Cultural Adoption of AI PdM & the Power of Strategic Partnerships - a Panel Discussion - Part Two

Trend Detection Podcast ·

Show notes

Welcome to the Trend Detection podcast, brought to you by Senseye Predictive Maintenance – which gives you visibility and insights into all your assets, from single machines to full plants to help you reduce downtime, increase knowledge sharing and accelerate digital transformation across your organization.

In part one of this episode, you will learn about:

How Senseye works with your existing setup Discover why Senseye is vendor-agnostic, how it integrates with any PLC or database (even email-based CSVs!), and how it builds on your existing infrastructure—no rip-and-replace required.

The power of AI + human expertise Explore how AI doesn’t replace people—it augments them. Senseye automates routine monitoring so engineers can focus on solving the truly complex problems, with results easily shared from shop floor to boardroom.

Capturing tribal knowledge at scale Find out how Senseye uses large language models to capture, preserve, and share expertise from veteran engineers—ensuring valuable maintenance knowledge isn’t lost as teams evolve or retire.

Scaling predictive maintenance globally Hear how real customers like BlueScope Steel are monitoring over 15,000 machines in real-time—proving Senseye can deliver massive scale, fast ROI, and measurable business outcomes.

You can find out more about how Senseye Predictive Maintenance can reduce unplanned downtime and contribute towards improved sustainability within your manufacturing plants, by visiting: www.siemens.com/senseye-predictive-maintenance

Transcript

Welcome to the trend detection podcast powered by Siemens. The Sensi tool, obviously AI there's a broad array of different tools out there for general use, but Sensi's quite specific, so I guess, tell us a little bit about the journey from where you started and, and how you got there. It's been a fascinating journey, really. Uh, we started about 10 years ago, and, and for me it's crazy to think about that. I didn't have any gray hairs. Uh, I've seen a lot of changes since then. We started as the founders in aerospace and defense. And we actually worked on these systems um and we developed the analysis software to help the UK MOD monitor its fleet of helicopters, and, and the OEM to understand how can we start to deliver not just a, Piece of equipment and a maintenance package, but how, but now that our customer who's the RAF was saying, well, I don't want to buy that anymore. I want to buy power by the hour, which was coined by Rolls-Royce in the 60s. So this concept's actually been around in aerospace and defense for a tremendous amount of time. And with that one customer saying, hey, I don't want to buy this, buy it like this anymore, I want to buy flight hours, and it is now up to you as the OEM to give me that. And we saw that concept potentially making its way into the industrial space. To give you an idea of how much it cost to analyze these very complex machines, the helicopters are essentially a, a machine that, that hates itself, it just wants to vibrate itself apart all the time. Uh, having spent some time in that industry, I've not been on a helicopter, and I'm, I'm not tremendously interested in going on one, because they're, you know, they're difficult, uh. All of this data. For only about 300 machines, required about 50 million pounds of investment every year to analyze and understand. And so it's a whole building, literally, or a whole floor of expert reliability consultants, looking at these things, OK, what's the gearbox doing, how many metal shards are detected, who is flying it at that time, are they, are they a particularly bad user of these machines? And we saw this as an opportunity. Particularly with IOT and cloud processing, to take that concept, and to build a system that would automate the role of the reliability engineer, not to replace them, but to augment their expertise, so that you're not having to spend 50 million pounds per year for 300 machines, because when you take that to the industrial space, we, we all know that it's not scalable. So we want to develop this scalable system that can answer a problem right now of how do I improve the efficiency of my machinery. But also answers this question that we feel is coming still, and it's taking a little bit longer than we thought, of how does industry shift to instead of buying a machine and a maintenance package, how can we start to enable other organizations to deliver capability. Yes, so they don't need to have that team, that whole floor full of specialist engineers, they have a tool. No, and if you're selling something as a service, so maybe you're selling welds per hour, or maybe you're selling rivets per hour, you know, whatever it is, predictive maintenance is an essential component of that. You cannot do that as an OEM without a deep understanding of what is your machinery doing at the customer site, how are they using it, how are they treating it, how do we need to change ourselves as an organization to provide that service. Everything comes down to the service. And we wanted to build a technology that would be ready for that when that time comes. So to deploy a tool like Sensei, uh, do I need, does everything need to be in the one brand? Is it a Siemens solution that only works with Siemens products? What, how does that work? It's a good question. Um, so we were not originally Siemens, and maybe I skipped over that part a little bit. So we, we developed this technology and we were Sensii, and then in 20. 22, we were acquired by Siemens, and one of our core tenants has always been, we are agnostic, uh, because every organization has their own history of solutions that they've chosen and even within the same facility, they might have these different solutions. And it would be very ignorant of us to come in and say, no, you need to get rid of all of that, uh, and, and you can only buy from Siemens. Well, Instead, what we want to do is take this approach that, OK, you've made some sensible decisions, and you've invested a lot of money, we don't want to throw that away, we want to start with what you have and build up from there. And maybe what you want out of this doesn't match with what we need, but we will choose and help you select the best of breed, and that might not always be Siemens, and this is a sacrilegious thing to say, maybe, uh, maybe this is the last webcast I'll do. But, but the idea for us is respect the customer's decision, respect the investments that they have made, and leverage that to the maximum of our abilities. So The system pulls data from whatever. PLC, SCAA, whatever database you've got, we have vibration sensors, there's so many different ways, we had one customer, a very large multinational automotive organization, they sent us an email every hour and they had a CSV file. It's as basic as that, it can work like that, and then you build up on a more sophisticated integration. So it's time series data, it can be IOT essentially, as long as it's time series data that we can relate back to the condition of this machine, that's what we're interested in, how you deliver that to us, that's up to you. So you mentioned a large automotive business, what other industries is this being deployed in? Literally everything you can imagine from hydroelectric power plants to confectiony manufacturing to, To anything, um, we cover, we cover uranium mining. Like I said, chocolate bars, ice cream, cars, you, you name it, we probably look at it, and that's fascinating for me as well as a founder to say, like, my favorite part of this role is really, I get to go into these organizations and see how this stuff is made. One of my favorite, Programs as a, as a, as a, I say as a teenager, um, was the how it's Made series, and, and you go into the factory and these people talk with passion about something that they've invested their whole career and their lives in, and I find that fascinating. So still watch them on YouTube by the way, yeah, yeah, yeah, but but that's literally my life now. I go in and I get to meet these people, and I get to ask all kinds of awkward questions and, you know, try not to press the big red buttons that everywhere, um, it's quite compelling, but it's fascinating to learn about this stuff and then. To see how we can improve that and how working with partners we can integrate these systems, integrate more and more of this information and give, give our customers something really powerful where they can, they can spend time on the things that they don't want to spend time on. That, that's key to us. How can you focus on the things that are really interesting, that a human is good at, which is ultimately weird problems. Condition monitoring. is not a new problem. It's very well understood. It's a very structured discipline. Something I've noticed about Sensor that I really like is um a lot of the solutions I can go buy off the shelf today, I spoke to a specific industry or a specific use case, whereas Sens is not sort of agnostic and use case as well, you know, you pointed at the data, it doesn't have to be a pump or something like that, right? And people don't like hearing that people like hearing that, OK, my machine is super special, my industry is super special. Ultimately, actually, um, condition monitoring is condition monitoring. And whilst we focused predominantly on the heavy industries, automotive, er, pharmaceutical and food and beverage manufacturing, It can really apply to anything because machinery and how you deal with it and how you look at reliability is, is a core discipline, no matter which industry you look at. So let's talk about implementation then. So is this something I buy, plug it in, and it's working? How does it work? What's the rollout of the plug? So the ultimate dream, and when we founded it, we had this idea that a maintenance manager can come along and they can type in the credit card details, and here's my er database URL and it will just work magically. And that's kind of the vision that we still have in mind of that's where we'd like to get to, but the reality is that it's much more complex than that. And that's why it's imperative uh for us to rely on partners to perform that integration work, that integration, that connectivity, none of this works without that. And that's something that we learned very much at Sensai, that as a start-up, we didn't answer that question. As Siemens and with our partners, we can do that, and, and I'm tremendously excited about the potential because it solves our biggest problem. Yes, if I circle back, I talked about productivity, you know, we've got a. A bunch of manufacturers, they want to apply technology. They don't necessarily have a big powerful workforce. They don't have a lot of like they might have one or two skilled individuals. They've got new staff coming in. They're adding a shift. Like they're trying to scale. How do they scale that knowledge and skills, you know, a tool like Sensor allows them to sort of leverage those that existing knowledge, you know. So there's a bit of a change management piece there, and it's like you're saying to them like it's not here to replace you, it's here to help you. It's ready to add value to your business and like, you know, really. Make better use of your time, you know, and, and aerospace and defense term, uh, it's a force multiplier. It's, it's, you have this thing, and now rather than monitoring the 20 critical machines, you can monitor 2000, and, and that's a wonderful story. So if we think of, you know, we all know about the tools like chat GPT, Copilot, those sort of things, um, does it work in the same way as it conversational? Can you talk to your data in real, you know, like human language? So that's a chat GBT large language models has been wonderful for us in that it answers, or, or it, it's a missing part of the puzzle that we always needed. But our technology and our AI is, we, you know, we started this 10 years ago when that kind of stuff just wasn't available. And so we spent a lot of time understanding how do we codify condition monitoring, how do we understand the metadata. We developed a lot, a lot of patterns, a lot of technologies specifically around, well, how do we teach a machine to understand what machine data looks like and try to do that automatically, because that's the only way that you can achieve scale. Part of our issue then was how do we communicate this to humans in a really, Easy way, and that's a massive challenge, one we've always struggled with, and we kind of deferred that solving that problem, till later down the line, we thought, well, we're, we're gonna have to tackle this problem, we have some scripted messaging that we can put in and we can make it smart, but it was still very difficult, and then, Large language models came along, and we fully integrated that, and we are the first Siemens product, actually, we released it in February 2024. That sounds awkward, isn't it? We've released it in February 2024, and, and we've passed all of Siemens legal and commercial and all of that, to have a fully integrated co-pilot that's aware of the time series data, it's aware of the maintenance manuals that you put in, it's aware of the contextual information that you're bringing in from your maintenance management system. And that is so powerful. It's incredible. I'll just pick up on what you said there. So it's connecting with the manual. So your machine or whatever you're monitoring, it's actually taking data from that and, and providing it to the user. Yeah, our philosophy is if the AI has access to it, give it to the user. Uh, my one of my co-founders, Rob, his, his, his pitch was always, we measure your success by how little time you spend in the application. We don't tend to say that anymore because it's often misunderstood, you know, we don't want you using it. No, we do want you using it, but we don't want you coming in trawling for information and trawling for value. If the AI knows it, tell you, it's not a chore, it's just easy, yeah, feeds the day to make your, make your life easier. So from a personal experience, I've been looking at AI tools as a uh a cheat code, a shortcut, you know, a tool to make things faster and more efficient, um, but that also brings challenges because there are, there's inherent fear with it. There's people don't understand which tools do you use. So I guess my question is, is, uh, what, what are the barriers to implementing something like Senso. on the like on the on the implementation and the data connectivity side. It's a journey, right? It's a journey like you said, you don't just switch it on and walk away. It's about, you know, bringing people along with you, sort of explaining the value to them, and then making, making them part of the journey. So they're really like engaged from the beginning and all the way through. So it's like, you know, the right support and training and, and also use case examples and teaching them how to use it, supporting them as they go. And then making them like part of training it, you know, because what I don't know what you picked up in there, but when, when you have a anomaly detected and you solve the problem, you tell Si how you solved it, so that gets fed into the model for next time to tell the, tell the person down the track. So you help them with how to give good information into the system. So we, we're working with them person to person to really get the most out of the tool. You said something key actually, it's about people. And we talk about condition monitoring, we talk about machines, we talk about machine data and condition monitoring, but ultimately, all of this technology and AI is about saving time and effort for people. Do you, as an expert reliability engineer, want to focus on vibration monitoring and the 123 times harmonics and trawling around in some old school system to look at sky lines? Our argument is probably not. The AI can do that all of the time, 24/7 on tens of thousands of machines within a single account, and it can tell you which ones you should be paying attention to. And out of those thousands, it might just be 2 or 3, and you can use your expertise then to really zero in on that. You see that scale, that productivity we're talking about. So how, you know, what's the measure of success? How do you know it's working? You, you talked about feedback. So, so the tool itself, what are we talking, return on investment, and that's part of the story, right? Yeah, yeah, and that's part of the journey that you have to take these customers on of what actually matters to you. Are you producing milk? Are you digging up dirt? Are you producing cars? How do you, we, and this is a conversation, and this is where the human element comes back in. How do you measure success and how can we set up the AI? So that it's as easy for you as possible to measure that success and capture it. So, OK, we found something cool on a, on a gearbox, and it was slightly cracked. It's OK to say, look, we uh, we fixed that, wonderful, and that's where we see most of the market looking at. But actually what's more, much more interesting is, what does that mean to management? What does that mean to the board level? How do you integrate that, that data so that you can say, yeah, I fixed the crack on the gearbox, and this is how we try to talk to reliability engineers as well, of, we're not coming for your job, we're making sure that you guys are displayed as the hero. You use, to find a crack in the gearbox, and that meant that you saved 10,000 L of water, you saved 5000 tons of carbon emissions. You also saved a $1 million contractual penalty penalty on, on emissions, and we see this all day every day, where we can say, look, somebody used Sensai. They fixed the problem, they added the KPI they added the work event, which, you know, helped Sensie to learn and build from that, and now your name as the maintenance reliability Eng engineer. Is associated with that big capture where you've saved a contractual penalty of a million dollars, and previously, that was a pretty hard story to convince, but as sensei, it's about bringing that human element into it. And, and I'll add to that, like if you've got, you know, you've got your, uh, engineer who's been there in a lifetime, you know, he knows his plant or her plant, but they might be misunderstood. They might not get that point across very often. But using the data, using the SCSI results and the calculation of the ROI, the return on investment, suddenly they're a hero, you know, um, and they can share that knowledge with the rest of the team, and they can really, um, make a difference to the business. And they are heroes. Like a lot of the, a lot of the difficulty we have is people get nervous about AI that they think, oh, it's gonna come and take my job. And, and maintainers are in a really difficult position. If they do their job well, then there won't be any breakdowns. Nobody notices, and then they say, oh we don't need maintenance, we can get rid of that. That's the first thing to get deleted from the accountancy budget. But if they're, you know, struggling, which they often are, because it's underfunded, because people say, oh, we're spending too much on maintenance, we don't need that. Then bad things happen, and then they get screamed at, so they've, they are under pressure from all sides, and a lot of the work that we do and a lot of the process and the methodologies and and the way that we work is to try to say, OK, you did this thing, you, you fixed the problem, and now let's automate how we reflect that into your organization, so that you as the hero can really easily be identified, and maybe you're not so good at taking your, uh, you know, your, your capture and your win, your fix, and relating that to senior management, but we can automate that. I just want to pick you up, Nick, on, on a comment you made. You talked about an engineer that's been there a million years, uh, and really that, that points to an issue that everyone's having at the moment, and that is a lot of skilled workers are getting to that age and stage and they're starting to retire and you're losing that. So starting to retire, but also, like I was talking about that productivity. If you add a shift, how do you duplicate that person, you know, or they go on holiday, you know. It's not just about when they retire. It's, it's all facets of the business, you know. So having these tools to help gather the information, intelligence, and the, and the processes, and I, I, I think there's always a challenge there to get some of those guys to let go of that information. But, you know, generally, they're pretty good at that. They love having the apprentice along. So it's like, it's like a tool for the apprentice. So you've got to explain it that way to them. Now they're helping the young guy that's coming up. And, um, as soon as you explain that, like, OK, got it. So obviously since I was ingesting a lot of data, vibration, all those different, um. Uh, feeds of information, time series data, but how does it capture the knowledge of somebody who's just got it up here and knows the machine and That's great. I mean, so yeah, so that's about that's again, I guess it's up to us engaging with them to help them put really good data back into to feed it, what, how to write, how to write the use case up, you know, what did they do? What …

Full transcript (WebVTT)