Kevin Clark with Falkonry

SUMMARY KEYWORDS

ai, data, ifs, kevin, learning, anomaly, industrial, asset, predictive, asset management, part, software, operational, criticality, companies, analysis, world, conference, 31st annual, drill

00:03

Welcome to the Industrial Talk podcast with Scott Mackenzie. Scott is a passionate industry professional dedicated to transferring cutting edge industry focused innovations and trends while highlighting the men and women who keep the world moving. So put on your hard hat, grab your work boots, and let's get there. I

00:21

once again, thank you very much for joining Industrial Talk. And thank you for your continued support of a platform that is dedicated to you industrial professionals all around the world, because you're bold, you are brave, you dare greatly. you innovate, you collaborate, you solve problems, and therefore you are making the world a better place. That's why we celebrate you on this particular podcast. Now we are broadcasting on site, the 31st annual SMRP conference here in Orlando. And it is a great event. It has wonderful individuals who are all focused on solving problems and making the world a better place. If you are interested in maintenance, asset management, reliability. SMRP.org is your first and only connection that you need to make sure that you can connect with individuals like the one we have here. How are you doing, man?

01:09

Doing good, Scott.

01:12

It's always good to see Kevin. Kevin's a nice guy. You don't have any real sharp edges around. You know, I try not to you succeed. Right. Right, Susie? Yeah, no, no sharp edges. No baggage. No baggage. No drama. I like it, man. Honestly, you having a good conference?

01:31

Good conference. Yeah, it's been well attended. Awesome. Attendees. It looks like the the number of practitioners is really high compared to

01:42

01:49

02:07

the adoption of, you know, reliability, asset management type of principles, increasing? Or? I mean, it just seems like there's a greater embrace of that solution?

02:17

There are those there are? And I think a lot. I think a lot of companies are trying to learn more today. I think technology becomes a big part of that. But I think we still have the some of the illusion out there that technology is going to fix broken process.

02:35

Yeah, sadly. But you're right about that. Yeah. I think they're looking for that silver bullet. I think that the some companies were kicked in the teeth as a result of COVID. Gosh, we are vulnerable. What do we do? How do we how do we create a business that's resilient? Hey, let's do some asset management stuff. How about we do some technology? And it's good and bad? I mean, at least it created the focus that is necessary.

02:59

Yeah, it gives us gives us something to work towards, you know, as we look at this technology, and it's advancing, and then we look at our competitors. And our competitors may be deploying software, I think it's giving some others drive to really go after it. Because as we've been talking software, it's becoming more and more of a competitive advantage.

03:21

Yeah. I've always struggled with, you know, there's a lot of noise out there. And then how do I differentiate one software to the other and it becomes somewhat commoditized. I just see the same thing, the same lingo, the lexicon is the same. You just, I just find it gets down to the individual and the human that that's the connection.

03:47

Yeah. And you'll see him jumping from software to software. And what I find interesting is when I see him jumping from a top five software to another top five software pretty quickly figured out it's really not the software. That's the problem. No, no, right off the

04:01

bat. Because you're right about that. All right. For the new listeners out there. Give us a little background on who Kevin is.

04:08

So I've been around a long time before

04:12

he's he's a little tired today. Yeah, a

04:15

little tired today. Poor guy voices a little tough. Lots of talking, presenting.

04:21

No, no sympathy on this side.

04:25

05:06

Segway time. Let's talk a little bit about AI. Because really because of falconry and and a lot of activity around that particular company. A lot of exciting activities around that company. Talk to us a little bit about AI. How's it changing? What? Talk to us about falconry? Yeah, make it happen, Captain.

05:25

Yeah. So I mean, it's, it's in the news, I just started with that. It's in the news that we're in the process of being acquired by ifs, we've announced the definitive agreements, and we're working on the close. So probably by the end of the year, we'll be looking at being wholly owned by ifs out of Sweden, powerful asset management company, well respected in Europe and beginning to move into North America. And that's why many people won't recognize ifs right away. But it's a it's a formidable asset management company. And faculty is going to play a big part in all of their platforms. So fuckery deployed out against every every platform at at ifs and then also be a standalone business unit at ifs.

06:15

What is the the AI and all of the stuff that's wrapped around AI? How's that impacting Asset Management? Maintenance? What tell us how that's going to work with ifs, I

06:25

can't do that without giving a little history. Yeah, please. So during my my years, I grew up doing critical criticality assessments, failure modes Effects and Criticality Analysis. And that, that we developed our our predictive programs, starting there, and that's, that's good process, it's good data that gets us to a job plan that's effective on your assets. The thing that we're dealing with now, is, when we do predictive, often times, we're doing it in isolation. And so we go out to what we'll put on sensors that that makes sense for an electric motor can identify temperature, vibration, accelerometers and other things that amperage, things that are going to matter for an electric motor failing. What AI does differently is it doesn't live by thresholds, it doesn't have a high and a low AI looks at all of the data that's coming in, typically, operational data. But also with predictive data coming in same types of things which could be vibration could be temperature, all of those together, AI looks at it, and it understands it from and learns it as a what is normal. And once it understands what's normal, then it can understand what is changing. So it watches for things that are not normal. What I like about capturing operational data, as opposed to just predictive data is operational data tells me exactly where I am, what products are running, what state status, the machine is in, what levels it's running. So I know everything about the operation, plus you throw in that predictive data, I now have a really good understanding of things that are beginning to fail in a way I didn't have before because I don't have the thresholds that I need, or that I had before in predictive now it's learning what's normal and watching for the abnormal.

08:36

So it's, what we're talking about is the ability to be able to take in all of these different data points. Ongoing real time. Right. And and the the platform, the AI solution, gives you that real holistic picture of your operation not done not just the asset, but the whole operation. Right. How does it display that? How does it? So how does it like me as a human going? Well,

09:04

you can display it a lot of a lot of ways, but we display it. In the most basic form, we display it as a heat map. So it's in color, right? So if you've got 100 signals that are coming through the API, it will show all 100 signals and it'll give you a heat map of what's failing or beginning to fail. So if yellow means I've got problems, red means I got big problems. Purple means I'm good. You couldn't go with green. Now no green. No green couldn't go with that.

09:35

I'd have to learn that one that's like purple is a combination of blue and red. Okay, so purple is good. So it's your classic dashboard, but not a dashboard as we know it but as a sort of add heat signature that

09:52

it's more like a heat so it looks like a heat signature, in essence, but it's more of an analysis tool. For an engineer for for a process engineer. For a technician, they can take a look quick look at that they can also be notified by the system that we have anomalies.

10:10

does give me the ability to drill down. So I'm taking this and I'm looking and I, whenever double click for lack of a better term, double click, and it drills down and gives me the where the little what's causing the anomaly.

10:21

Yeah, the anomaly is typically more than one signal. Right? So it's, it's usually a clump of signals. And it could be other things that are that are.

10:30

Give me Give me an example of another thing, because I get the vibration. It's vibrating. Not good. Right? Yeah, you know, and, well, in

10:37

often we have operational data, not predictive data. And that's a whole different conversation, where in most organizations, we find operational data located in one place, we find predictive data located someplace else, they're typically not together. Now, you know, the word on the street is, and I've heard this from numerous customers and people at events that they're working on centralized data repositories real time. And if that's the case, then AI is going to be so much more effective, because AI wants more data.

11:11

It's vital. Yeah, I don't, I don't see how anybody can effectively. I mean, you can, I'm not, I'm not saying that I'm a human can't do it. I'm not saying that. But I think there's a time element to it, the quicker I can do it,

11:25

I was just gonna give you the example of the other thing. The other thing might be, maybe you got amperage, draw on on one motor that's going up. But at the same time, you see, maybe a fluid level is low, on another. And so the two of those together create the anomaly. And it might mean something, it might be something you've already identified before. So it might be an unknown anomaly. And then there's things known as unknown anomalies, and those are the ones that we have to do deep analysis.

11:59

But in that same scenario, it might not be the condition of the asset, it just could be the condition of the operations or, you know, the feedstock not coming through or something something, you know, causing a, it's

12:14

that's why the conversation quickly goes got to, well can can AI detect quality? Yeah, yeah. Once again, of course, it can. But it all it's all dependent upon the learning and teaching by humans, on what those anomalies mean, as we as we, as we begin to identify more and more more. And the more data you give AI, the more it learns, the more normal makes sense to it.

12:41

How do you take let's say, I'm interested, and I'm interested in that that holistic view and I want, I want to be able to start begin pulling information in such a way that I can see the the overall operational health, not just asset, whatever operational health of my operations. Do you start incrementally? Or do you begin? How do you how do you how do you deploy? I mean, I don't even know where to start. Yeah. And then train it.

13:10

Yeah, right. It's data, right? So we can take historical data, we don't like to do that, because it really doesn't give you the best, but it helps teach AI quickly. So you can use historical data, and then it quickly learns what normal looks like. But we even if we're doing a pilot, Scott, we connect live data, we always recommend connecting live data. And once we're able to collect live data usually within it could be days, it could be weeks, depending on how fast the data is, if it's slower data is gonna take longer to learn what makes sense, right? But days, weeks, you're gonna have learned AI, when?

13:51

When do you know it's learned?

13:54

Well, it's always under journey, right? Yeah, of course. Right. That's, that's the point. And that's, that's the thing about about predictive and that's why AI is so powerful is because AI never forgets. And so you could go six months, you could go a year, you go two years. And finally, that one anomaly shows up that you've never seen before. And it might even be a pattern that it comes up every once once a year. And that could be a pattern, and it could be telling you something, but you never saw it before because you weren't monitoring the right things. But now with AI you're monitoring essentially everything. So it never stops learning. Just like with the human body. We never stopped getting sick. We never stop getting hurt. You know? So there's we're always if you were to attach AI to our body, it would always be finding anomalies, things that are not normal. So same thing with a machine it's always going to have something

14:55

but when when do you determine that okay, we've we've run it long enough. We think that it has has enough information. So it's, it's sentient now or whatever you want to call it. It has learned. So how do you know that and just sort of let it go? Or do you just like, yeah, it's running off? I know. Yeah. Yeah. Thumbs up, we're ready to roll.

15:17

Right? I'd be willing to say that we haven't seen one. That is I would call it the perfect model. Because they're all in the midst of learning. Because this is this is new technology. It's been around a long time. But it's newer in the application of drilling down to the anomaly detection on an asset. So that part is still fairly new. But I think it would be difficult to say there is one gold standard asset out there that you would use across all assets like it. Right, so once CNC were, you know, it's perfect. Ai? Can you take that against all other CNCS? I don't think it exists yet today.

16:03

Yeah, but But again, it's a journey. It's absolutely journey. Yeah. And and the, I'm assuming the value, is there. Real? The real gains are there.

16:18

Yeah. And you know, the thing about the thing about monitoring both process and asset data, is that it continues to get better. And it continues to give you more insights as you go, right. So if you're, if you're learning, if it's learning what the anomalies are, and you're helping to identify what eat and giving context to each of those anomalies, the next time it comes up, you know, more.

16:46

What's interesting would be the that moment in time, where you see red, and, and the drill down and you go, Oh, that's odd. Right, aha. Or it must be a real interesting, I don't know, reasons realization, just seems it

17:06

is yeah, the other thing, I like to tie it back to to criticality assessments and Failure Modes and Effects and Criticality Analysis that, that when we identify an anomaly, and we don't know what it is root cause we need to go back and identify it, right. And we have so much more depth of what went wrong, because this is time series AI. So we know exactly when things happened. And we know exactly where they happen. And we know exactly what sensors they were. And we also know exactly what happened to the product. There's just so much more that we know. So root cause analysis becomes much simpler, because all of that data is right there, and then captured. If we were just doing it on the SCADA system, Scott, it would be difficult, super difficult to go back and correlate all that data. Yeah. And bring them together. Yeah, yeah, that's what AI is to enforce.

18:01

And then in in that root cause analysis, that it continues to learn. Can you manipulate the model in such a way and say, Hey, here's something new. Can you? You know, infuse more stuff into it? Oh, yeah. Yeah, you can always build

18:19

18:34

You're wonderful. Like, you'd have no have no pointy parts.

18:37

Come on.

18:38

No, no, nobody. Nobody has ever come to me and said, Kevin's pointy. They've come to me and said, I'm corny. Not you. I'm gonna have to try harder. If you wouldn't, that's not your nature. How did they get ahold of you there? Kevin?

18:51

You can either get me at Kevin Clark get falconer.com. Okay. Or you can reach out

18:57

on LinkedIn. Right? Yeah, you can

18:58

get me on LinkedIn. I think it's Clark, KD but I just look for me now. You can find me. Oh,

19:05

have your stat card out there on Industrial Talk.com So if you're not we're going to have all the contact information for Kevin out on Industrial Talk.com We're broadcasting from SMRP 31st. Yeah, so 31st annual conference right here. 31. You were you were there back then. About 20 back? Yeah. Okay. All right. If you're in the world of asset management, maintenance and reliability or everything in between, you need to be a part of SMRP Go out to SMRP.org Find out more be a part of the conversation. We're gonna have another great chat. Just in a moment. So stay tuned. We will be right back.

19:43

You're listening to the Industrial Talk Podcast Network.

19:52

Always a great conversation with Kevin the legend. It's sort of rhymes. I'm gonna go with it. Kevin. The legend because he, he brings to the table tremendous mad skills in the world of asset management and maintenance and reliability. All of that he's got. He's, he is one that you must connect with to. Because here at Industrial Talk, we're all about collaborating. We're all about education. We're all about collaborating. And we're all about that innovation. And there's Kevin right there. Educating, collaborating, wants to collaborate, and definitely falconry and the innovation that's happening. They're a must. You got to look into it, just do it. That's your call to action. Easy peasy. Go out to Industrial Talk, all the contact information for Kevin out there, easy to connect with. No friction. We don't want friction. We want you to succeed. We want you to be a part of Industrial Talk Most definitely. So be bold, be brave dare greatly hanging out with Kevin changed the world. We're going to have another great conversation shortly. So stay tuned.