesome. And tell you what, for posterity, I'll do it one more time. All right. So what we're seeing.
Zack Scriven: And then if you can minimize that little riverside thing in the bottom right. just like perfect. There you go. Awesome.
Glenn Gardner: Yeah, totally. ⁓ So what we're modeling here again, please in your mind, I'd like you to picture a pump, a compressor, a motor, any rotating machine, and specifically the bearing that's supporting it and the structure below it. That's what we're simulating here with this spring mass damper. And you can see that without touching any math, it's very intuitive. If the imbalance is small like this, the resulting vibration is small and life is good. Now, what happens when the imbalance is much larger? Although the math is hard, the intuition is very obvious. I'm going to displace this spring significantly and you're going to see a really big vibration. This is not a surprise, but this is what happens when you have a really badly unbalanced rotor. There's lots of reasons that can happen. Folks like Mark Kincaid, who's on this call, study this all day and night, but they can include things like buildup of debris on a fan. They can include the wear of the veins on a centrifugal pump. It could be that it was a poor manufacturing process in the first place and the rotor was never balanced when it came from the OEM or that it was poorly repaired and not balanced properly. Many, many sources of that, but one of the most common vibration problems you'll find is unbalance and vibration is really good at figuring out that problem. Now I'd like to do a slightly different failure mode. Another very common failure mode is bearing wear. We've all seen ball bearings. Ball bearings are relatively simple devices. You just have an inner circle. We call it the inner race. You have an outer circle. We call that the outer race. And then you have spherical balls in between that are rolling and reducing the friction that the shaft feels as it rotates. Now, in that case, we're dealing with something that's much more transient. If you picture a flaw such as a crack on the inner race of the bearing, what's going to happen is every once in a while as a ball goes past that little crack, you're going to see a ping, a transient. And that is represented in vibration like this. So we're going to show a transient with some damping here. Rather than vibrating forever, when you have an individual ping, when you pass over that particular flaw, you're gonna see this, which is a transient time series displacement. And in your mind, what I'd love you to picture right now is every single time one of the many bearing balls goes past that particular crack, you're gonna have another transient. And so that bearing is continuously experiencing impact and transient, impact and transient forever. And that's another type of thing where the mathematics are complicated, but intuitively it's super obvious that we can pick up the presence of those small transients inside of a vibration measurement. I share both of those simply to give you a feel for why vibration is such a powerful technology. We just went through two of the very basic vibration failure modes. There are many more. There's maybe 50 common vibration failure modes that we can catch every single time. And there's maybe an additional 50 on top of that. They require a little bit of the art of the trade, a little bit of experience to spot, but they can still prevent failures well in advance. And my goal here is to give you an intuitive feel for why it is that vibration is such a great technology to do that. So I'm going to stop sharing for just a moment and do a little framing before I hand the mic over to Mark Kincaid. ⁓ The other thing that I want to share is CMMSs. So half of today's talk is about vibration and predictive maintenance. The other half is about the sister technology, which is the CMMS. An important theme that we'll repeat multiple times is that predictive maintenance is there to identify a machinery flaw early in the degradation process so that you can repair that flaw quickly and get back up and running inexpensively and without any collateral damage. In order to do that well, you must have both sides of the same coin. You must have a predictive technology. like vibration to identify the flaw in the first place. And you also must have the ability to plan, schedule, and implement the repairs. Vibration is the predictive technology. The CMMS is what manages the workforce on the front line and plans and schedules and implements the work so that you can repair that fault. Two sides of the same coin, you must have both to have an effective predictive maintenance program. And that's a big theme of today. So. I'm gonna pause there. Mark or Kyle or Zach, please feel free to add a little bit of commentary to my opening speech. And otherwise, I'm gonna hand the mic in sequence first to Mark Kincaid, who's gonna go into a little bit more depth about what it actually looks like to do vibration analysis in a real system. And then over to Kyle for the same story on CMMSs.
Zack Scriven: Perfect. Yeah. Do either of you have any, I think you want to add to Glenn's section before I chime in and then we'll pass it to Mark.
Mark Kingkade: Yeah, I'd like to say one thing that Glenn's talking about is no one individual or no part of the program individually can make or break the program for you or make or break successes. The important part is the open communications between all stakeholders. We can find the problem, we can address the problem, ⁓ then it's gotta go to the maintain access, the CMMS to get it implemented into the system and then we need while we're doing that we need to have open communication between operations, production, safety, environment, everybody else to figure out how soon we need to address this and the impacts of the failures if they happen.
Glenn Gardner: Well stated, Mark. Well stated,
Kyle Beers: Mark, I thought about something similar, Glenn, when you were talking about what this looked like in the past. And the big thing of what we can do now is connect these stakeholders. I remember I started my career in reliability and similar to you, I had vibration reports. And next thing you know, I had to turn it into a work order, but I had to jump through hoops in order to do that. to see this kind of all come together in the workflow that we've created ⁓ is pretty great. So really looking forward to. get more into it but Mark I can I can hand things back to you.
Zack Scriven: Yeah, Mark, please, when you're ready, share your screen.
Mark Kingkade: There we go. ⁓ This is just a basic kind of concept. This is what I live in. I sit at my desk or out on the floor and I look at squiggly lines all day. ⁓ And those squiggly lines tell me things. You just got to learn to listen to them and understand what they're trying to tell you and understand when you need to react. But the top...
Glenn Gardner: you
Mark Kingkade: Basically, you're working with your frequencies. That's the number of events per unit of time that happens. Like Glenn said, your outer race defect, your inner race defect, ⁓ they all have their own individual frequencies, your unbalanced misalignment. They all have discrete frequencies that we look for, and that's how we address them. And then we look also at the amplitudes that are showing up in the waveforms in the spectrums. Your amplitudes are along the left side of your of your graphs here. But it starts out with ⁓ your vibration signal when I collect my data, it starts out with a waveform, this top one. And if you look at it, it's really, really busy. And it takes a lot more work and education to be able to decipher all your different things that are happening from this single ⁓ waveform. But then after the waveform is taken, we derive what's called the spectrum. Waveform represents time-based events that are happening. It's over milliseconds here. And then we move it into frequency in the spectrum, and that's frequency-based. So along the bottom here is your operations, know, the number of times it's happening during a rotation. we typically, you can call them orders. or you can look at it in CPM or you can look at it in Hertz. this is basically telling me that anything that's happening up here, when it's turned into a spectrum, that means you've got events that are happening on a regular basis and you have ⁓ several of them that happen over a period of time. So it basically takes your waveform and makes it into the spectrum where all your events are turned back to, it's just easier to look at your different defects that are happening. So my next one here is, go ahead.
Glenn Gardner: Yeah. Mark, a quick comment. I always found it remarkable when I was first exposed to, you're showing a time waveform and a frequency spectrum there, like the fundamental building blocks of vibration analysis. I even today find it remarkable that what looks to the naked eye, like just complete noise, you can run it through algorithms and or a trained human that carefully looks at the orders and octaves and orders of magnitude. And there's real information embedded in both of those signals, not just real information, but a ton of real specific information. When I look at it right now, I can't help but think it's really genuinely remarkable that what looks like totally random static noise up top has a whole bunch of information that directly leads you to the failure mode.
Mark Kingkade: It's, I agree, it's, mean, your machine is quietly talking to you and it's, you just need to listen to it, because it'll usually give you plenty of warning if you're listening and understand what it's saying.
Glenn Gardner: Totally.
Mark Kingkade: So this here is just a pretty simple one. ⁓ They replaced this gearbox in a kind of an emergency setup and what happened is this base here was flexing back and forth. Along when you look at the spectrum here, this bottom one shows it how it was ⁓ when everything was solid. But when we see this peak here in this top spectrum, or this. the top spectrum there, that peak there is related to two times the output speed of that gearbox. That gearbox was running at about, it was like 9.8 RPM, we'll call it 10 RPM. And then after the work was done, you see this jump here, and it's pretty significant in this one, it's low speed, but it jumped up to two times running speed. And that's where we're saying when you look at the spectrum, two times running speed is a really good indication of misalignment. In this case, it was the misalignment between the gearbox and the shaft that it was related to. This was a toaster. shaft it was hooked to and the vessel that it was underneath was approximately about 60 foot tall. So you got that shaft that's oscillating back and forth there. So it's causing misalignment. But the root cause of the misalignment was that structure that the gearbox was mounted to was flexing back and forth. So we were able to go in and add some stiffeners to that, those two I-beams that are underneath there. If you look at these I-beams here, we added some stiffeners and things like that stiffen it up and then it dropped that high two times running speed back down to more normal what we're seeing there. ⁓ But if you look at this, if we're looking if it had a bearing issue or something like that,
Glenn Gardner: Mm-hmm. Love it, love it.
Mark Kingkade: then a lot of your noise and your peaks that you're seeing would be predominantly here in the upper part of the spectrum. If you see several running speed harmonics ⁓ that would just be a whole bunch of peaks along here, along this bottom line, they can be an indication of rotating looseness. ⁓ If you've got a high one times that's high in the horizontal direction and not as high in the vertical direction, that's your indication like what Glenn was talking about that comes out to your imbalance of your fan or your rotor or your ⁓ your impeller or something like that, but basically what we're doing as the vibration guy is we're taking and collecting this data and many times ⁓ you're gonna look at stuff and And if you go talk to operators or people that aren't really familiar with what you're doing, they're like, there's nothing wrong with it. I can't hear it. I can't see it. I can't really feel it that much. So they have a tendency to not want to believe you until they understand what vibration and can do for it. My background's been in vibration for a little over 30 years now. I come off the floor as a floor mechanic for 45 years in industrial maintenance. And I was with that group. In the beginning, it's like they'd come up to me and tell me I've got a problem developing in this motor or this gearbox or this blower, and you got a bad bearing. And it's like, you're crazy. There ain't nothing wrong with it. So when I got started, it's like, let me tear this thing apart, and I'll prove to you that that bearing was fine. And then you tear it apart, and it's like, crap, there's just a little spall in there. You may just have a little pit on a ball. You may have a
Glenn Gardner: Thanks
Mark Kingkade: cracked race or a chip tooth and stuff. outside of it, when I'm on the floor and stuff looking at it and feeling it, it doesn't seem that bad. But that's the very early indications. And that's where the communication comes in and the education. Yeah.
Kyle Beers: And Mark, ⁓ correct me if I'm wrong here. You said something that was really important that I think relates to what Glenn said in the beginning is that we've come a long way in 20 years and we used to do routes for vibration and now we're doing the continuous aspect. So you said something that was really important is that I essentially have to convince this person that something's wrong with it. And to your point, those squiggly lines may not mean a whole lot. But what you do have, I think there's a third piece of data here where we can actually see this data over time. and ⁓ assess as it's getting worse. And I think that's something that really speaks to the audience when we have to prove that an issue is worsening.
Glenn Gardner: Yeah. It's a great point.
Mark Kingkade: Exactly. Because back in the day I was running my routes every 30 days typically. And now with the online systems that you have out there today, at a minimum, most places are running that same route every day. know, and I'm not having to spend, you know, an eight hour day out there running this route because this thing is collecting it. And so I can see my trend changing a lot earlier than I could in the past. And that's
Glenn Gardner: Yeah. Yeah, 30 days is great. Just like Kyle was saying, I used to have to fight to get 90 days to get quarterly measurements, let alone every 30 days. But in either case, like you're talking three to 12 measurements per year. Like that's not much of a trend compared to having a measurement every single day or even faster if you choose. I think Kyle nailed it. You see the dynamics immediately. I'm sorry, Mark, please go ahead.
Mark Kingkade: Right. that- Right. And then No, yeah, just exactly like what you're saying. The trends develop over time. And so now I can watch that trend happen over the last week, the last month or whatever, and see how quickly it's changing, which is going to give me a better indication of the severity that's happening, you know, and how soon I have to address it, how soon we got to get that information to the guys like Kyle to get a work order, you know, get into operations and find out when I can take this machine offline. so it doesn't disturb their production schedule and stuff, but also get it addressed early enough so that my repair costs are a lot lower than they would be if I'd let it go longer. ⁓ That's.
Zack Scriven: Hmm. Yeah, Mark, I got a question for you that came in Yeah, so ⁓ I mean you're kind of touching on like, know the the six big reasons of savings for implementing a predictive maintenance program which I know Glenn was gonna save that for us the end cuz like I mean, honestly, you know, it's crazy is like, you know, I'll get to the question here but it's like if Predictive maintenance and CMMS used to cost millions of dollars and it still was done So it must have still been worth it somehow But I think the trend is now that the cost has come way down, mean, how critical was this piece of equipment here? Was this like a super high or like more of a medium priority critical?
Mark Kingkade: No, no, this was high without this, the plant went down. You were just, you were done. Yeah, because, yeah, this was towards the end of the production and it's after everything had been worked on and the formulas and stuff like that. So if I lost this, then I can't put product out the door to the final stages of the operation. So this was a DT.
Zack Scriven: Got it. So this is high critical. It just costs way less to implement.
Mark Kingkade: and you have DCs inside of the extraction plants and stuff. these are the two most critical assets in this facility for what was going on. And it's not easy to work on these things. A lot of them are in explosive environments. So I just can't go out and run a weld and throw a brace in there. I have to do a lot of pre-planning and things, purging. And that's where the safety aspect comes into it too.
Zack Scriven: Yeah.
Mark Kingkade: We bring in environmental, what happens if this fails and stuff gets released? know, can't safety side of things since this is an explosive environment, I can't just go in and throw a brace in there and weld it up and be good to go. I have to clean and purge this area, make sure it's safe for the guys to go in and do the work. You know, so that's the big part of it.
Zack Scriven: Yeah, awesome. Thanks, Mark. So the question from Jim is, and I think Glenn might have the answer to this one, but he said, isn't baseline current and actual current just as valuable in addition to vibration?
Glenn Gardner: Yeah, you're good.
Mark Kingkade: Yes.
Glenn Gardner: ⁓ Absolutely. Yeah, please go ahead, Mark.
Mark Kingkade: No, I was going to say, yeah, it's exactly true. mean, there's no one technology that's going to do everything for you. In my career, I've done vibration. I've done ultrasound. On this gearbox here, we did ⁓ vibration. We did ultrasound. We did oil analysis. We did motor testing, MCA testing on the motor and the electrical side of things. So you have several technologies that will be applied to most of your assets. And that's part of your maintenance strategy that you've been develop over time. You know what technologies work best for this. So in this case on the gearbox, you know you're looking at basically oil analysis, vibration and ultrasound. If we're just looking at the motor, you know I don't really have oil analysis but I can do grease analysis on the bearing if I can get access to the grease. I got motor testing there, I got thermography to look for hot spots and windings and stuff, then you've got the vibration. You know, so it's, it's an all encompassing.
Zack Scriven: And Waits is specifically focused on vibration, or do you have products that span other areas of predictive technologies?
Mark Kingkade: No, is, know, of bread and butter is vibration, but we also, the sensors that we install measure the temperature at the location of the sensor. So we understand the temperature at that bearing location or at that gearbox and stuff. ⁓ We also pull in other things with what we call our universal adapters. If you've got a tap in there that's already monitoring current and voltage. we can pull that into the dashboard and then that goes into our trends too. ⁓ So it's a way to get as much data as you can quickly in 24 hours a day, seven days a week, so that it streams to the dashboard so all stakeholders understand what's going on with that certain machine.
Zack Scriven: And then Jim asked if there are multiple bearings, could it identify from which bearing from patterns?
Mark Kingkade: Yes, yes, each bearing will have its own characteristic ⁓ fault frequencies. And then that's also related to the shaft speed. So as we, on this gearbox here, I could pull it up. There's actually about 68 different fault frequencies in that gearbox that I can pinpoint to point it to which bearing it is, which gear it is, which shaft it is. ⁓ If it's an inner race, an outer race, if it's a bearing. roller bearing or if it's a cage frequency. So it takes a while to set this stuff up. And now, you know, in the past I had to know all that stuff and then I was the person that had to track it. If I know all those frequencies and I plug it in with all the algorithms and stuff like that that we use today, the computer is sitting there monitoring it in the background. So in this facility, I did 167 pieces a month when I would go in there. with the computers and the algorithms and the AI that we have today for these things here, if I plug all those frequencies in there, then when I show up to that plant, now maybe I only have to really look at five different machines that are actually showing a change because the rest of them, the other 162, have showed absolutely no change over the last month. So now instead of trying to spend a week doing 167 machines, ⁓ Now I can do just the one day, I can spend a lot of time just focused on the machines that are showing a change. So now I've got all this extra time. And like Glenn says, we're trying to do more today with less people. So this way I can focus on the stuff that's actually showing a change. So any more questions then, Zach?
Glenn Gardner: Those are great questions from the audience. Yeah, solid.
Zack Scriven: No, that's pretty good.
Mark Kingkade: Okay, then this one here, I just wanted to throw it up here. This is a gearbox that actually they broke the case in half. I mean, trashed the thing out. For six months prior to this failure, we had guys telling them that, you know, you've got a bearing that's starting to go bad. Oil analysis was showing gear wear and stuff, and it was showing it. We had chrome showing up in the oil sample and stuff, which was coming from bearings. and things and the site was, you know, it's like, hey, I'm not seeing anything. I'm not feeling anything. I don't think it's really that bad. So for six months we were pressing about, you know, hey, this needs to be taken off service. But again, this was a critical asset for them. So if they take this thing down, the plant had to go down. In the end, basically three days before this thing blew apart, like you're seeing it in these pictures, you had the bearing cage frequency for the second intermediate shaft, which is this bearing right here, that bearing cage frequency had jumped over 1000 % in the matter of about two or three days. So that was the end of the failure. For six months, we've been telling them and they weren't doing it. And then all of a sudden it jumped by 1000%.
Glenn Gardner: The end is near.
Mark Kingkade: Even at that point in time, the site was not willing to accept how bad it was. This was a Friday afternoon when the technician went in and told them, hey, we need to really shut this thing down. The maintenance manager at the facility said, I don't think it's that bad. We're not going to screw with production. We'll come back Monday and we'll discuss it more in depth Monday. Well, Sunday afternoon about 11 o'clock in the morning, they got a call that's like, hey, the vibration guys need to come back on site because our gearbox just blew up. So you're taking that guy off of his Sunday afternoon with his kids and his family and stuff and he's going, but the whole thing was everybody wasn't communicating, everybody wasn't putting the stuff forward to go ahead and make a... ⁓
Kyle Beers: you
Mark Kingkade: Informed decision on how to how to address this in the end. This was a 1.5 million dollar failure There was no gearbox that was readily available in the United States They found one in another country had to fly it from that country to that was on a 50 Hertz setup They had to fly it from that country. They had to fly it to the United States They had to tear it apart and regear it so that they could get the same 10 rpm output using a using a ⁓ 60 Hertz motor. So that's again that's where the conversations come in and this is the part that that we you know hand it off. Once we see the problem we give it to Kyle and let Kyle and his group write the work orders do the job kidding and things.
Glenn Gardner: Yeah. And Mark, you stole my thunder. was going to ask, what's a lead time on a replacement for something like this? I don't know, three days a week, something like that.
Mark Kingkade: Yeah, and it's gotten worse since, you know, in the last couple of years, the lead time, even on some small stuff has gotten crazy. You know what? I used to be able to get in a week or two. I may have a little half horse pump that's critical to my operation, but now that lead time, maybe, you know, two or three months.
Glenn Gardner: Yeah. Yeah. Yeah, it's no secret that supply chain is harder than ever for every physical good you can imagine. And a person earlier asked about the return on investment calculations. Usually it's swamped by things like this. There's a rigorous methodology I can go through at the end, but the gist of it is don't let this happen. Like when this happens, your whole process goes down. The downtime is the most costly thing.
Kyle Beers: Well.
Glenn Gardner: and you typically can't get complex machinery like this any quicker than like six months. I don't know how many months they were down in this case, but.
Mark Kingkade: In this case here with what they did in the hoops they jumped through, we were down for about six weeks on it. But during that six weeks, the facility had to buy product from their competitors to fill their contracts. You know, so they were getting a double whammy with the plant being down, the cost of doing the repairs. Now I got to go out to my competitors and buy the products that I need to fill my contracts. So it's just a snowball effect.
Kyle Beers: Hmm.
Glenn Gardner: Yeah, totally.
Zack Scriven: Thanks so much, Mark. ⁓ Just out of respect to time, I'm going to pass it over to Kyle right now just so he can get into it. But we're doing good on time. So ⁓ this is a great conversation. I'm looking forward to rounding out that story and then being able to pass it back to Glenn to tie a bow on it and then answer any questions that the audience may have. So Kyle, the floor is all yours.
Mark Kingkade: Yep, sorry.
Glenn Gardner: You Great.
Kyle Beers: Yeah, absolutely. Seeing the screen, okay.
Glenn Gardner: Yep, you are live. Looks good.
Kyle Beers: Wonderful. And Mark, you set me up ⁓ great. So thank you. ⁓ But for anyone just joining, my name is Kyle Beers. I am Solutions Consultant with MaintainX. And a lot of what I do is help industrial teams connect their machine data to their maintenance workflows here inside MaintainX. So very applicable for today's conversation. But just to kind of pick up where Mark left off, I want to talk a little bit about how the information from Weights gets to MaintainX and what does that mean, right? that diagnostics, those insights have been generated in the weight system for all the insights that Mark generated. Now, what we're left to do is to pass that information off to maintain X as the, we'll say the AI powered execution layer. And the idea being that we want to give the technician the right information to do their jobs correctly, right? And we talked about earlier, there's multiple stakeholders, there may be reliability groups, there may be ⁓ other folks leadership that need to communicate this information in a streamlined and effective way. And that's what the WAITS team has really done. They built this direct integration between these two systems. So the moment that diagnostic insight is flagged, that work order is going to be automatically generated and assigned to the applicable team. This is way different rather than that alert sitting in a dashboard or buried somewhere or in a PDF report or having to work out of two systems, right? We're automating that action. And I think The big reason why this matters more than people realize is that, yes, the hardware is very important. The diagnostics are also very important. But in my mind, predictive maintenance initiatives, they fail because the information never gets into the hands of the people who can actually do something about it. And Mark, you told the story the best, right? If we knew this machine was going out and it had a six-month lead time, and there's all the other things we can do to make sure we have our ducks in a row, ⁓ we want to do that. when that work doesn't get done or it gets done late, or without the right context, you're leaving a tremendous amount of value on the table. ⁓ And like you just explained, that's the difference between a costly unplanned failure versus a planned corrective repair that you're in complete control over. So I wanted to highlight that workflow for everyone. ⁓ Glen Mark, ⁓ Zach, any comments before I kind of get into? A little bit about the work order inside maintainX and how we execute it.
Zack Scriven: Mark's got a question.
Mark Kingkade: I would just like to add, Kyle, that with the system like MaintainX, for the first time in my life, when I first started, I'd have an IR report, I'd have an oil report, I may have an ultrasonic report, you know, I have a MCA report on a machine, but nothing was ever in one spot to make it easy to get that information. Now with systems like MaintainX, I can go and put all that information that I have on that machine in one location in my CMMS. So when I start seeing things ⁓ as an operator or I start seeing things as a maintenance tech, I can go in and see what all these other technologies and stuff have already flagged on it. And that's where the real power of MaintainX comes into play for us.
Kyle Beers: Yep, you bring up a great point. ⁓ Centralizing that data in a place where it's accessible to all those stakeholders. So yeah, I appreciate that, Mark. Let me move on here and just kind of talk about, again, what happens next, right? So what you're seeing is the maintainX mobile application here executing a work order through the use of AI AI co-pilot. ⁓ So really what I want to highlight is that when that work order lands in maintainX, what are some of the things that are happening and what are some of the things that technician is focused on? Well, the first thing being it's assigned, that work order is not just created and is ⁓ out ready for someone to grab, it's directed towards a a team of individuals that can execute on it quickly, right? This is going to help reduce the reactive nature. So in this flow, every alert is going to become a planned repair, rather than that emergency call out. And that's going to move the needle closer to our predictive maintenance strategy. then second I want to highlight is that with using maintainX Co-Pilot, there's an incredible amount of information already in your asset profile, already attached through maintenance manuals and asset history and even recommended parts. And that's going to be there directly in your hand. It's going to help close that gap between ⁓ tribal knowledge that's constantly walking out the door and allowing us to really have a lot of confidence in the work we're doing. ⁓ Next, I want to add that the time and parts are all being tracked when you're going through this work order. Every labor hour is being logged. All your parts consumed is going to be tied directly back to that asset. And why that's important is that You know, for the predictive maintenance program to be successful, we need to make sure that we have metrics that are rolling up that can help justify the cost being spent, what you're spending, what you're avoiding, what the program is actually delivering. So, I think this is a great tool to help track those costs as well. And last but not least, you know, it's our system of record, right, for continuous improvement. We talked about catching the failures, but a big component, I think, too, coming from the reliability world is defect elimination. So when we're talking about predictive maintenance, right, we want to assess what the root cause was and capture that, why it failed or maybe almost failed and what was done about it. you know, I think that feeds nicely into our predictive maintenance strategy. So a little bit about what you can expect once that insight from Weights is creating that work order inside maintainX.
Zack Scriven: So one question. So. ⁓ I mean, maintain X is so far ahead of the competition that you could not, like, even if they only tried to build a maintain X competitor, they just wouldn't be able to catch you. So what I like about maintain X and this kind of touches on to where Glenn's going to talk about like the financing and the financials aspect of it. You really touched on it. Like what are those different KPIs that are rolling up and to demonstrate how we're using this system to save, you know, not just on increased availability, but reduce defects. it's got to be repeatable, right? If you're doing preventative maintenance, you need to be able to repeat that preventative maintenance using the same inventory parts, making sure you have those on hand. So it's that full life cycle. It would take you years to code that yourself. So taking this off the shelf, plugging it right into your stack. My question is from an architectural standpoint, and this is kind of a layup, but cause I know you, I know you guys as platform, let's say I have other sensors, other, other applications outside of weights, right? ⁓ weights is my vibration solution. From an architectural standpoint, which way are you bringing those data points in to maintain X from other systems?
Kyle Beers: Yeah, great question. think Glenn, you might get into this a little bit, but we want to take those raw machine signals coming from your PLCs up through your SCADA system like ignition as well. So this is one use case with weights where we're unlocking predictive maintenance. There's a lot of things we're doing with other machine data that tie into usage-based PMs, right? We're not working off a calendar anymore. We're working off of runtime hours or cycles. ⁓ Things like that, ⁓ those machine signals that are staying at the machine level, we're now using that in maintenance to drive smarter PMs to your point. And we're really climbing that ladder from a digital to a manual to a digital to a usage-based condition and ultimately predictive. we have to take a, I think a phased approach. And so we want to utilize what data is available from the machines first, ⁓ if we can. ⁓ But as Glenn said, it's never been so easy to implement predictive maintenance. So ⁓ this is just another avenue.
Zack Scriven: We've a couple of, the chat's really active right now. We've got a couple of questions come in. Kyle, did you have anything else you wanted to touch on before I drop these questions?
Kyle Beers: ⁓ I did have one more slide, but let's keep the questions in and just keep us honest with the time here, I guess.
Zack Scriven: Okay. Yeah, I think as long as we have about 10 minutes for Glenn, think that should be good. Glenn, for you, So Michael said, why bother doing vibration monitoring and an oil analysis if you're not going to act on the results? Prediction without prevention is expensive monitoring. And you can't approve what you can't see. So ⁓ serious amount of violence happened in that gearbox to split up the case like that. Classic case of production ignoring maintenance, unfortunately. Thank you all for being here. We're trying to advocate for this. Sergio asks, was the lack of communications between the PDM and management addressed?
Mark Kingkade: Yes, after this one failed, ⁓ there was a lot of stuff that went on and it went all the way back to there was issues with the way the gearbox was originally stored before it was put into operation, you know, a couple of years prior to that. But there was a whole realignment across the whole business unit of the company that this happened at. ⁓ It wasn't to where they were beating everybody up. It's like, hey, we can't do this again. We have to have those open lines of communication.
Zack Scriven: Last but not least here, yeah, Michael added, hey, there was a lot of things additional to that gearbox scenario, including lack of criticality analysis, lack of critical spares, and no real asset management policy to guide good decision making. This is all about taking data and turning it into informed decisions. That's what Industry 4 and Digital Transformation is about. We have Brandon Walsh. Thank you for sharing these practical advices on end-to-end predictive maintenance use cases, Wates. highly applicable to our factories, rotating CNC machines. Anyone in the audience, ⁓ if you guys want help implementing this solution or implementing a predictive maintenance program, that's what Abelara does, right? Glenn, he's a consultant. He's happy to sit down with you, assess your needs, and recommend a solution to move forward. We're partnered with both Weights and MaintainX. So y'all will be receiving an email with next steps. So thank you for being here. Does the model consider temporal relationships or frequency-based relations to predict failures? We got some real technical people in the audience today.
Kyle Beers: Hey Glenn, you're on mute if you're trying to speak right now.
Glenn Gardner: I got muted somehow there. ⁓ Absolutely. The weight system has all sorts of sophisticated analyses, including looking at time-based domain ⁓ phases, which is really important, and it's like a micro phenomenon between two different bearings. Short answer is yes. There are many ways that you address time relationships, both at the macro perspective of something degrading over time and all the way in at the micro perspective where you're looking at phase differences between two adjacent bearings to figure out which one has the problem. So affirm yes.
Mark Kingkade: And the other thing I'll throw there is because we're monitoring this stuff 24 7 365 You know if you have any any jumps, you know, we can watch the trend happen really slowly or if all of a sudden You know, we've been watching it trend up and then we have something where our amplitude is jumped by like 25 or 30 percent in the course of 24 hours or 48 hours then we can push that to the guys like Kyle and maintain x that you know, hey, this one that's scheduled for repair, we need to push it up and how quickly can we get it, you know, the job kitted and get it set up and on the schedule so it doesn't have that big impact on production.
Zack Scriven: Awesome. Thanks, Mark. Kyle, do you want to touch on your last slide, and then we'll close it out here and answer any final remaining questions?
Kyle Beers: Yeah, sure, we'll keep this very brief. But the idea behind the slide that you're looking in front of you is that most facilities have data, and most facilities have a lot of data, arguably too much. And the problem that we're seeing today is that it's unstructured sitting in these disconnected systems. And getting it into a format that AI can actually use is still the hardest part of the equation. We're trying to make that easy at maintainX. We look at that through two lenses, our asset intelligence and our work intelligence. Asset intelligence in today's example is bringing those insights that machine data from weights and using the AI that's built into maintain X to make sure that we are looking at failure history. We're looking at maintenance strategy and we're understanding what we should be focused on next, right? Getting to that predictive state. The work intelligence is what maintain X has been doing for a long time. I mean, if we think historically, About 60 % of a technician's time is admin, finding information, figuring out what to do next, writing up these work orders. So if we can give that time back for them and allow them to be more effective in the field with the wrench time, that time on the floor, ⁓ this is where that work intelligence really comes into play. So I just really want to ⁓ leave that ⁓ for everyone to kind of see how this all converges. But ⁓ Glenn, I'm happy to pass things back to you.
Glenn Gardner: Sounds great. Yeah. Yeah, I do want to allow time for more questions. So I'll try to wrap it with one really important comment that I that I hope will hit home. So first of all, we talked through the fact that it is factually accurate that compared to say 28 years ago, maybe even 10 years ago, it is one tenth. as expensive and one-tenth as fast to implement a predictive maintenance program. We've lowered the cost by 10x and we increased the speed by 10x because of many technological changes. That's amazing and all of us are the beneficiaries of that. However, please don't misinterpret that to mean that the role of the craftsperson that's directly in front of the machine is less important today. Quite the opposite is the case. The role of the craftsperson that's laying hands on the machine directly in the factory has never been more important. Mark and I before this were on an unrelated call where we were actually stepping through 12 different handheld technologies that you literally can't do with a sensor things like using precision calipers using precision shim gauges ⁓ Looking at belt wear and a visual inspection of belt wear or using a gauge to examine a V belt wear There laser alignment. There's many of them Mark and I had a full talk about it a dozen of them for an upcoming training course The important point that I want to make sure hits home Number one, it's amazing that with maintain X and weights, we've now automated a lot of the hard work behind the scenes. The important thing is what's been automated is like the really advanced mathematics, the really difficult level three, level four vibration trained analyst type stuff. And that's great. You can go right to a company like weights who has an army of 60 or so level three certified vibration analysts. That's the part that's now centralized at a vendor like weights. The part that is not and never can be centralized is the wisdom of eyes, ears, smell, everything, all of your senses on the machine, looking at the belts and using the handhold tools of precision maintenance like laser alignment, like shim gauges, et cetera. And so we firmly believe that the role of the craftsperson is more important than ever, not less important than ever. What's changed now is you don't need level three and four vibration analysts on site like you did before. And I will close with that.
Zack Scriven: So if we have any other questions, I have a question because for those of you who aren't familiar, I can't imagine that those vibration analysts are very, they're probably pretty expensive people to have on staff.
Glenn Gardner: They, yeah, expensive, rare and retiring. The other big one is they are rapidly retiring. The technology wave kind of saved our hides, to be honest. Like this technology wave couldn't have been timed better. ⁓ There would be chaos if not for the fact that this technology wave happened to counterbalance the retirement of many trained vibration analysts. Boy, were we lucky that this happened at the same time.
Zack Scriven: So would it be fair to say though that a lot of the savings is part of that is attributed to the change in business model as, ⁓ know, like if you kind of want to talk about what it looks like to implement this, like, hey, I'm going to sit down and I'm going to roll this out. You know, we talked about maybe touching on, you know, an idea of what that would look like, but the business model has completely shifted as far as, ⁓ you know, as my understanding of it.
Glenn Gardner: Yeah.
Kyle Beers: you
Glenn Gardner: Totally. It has, yeah. Now the benefits are, it's funny, the ROI calculations on ⁓ the benefits to the end user look almost identical. What's different is how quickly you can implement the program. So the ROI is still the same or the basic foundation of the ROI, uptime, avoided maintenance costs, health and safety concerns, et cetera. That's all the same. What's different is like if you were to zoom back to when Kylan and I first started, to build a proper... reliability program, had a, there was a gentleman I think named Mike in the comments who I can tell is an experienced reliability engineer just based only on his question. We used to have to start with, okay, let's build an asset register from scratch, go to great lengths to document all sorts of nameplate information, literally writing them down on like pieces of paper in the field, putting them into a CMMS. Then you want to do a rank of criticality of all of your assets. Some are more important and some are less important to uptime and to maintenance expense, et cetera. That was a big body of work. And then you had to do RCM and FMEA analyses because everything was constrained by the fact that it all came down to everything was constrained by the walk around approach. So in the world where you needed a me or a Mark or a Kyle to literally walk around to the machines in order to take the measurements, you had to be really careful and do all of that analysis to know which machines you could even monitor. It wasn't like we had the luxury of being able to monitor thousands upon thousands of machines because it just wasn't cost effective. The reason it's so much faster now, I still always will love the approach of build an asset registry, rank things in order of criticality, do things like RCM and FMEA analyses. Much of the wisdom inside of there is now baked into the platforms like weights, which makes that super fast rather than nine months of analysis like before. And even more importantly, we're not constrained. It's so beautiful. We used to like have to agonize over, we can only fit 20 machines in this route quarterly. Like which 20 is it gonna be? And now it's like, whatever, just throw another sensor on it. Like the algorithms are taking care of it. It's a piece of cake to increment the program and just throw sensors on everything because you don't waste a human's time going out, analyzing the data, et cetera. So those are the two fundamental reasons that it's faster and cheaper now to implement the program.
Mark Kingkade: Yeah, I'd like to add just what you're saying, Glenn, the other advantage of now that we've got the wireless at one one organization I worked out, we had six plants around the country. We had two analysts that were traveling and trying to collect the data. But, you know, I'm sitting in Iowa and, you know, all of a sudden the machines making a weird noise down in Georgia. I can't get down there, you know, in a couple of hours and stuff. But with the wireless sensors. We can pull that information up and we can start looking at it. And I can have a discussion with myself and the people in Georgia and the other analysts that were on our team that may be like in Nebraska or maybe they're in Minnesota or something like that. We can sit down and have the discussion ⁓ with the stuff that Maintain X can do for us and control and see what parts we have available. and things you know, and the parts may not even be available in Georgia, but the part that I need to fix that machine in Georgia is actually available in Nebraska. ⁓ And by utilizing maintain X's ⁓ methodology and, you know, over across the whole organization, they can identify where our critical spares are. And then we can start the process before I even step out the door. of getting parts shipped down there and stuff and address it from a distance.
Zack Scriven: Thanks. Yeah, so we got one more question in which we have time to get to and it'll be perfect on time. A couple of comments came in though. Jim said in the 10X savings pitch for implementing a maintenance platform, one question always gets glossed over. Who is actually responsible for getting all the pump motor and device data into the system? Most modern maintenance platforms are fully capable of delivering everything that's demonstrated. But the weak link is that integration piece. Agreed, Jim. You're preaching to the choir.
Glenn Gardner: Yeah, I can answer that. Go ahead, Zach. I can answer that quickly. Okay, continue then. There is a good answer to that, but please continue.
Zack Scriven: My- That wasn't the question. That wasn't the question. Michael did say, I did a VA on submarines with a single cable and accelerometer connected to a raft with a gear one point at a time. It is way faster and cheaper now, Michael says. And then the question, ⁓ yeah, Alan Friedman, the vibration trainer, says, weights will have ⁓ more level three analysts a week from now. So awesome, Alan. Hello.
Glenn Gardner: Alan, hey. I love it. Alan's fighting the good fight. The level threes are not going down without a fight. Alan's continuing to crank them out as fast as he can. Good to see you, Alan. Thank you again, Alan, for allowing us to use your vibration simulator. That was really cool.
Zack Scriven: Yeah. Yeah, awesome. And so I'll ask Elizar's question here and then Glenn, if you wanted to comment on Jen's. But yeah, Elizar asked, in plants with established CBM programs ⁓ using vibration, oil analysis, and thermography, how can you be effectively integrated to automatically trigger and prioritize work orders based on condition thresholds or fault severity? To me, I would say you'd probably want a platform like Ignition to be able to integrate those with using the Ignition module that connects to maintain X. There's also other tools maintain X has an API you can integrate with And maintain X are called meters or signals. I believe right meters a meter is like a like a count So you could bring in so you could bring in those counts through MQTT. if you're using unified namespace There's a host of different ways to integrate those and that kind of touches on what Jim was saying was the gap was That lack of plan to integrate those
Kyle Beers: You nailed it, Zach. Meters.
Glenn Gardner: Yeah, meter based PMs exactly.
Zack Scriven: I'll just say this, our clients don't have that problem because we architect the solution properly, I am, know. Anyways, I'll pass it over to you guys.
Glenn Gardner: Yeah, I mean... Yeah, Yeah, Zach, what was that last person's name? That was a...
Zack Scriven: Jim Barrett, yeah Jim Barrett, he's been, yeah, he's a good friend of mine.
Glenn Gardner: Okay, so yeah, that's an advanced, ⁓ I don't happen to know Jim, but I already can tell Jim's an advanced practitioner. ⁓ The real answer there is one of the tricks, one of the reasons that I, excuse me, maintenance MT and OT have always been divided is they use fundamentally different data structures. I'll skip the long version of the speech, but if you go way back into your linear algebra classes, you'd hear there's scalar vector matrix tensor data. These are different data structures and you can't store them in the same way in the same database. The reason that's relevant here is thermography, for example, is matrix data. Motion amplification is tensor data. Vibration is vector data. These are all fundamentally different data types. We have found that it's not practical to integrate all of them into one database. And so we definitely have a bias towards keeping the data in its native database, which was built for purpose. So keep thermography in the thermal imaging database, keep Vibe in the Vibe database, and then send back out scalar summaries. It's very easy to do the analysis inside the built for purpose advanced suite and then centralize the presence of a flaw, a severity ranking on the flaw. and or some simple summary data. In vibration, it tends to be RMS. In thermography, it might be the max temperature of an individual pixel. In either case, the main message is keep the analysis in the purpose-built tool. Send scalar data out to some centralized system that could be maintain X, that could be ignition. You have multiple options. But we have a really strong bias towards that approach. We have not seen anyone succeed at putting all of the hardcore technical data into literally one database that's capable of analyzing. at all.
Zack Scriven: Awesome, thanks, Glenn. So I brought up this screen here from reactive to predictive. We touched on this throughout the webinar, but I'll just throw it up for the audience here as a cheat sheet if you guys want to screenshot this. From reactive to predictive, these are the six benefits. There's more. But these are the way you can think about the costs that you're going to reduce and what you're going to increase. you're going to reduce operating expenses. You're going to reduce capital expenses and energy consumption. Just the. existence of a predictive maintenance program, it's shown that it reduces your energy consumption. And then on the flip side, you have gains of revenue and production, you know, through machine availability, increase of human safety. I know Kyle, talked about that, Mark as well. And then environmental compliance. So these are the six benefits. This is what you should be thinking about when it comes to how to justify that implementation. And then there's also, you know, this big six predictive technologies, but we can leave that for another webinar.
Glenn Gardner: there.
Zack Scriven: I'm gonna stop sharing here. And I got that from the school of Glenn Gardner, so.
Glenn Gardner: Thank you, Zach.
Zack Scriven: Awesome, yeah, I guess we're a minute over. if that's it, we'll go ahead and close it out. Thank you all for joining today. And Glenn, Kyle, Mark, thank you guys for an awesome session.
Glenn Gardner: Yeah, thanks to our guests. Kyle, Mark, that was awesome. We appreciate it.
Mark Kingkade: Thank you.
Kyle Beers: Likewise, thanks guys.
Glenn Gardner: Thanks, team. Take care now.
Zack Scriven: All right, we're going to just stay online real quick. I'll go ahead and stop.