Torsten Seehaus with Fluke Reliability

SUMMARY KEYWORDS

customers, vibration, model, data, machine, fluke, equipment, torsten, industrial, automated, wrong, bearing, future, aAzima, maintenance, predict, measure, failures, type, industry

00:04

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 go.

00:21

01:25

thank you, Scott. Again, again. Yes. So my name is Torsten Seehaus I have been with fluke, for the past 10, maybe 15 years and in the industry held several roles in product management and sales across multiple different product lines and industries, which were taken care of. And then we acquired Azima. DLI, six months ago. And with the acquisition, I stepped in, in the role of leading this business full time. So I've been doing this since August last year. It's

02:03

an exciting time. I love that product. What does dry stand for?

02:08

02:18

I just use Azima is easier. Yeah. Okay. So take us through what Azima is

02:26

the same as a business with a focus on measuring vibration to provide customers insights on when something could go wrong with their machines. They've been doing this for probably 60 years now started out with the US Navy on some of the aircraft carriers. Back then everything was very manual, people were on board 3d measurements, then trying to get the data on a piece of paper and analyzing data, providing a report back to the customer to understand what should be done with my break in the future, to expand the business into commercial applications. So manufacturing mode, first and foremost, lots of applications, food and beverage, oil and gas refineries, multiple different industries. And there, they've been really successful in expanding the business. So now they have a lot of customers relying on a Azima to provide an insights early on what could go wrong, so they don't have to wait until something breaks or just have maintenance and night might not even be necessary. So making sure customers are doing something at a moment in time when it's important.

03:36

Yeah, sort of paint that picture that use case that Azima fills, it's like it has a data lake, it has substantial information that is embedded in it. So take us through what that means.

03:52

So think about a data lake. So it's, it's data, which is collected over decades, essentially, from the multitude of different machines. So most likely 80%. And we have machines out there provided data to Azima the types of machines and every machine out there, but the types of machines which are out there. So that means that with all the data they have on what could go wrong vibration data on, there's pretty good insight on predicting what a similar machine might be doing at a different side of different customers. So they can draw on all this knowledge they gained over time to make sure that when they provide insights to customers, similar machines, they are mostly accurate. So you want to avoid these, what's called false false positives and false negatives. So you don't want to do something when there is no need to we don't want any bearing on an expensive machine if there is no need. But you also want to make sure you do it if it's required.

04:51

The the information the data that exists within Azima it was accumulate added over a many years, right? Is there data in there that is just a target and say, Hey, we need that information to be a part as NEMA two as well, we're just going to go and put that in there. Does that exist to?

05:14

Not yet, it's an interesting thought model for the future, because operation is only one element of how you can measure and predict something could go wrong. There are other measurements out there, as Azima has been squarely focused on vibration, and nothing else. But of course, there's more we'd like to add in the future. But right now, it's the data lake itself is vibration, megawatts.

05:33

But the people's understanding of AI is clear. It's like, okay, I feed this information into this AI model, and then it sort of tells me information, but this is a different sort of application, it does the same. However, it it still requires that human intervention.

05:58

It does. So there's, you're you're throwing something into the model, in this case data. And you get a response out of the model, which is saying, based on what I'm hearing, and what I'm what I'm seeing, most likely, you will have an issue with some, some bearing or some other failure which might occur. The AI part comes in when everything is automated, so many of the failures can be automated already. So around 60 to 70% of the failures don't need any human intervention to tell the end user something is wrong. There, whatever it's, let's say severe failures, where you want to have a human analyst looking at over Yes, it might be an expensive machine, don't just want to react to what the model tells you. And the future may be we, our goal is to increase automation level. So you want to maybe automate 90% of possible, but most customers value the, the ability of having someone to speak to to look at the data again, before they make some very significant and expensive maintenance decisions.

07:06

So the the user interface Azima, let's, I'm gonna, I'm gonna paint this picture. I go out there with a diagnostic tool, I measure the vibration on a rotating piece of equipment, it records that vibration for a period of time, I take that data, and then I can just define what the rotating piece of equipment is. Here is the results of my test. And that's where really does its magic. So how it works.

07:39

Yeah, almost the one of the inputs, which is required initially is telling the model, what kind of machine are you measuring. So it has a setup phase initially. So when you think about an installation, you will always go through the entire setup. And you will tell the model, what I'm measuring here belongs to a certain type of machines, a fan, it's a motor, it's a compressor. So the model already knows exactly what type of machine is being measured. And it's pulling data from the data lake, which have a similar kind of machine type.

08:16

Does it get down to the level of specific models and nameplates? And it's like, okay, this is a, you know, Model X, Moto y, whatever, and then be able to sort of let the magic happen.

08:33

Yes, that's, that's exactly what it does. Correct.

08:37

And then once it's in there, right, it's, then then you can have the luxury of automating it. So you don't have to set it back up, it's already in there, it knows that the information that I'm pulling off it, it could be a device, pulling off the vibration details, feed into it into the system, and then being able to have that sort of notification that I'm a vibration technician. And I'm going to say oh, it's telling me that I got to look at this or there's something to be said about this asset.

09:10

That's, that's correct. So your wallet once it's set up, and we have multiple customers who have been using the system for 10 years or more, once it's set up, you never go back we changed the machine do something else but it's the machine stays the same when a piece of equipment stays the same, you will never tell the model again what type of equipment the model is looking at. It will always use the information what has been recorded initially. And then it will measure vibration and it will not only give the customer information about something might go wrong, it will tell the customer exactly what will go wrong and when. So we'll say your bearing needs to be changed in 28 days from now. So customers have more time to plan maintenance, but it will give you a specific except on what action has to be what actually needs to be done, not only that something is wrong, and then go check it out. It's actually giving customers insights immediately saying something's wrong. And exactly this is what's wrong. And this is what you need to do.

10:12

So it gets into that detail of, like, I'm looking at this vibration result. And it can determine it appears to be this particular problem and give that sort of indication. And in that indication of the problem, I'm able to say, okay, so I need a bearing I need whatever it might be, and I can prepare for the maintenance of it. It might be critical, where you better get on it now or yesterday, versus something that is predicted in in a month.

10:43

Exactly. Well, we we want to avoid on the Azima. Side is you have to do it now. I mean, this is this is what customers did like Don't lie. You don't want to do that. Yes, correct. So customers don't like it. They don't want to be surprised if any of that right. What I would like to hear is and sometimes we can predict the future like next year. So we see something which is initially starting to occur can Yes. And based on the data lake, we know exactly how long it took for water types of equipment to get to the same stage. So you, you want to have a prediction, which is telling the customer we see something, you don't have to act on it, it's still working. Right. But most likely, in a year from now, nine months from now. There's something to be done. So if

11:28

you haven't seen it, fortunately for me, I didn't catch it on camera, but my lights are flying outward, because the the doors are open outside and they're blowing in so but you didn't see it, but I had to tell you about it. Where do you see it going? Where do you see it evolving? Like what is it? I know you got to get it into? I mean, there's got to be more customers get it get more familiar with it? Where do you see it going.

11:55

And that's exactly what we see right now it's getting more familiar, customers are more reacting to issues like I have to do it now. Or we have time based maintenance where they are saying, I'm going to change something in six months from now I always do it every six months, which might not be even necessary. So this is a big market rate, just getting these customers to the next level and making sure they they use predictive maintenance to their benefit because of the voice downtime, it saves maintenance costs. So there are lots of benefits. Well, we can see here the conference right now as we have multiple different opportunities for growth and for developing future products. So that make it even easier. For people who are the frontline who do the maintenance, they might not necessarily like to look at vibration data. So right now we are giving a

12:49

rare breed, just FYI.

12:54

Correct. So these people, now we give them not only inside what needs to be done, but also when needs to be done. And what's happening. So now, but now they still get the information, but now they have to act on it. So that means they might use their CMS system, their work order management system, someone has to type something in there for someone to do something. So it is again, an automated piece. And now you've put a human in between, again, manual steps, their failures, the work that has to be executed, how do we know it has been executed? We don't know yet. So there's a link between what we can do now from a reparation perspective, sending a work orders when they need to be sent out. And checking back to work a lot has been done checking didn't really fix the issue. So the model can learn from itself was my prediction again. Correct? Did that solve the issue? Yes or no. And so we can actually adapt the model again. So you have more feedback loops in the future as well

13:54

see that that's that's that's pretty interesting, because I think the the real along with a lot of things, but I think the real point of is is being able to know how to plan your work proactively at the right time with the wired equipment with everything in it just happens you don't get you don't get into this, this backlog you know misery that so many companies are in it happens it generates the work order at the right time for the right piece of equipment boom done. And your your your directing your work in such a way in the people who do the work more effectively.

14:37

You do people and also inventory right if you have Yeah, lots of customer, multiple sites, you don't have spare parts and all the locations for all the equipment you might have to service some customers all across several regions and countries and so it's even further away. So how do you get the right spare parts in place at the right time? Seattle things so you might actually be able to log in inventory costs.

15:02

I love it. You were you were fantastic. And you were super, super flexible. Thank you. Did you like the award ceremony?

15:11

I did. Yeah. It was really great. So seeing all these customers up there getting their awards? Yeah, definitely.

15:16

How do people get a hold of you out on LinkedIn?

15:18

I am on LinkedIn. Yes, you can find me out there. And of course, LinkedIn, I think the

15:23

15:49

You're listening to the Industrial Talk Podcast Network.

16:00

Right, his name is Torsen. Fluke reliability is the company Xcelerate 24 was the event. And the solution is Azima. Cool stuff. That conversation was chock full of great information. You gotta love it. You gotta just love the fact that we have this platform to be able to highlight these incredible solutions to help you succeed. Vibration AI, what can you try to throw something out that and that is just doggone cool. Such as Azima. Check it out, go out to Fluke reliability, and find out more and reach out to of course, you won't be disappointed. Building a platform, you have a podcast, put it on Industrial Talk, you have a solution. Let's highlight it on Industrial Talk. You need to you need to succeed. That's what this platform is all about. Be bold, be brave, daring, greatly hanging out with porcelain and you will change the world. We're gonna have another great conversation short