From the Field: What Customers Really Want in Predictive Maintenance - with Tom Jacques

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.

This episode covers:

Cut Through the Noise: Discover the true, underlying challenges customers face in maintaining their assets, moving beyond buzzwords to focus on what genuinely impacts production and downtime.

Experience Rapid Intelligence: Learn how Senseye's software can "learn" an asset's normal operation in just 2-5 days, providing insights that might seem too good to be true, and how customer feedback directly shapes its evolution.

Debunk Predictive Maintenance Myths: Understand that predictive maintenance isn't about eliminating all failures or replacing human expertise, but about optimizing maintenance timing and acting as a powerful decision-support tool.

Witness Industry 4.0 in Action: Hear how digital transformation is delivering real, measurable success for companies across various sectors, making advanced asset monitoring more accessible and impactful than ever before.

Strategize for Maximum Impact: Gain insight into how identifying critical pain points and understanding the ripple effects of asset failure are key to selecting the right machines for monitoring, ensuring significant return on investment.

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

One final point to your question is we've got far more success cases now than we had 234 years ago because of how many different types of assets we monitor, how many different types of industry we're now working in, um, which is still, you know, it's, it's been many for a long time, but now we start to, to really, um, really get deeply involved in, in different types of assets. Welcome to the trend detection podcast powered by Siemens, and now your host, Nel Sullivan. Hello and welcome to the latest episode of the Trend detection podcast. Um, I'm really pleased to welcome back Tom to the podcast, who's been on, I don't know how many it's been now, Tom, but, um, it's a pleasure to have you back. Welcome back. How are you, how are you doing? Thank you, Noel. Yeah, pleasure to be here. I think this is probably our 6th or 7th, something along, along that line. Yeah, and this one's gonna be the most memorable. I can, I can feel it. Exactly. I mean, today, today, yeah, today's a really cool, um, topic actually, um, and it's, it's really at the heart of the podcast actually, cos my, my, um, when we launched this podcast however many years ago, it's all about sort of capturing the insights of all fantastic colleagues within Vince Sensei, now Siemens, um. You know, you guys have all the knowledge, you speak to customers every day, all these different conversations, and it's basically, I want to share some of that insight, or, a lot of that insight with the world, um, and we do that anyway across all of our episodes, hopefully, but, um, for this episode, we're, we're sort of digging into that a lot more. We're actually thinking exactly what is affecting the customer in the field, what are their thought process, what are their, their challenges, and it should always be around the customer as well, um, of course, um. So yeah, so I guess to open the conversation, um, let's start with those common challenges customers face. What, what are customers saying to you at the moment, Tom? What, what are the big challenges they're facing? Well, this is an interesting question because, you know, in terms of what my job is, it's really to break down for customers what it means, how it can impact your maintenance, how it can impact your production. And then to talk about this now on a podcast is brilliant because it helps me get, get the stage to talk about how many buzzwords there are, how many different confusing elements there are to this, but actually at the crux of it, it's, is the machine doing what the machine should be doing? Is the process outputting what it should be? Are you getting lots of downtime or lots of over maintenance issues? I think quite often the challenge becomes you get too wrapped up in, in, in the complexities of it and you lose sight of, of what you're actually trying to achieve. So the first thing I would typically say is that you, you just look at the data you've got, you look at what, what gaps you've got and then really try and understand what exactly are you trying to achieve and that's the first step in any part of these processes. But in I guess in terms of the challenges though, what, what are sort of customers really, you know, saying to you, you know, when you first have these initial sort of conversations, what, what are the things that are really sort of, you know, I think the first, the first challenge to, to that point, the first challenge is often that can't possibly be real, you know, that you've got a piece of software that can learn my asset in 2 to 5 days normal operation. It can basically understand what the asset should do effectively on its own with a few prompts and the ability to, to have parameters set at the right sensitivities and so forth. And then from there, it, it has user input. Every time something gets changed or something gets fixed, it will incorporate that piece of information. It, it all sounds too good to be true. I think that challenge, the, the ability to understand that and see it for yourself is, is the biggest hurdle that we face. I think selling a product that sounds like I mentioned it too good to be true is probably the biggest difficulty I think to, to what we do. And, uh, and it's, yeah, and it's, it's interesting. I mean, we always, we always talk about on here about, um, you know, the dialogue with the customer and it's, you know, you know, you look at sort of vendor customer relationships, so it really is a partnership. I know it can be a bit of a cliche sometimes as well. But I mean, I guess you've seen, uh, have you seen examples of where the feedback for customers has been sort of, you know, directly influenced Sensii, I think we could probably talk about specifically today, right? Yeah, definitely. I think when you've got different graphics, different methods of showing that data, um, different interfaces that the customer wants to see, you know, you could have the smartest algorithms in the world, but if you can't then interpret that information in an interesting way or a simple way. You know, having a traffic light system to start with is, is something that customers love having, so you can see, is it in a status of red or orange or yellow. That's very simple. It's something that we use pretty much unanimously across industry for different software, so something that aligns with that is standard, but then having reports is a pretty new feature of Sensei, so all of this stuff can be shown to you in a report, you can see different areas of the plant, different types of machine, different fleets of assets, all sort of bundled together in whatever kind of report you want, plus cases, open cases, closed cases, um, you can chat to copilot. All these things have been driven by what customers want to see, so this is the direct feedback that you mention. And I think that's crucial cos it shows customers that we'll work together on this. um, you know, typically the way many of our customers would shout in a certain direction. That's the direction that we would then engineer the product. So that's how SensorR has evolved into something that's very user-friendly, very, um, well suited to industrial environments, and that could be in pharmaceuticals, it could be in steel. These environments are different, but obviously maintenance at the core is, is effectively the same process. Is this part degrading? Does it need to be changed? Has it been changed? And then making sure as well that that integrates with existing workflows, so that's another key part of, of what Sensi is very good at. And I guess, again bringing it back to your specific role, Tom, I guess how do you sort of interface with um with, with the product team and making sure you know, some of these communicate, you know, some of these, some of these feedback's fed between both um both sort of departments. So one of the latest features we've had is the ability, I say latest, it's been around for months already, is to have, have spectra be visible in Sensei. You can, you can make very good use of data that's come to you. At 10,000 Hz frequency or, or 46,000 Hz frequency, whatever it might be, depending on sensors. We can, we can ingest that, we can do very clever things with it, you can see the FFT of that, so the Fast Fourier transform. This is something that customers wanted. Oftentimes you have vibration analysis that's done as a, a manual measurement, maybe it's a monthly check, someone will come in and do that inspection, they'll come and, and, and kind of do all these measurements and leave, leave and come back to you with a report of which bearings are failing, which shafts are starting to oscillate more than they should and effectively give you a, a strategy, a plan that you should now follow. Now one of the things customers wanted was to have that in the same piece of software that's telling them of tank level of, you know, motor vibration, uh you know, generically from a, you know, from a memSense or something like that. And now the ability to, to pull that back in, and my role then is to feed that right the way through, but customers want to see more of that, they want us to develop that tool, um, and this is really where, where we're going because if you can start to take in manual measurements, things like oil samples and all the rest of it, which are from complex sensors online that would typically have been offline. You can start to completely remove manual checks that are, are to do, you know, oil contamination or to do, you know, things like spectra analysis and feed all of that stuff in, which is, which is complex information into a, into a, a piece of software effectively. And to do that is, is really difficult. And to build a product around all of that stuff is also quite difficult because it's lots and lots of different variables all culminating together, as we've said in the past, plus PDF documents and so forth. So, I, I think quite often I, I end up being the voice of the customer because it's my role to explain the latest conversations we've had, questions about can the software do X, Y, and Z, and just making sure that comes back to the, to the team developing the product. And that's part of the, part of the way we've developed such a good product because. It shows us what the industry wants to have, what it, what it wishes it could do already, and, and that's the direction we've gone in, you know. So in a very long answer to your question, it's, it's important that we capture those feedbacks, so we, we don't just spend 6 months developing something and then deploy it. I think, I don't know how many years ago that would have been, but maybe 5, 10 years ago, that was the method. And there's very few gaps within that for a feedback cycle. So the goal now is to constantly keep understanding, are we going in the right direction? Are we, are we iterating this in the direction that customers want us to? Is it, is it a big customer or a small customer or a, uh, you know, generically across the board, many, many customers want the same feature or functionality. And you have to listen to customers, I think that's really, really important and we do that really well. And, yeah, and of course, yeah, I completely agree. I mean, listen, listening to customers is key. Um, I guess 11 other way to look at it as well though is when you've got lots of customers asking for lots of different functionality. I know it's more of a product team, you know, prioritization in terms of the roadmap and etc. but I guess from in, again, specifically to your role, Tom, how you sort of manage expectations and find, I mean, the, it's in your title, right? Solutions engineering, how you find solutions to things that. Might not be possible now in the app, but it could be in the future or it might not be possible in the future because other priorities, etc. Yeah, and I think we, we've definitely touched on that in the past, and I think that's really the crucial part of it. No, I think that's a great question. So what we typically would say is, is common is things like MTVF, so meantime between failures. But I think more and more as we go on now we start to see customers looking at things like unplanned versus planned maintenance ratios, how much time do they spend maybe fixing things that shouldn't be breaking down, so very reactive or or kind of doing too much. And then another emerging one is how effective are technicians and maintenance engineers when things break down. So, I think some of, some of the unexpected metrics that, that are starting to emerge are things like, OK, it's taken us this long to determine there's a problem because Sensize helped us really speed that process up to, to troubleshooting a problem. And then actually how fast can we, can we understand what's going wrong. So you don't just have a, an increased, you know, um an improved process to understand what's going wrong and when and capturing it before you otherwise would have, but because of all the knowledge capture, you're now able to determine much faster uh or much more quickly what's actually going wrong. Uh, and what you need to do to fix it. So if you needed to get a spare part ordered and, and into stock, that's what you can now do. Um, and again this goes down the road of, you know, that sounds almost too good to be true. I think we've, we've talked about that a lot in the past. No, exactly, yeah, um, a victim of its own success in a, in a way, right? Um, and that, and, yeah, and I mean, now we've got, you know, look, we've got so many great, um, public case studies. I mean, Bluescope Steel, I always mention Bluescope Steel, um, Saxon Mills in Germany as well. Check those out on, on the website. They are, they really are sort of fascinating stories, great examples of, of what, what's actually possible, um. But I mean, I mean, you, you mentioned there are a lot of different things, a lot of different areas that predictive maintenance can influence, but is there often sort of misconceptions about what can be achieved for predictive maintenance, or, but it could be in terms of what could be achieved or the timescales in which it can be achieved. Is there, Still varying expectations on that level, would you say? I would, I would love to say that we're very much at the point of plug and play, and you could just, you could just set some sensors up or plug some data in and it, it all just works effectively on its own. Now because of the the the nuance, the different asset types, the different industries, I mean you know how many different types of customers we're working with. And there's, there's differences across the board. There's different types of machine, there's different data frequencies and all the rest of it, so. We have to first of all put a project together that can factor that stuff in, and I think customers more often than not are happy to work with us on that scope phase because at the end of it, you end up with a project very quickly, you know, within just a few weeks if we can get on site and have a look at the machines or, or someone else can take photographs and really explain to us what's going wrong, how the machines work. As soon as we understand what the machines do, what sensors are, are in uh in position, what data is available, we can build a project that's about as close to plug and play as you can get. And then from there we can deploy the project, set it, um, running and do the operate and refine part of our process to really work together, um. I think some of, some of the misconceptions might also be that, that PDM eliminates all failure. Now I would always say until we remove friction, which is, as far as I'm concerned, impossible, we're always gonna get mechanical degradation, so bearings are pretty much the closest thing we get to frictionless. But still within that, they break down, they degrade, they get hot. You have to keep them lubricated, you know, and there's also debris and all the rest of it that gets in there. They can be overloaded, they can be, uh, warped, whatever it might be, these things will degrade. If you look after them, they last a lot longer. But they're still gonna have a, a lifetime, they are still gonna need replacement at some stage now. If the bearing is still fine, every 6 months you don't need to change it, and in some cases, if you need to open something, get inside, uh, allow debris and, and other contaminants to get in, you, you're actually making the problem worse. So having a system that's got sensors and it tells you, you know, at exactly the right time, the bearing now is coming towards end of life, maybe based on all telemetries and what we know about this bearing, maybe you've got another month left. You can start to plan all of your operations around that. And I think Uh, maybe a misconception is that that's now gonna, that's never gonna be needed. What we're doing is we're optimizing the moment when it's needed. So it, it, it comes down to really forecasting and, and predicting. That's the name of the product, right? So it's predicting effectively when something's gonna be needed, and that's based on anomalies and trends and other metrics from the system. Um, I always tell customers it's not about predicting everything, it's about reducing that surprise element. Exactly, and I, and I guess the other, the other layer I'd maybe add into it as well, because in it, especially when you talk, talk about AI and, and gen AI as well, it's, it's also, it's a, well, and decision support tool is another phrase that, that often comes up and it's a good way to sort of describe things. It's not, you know, the AI is not replacing anyone, it's not being. You know, replace this this exact moment and takes the human decision out of it. It, it's very important to emphasize the human element of this. It's advising humans to decide themselves what they, what action they need to take, right? That's, that's what we're saying here as well. That's, that's another good misconception. No, I, I completely agree. And how often do we see it where if it's got AI effectively it can take over and do all these things? Maybe one day, maybe, but I'm happy at the minute to say we're not at that stage just yet. And where we currently are is having a very, very clever piece of software, which I think is ahead of the curve. It's, it's definitely far, far removed from a condition monitoring piece of software that tells you, you know, you've reached 80%, therefore equals alert. It's a, it's a very incredibly clever system. Uh, you know, I love working for Sentenceive for that reason because this product is so, so versatile and it's got the ability to continuously learn stuff, so. Maybe, maybe some of the, the misconceptions are that it's fixed, and that product is then just set up for that machine, but it's, it's not the case, it, it learns and develops and if the if the asset changes or if you replace, If you, if you change the bearings for example, from a, a, a rubbish quality bearing to better quality, you're gonna expect different, um, patterns of vibration, different harmonic signatures, and also a very different lifespan, that bearing should last you quite a lot longer. And that's where as well you wanna make sure customers understand, you know, that's the same, um, bearing housing, it's the same shaft, it's the same machine, but we're gonna expect different results from that. So we can, we can tell the system to expect, uh, a different outcome, to, to perhaps relearn something or to, um, factor in spin speed of the shaft in addition to temperatures and so forth. So. It's a very, very clever system, so making sure that customers fully understand, um, how, how intelligent it is and how we also don't have to have a knowledge of that machine prior to, to monitoring it. Yeah, uh, often we get asked, you know, I've got a big bucket of data, do you want to see that first? It's helpful, maybe in some cases we see particular high spikes or, or deviations away from baseline that could be indicative of a failure, that's useful. But equally, we don't necessarily need that, we can, we can monitor a gearbox for about a week, let's say, with a few variables if it, if it has them, different products might be produced on that production line, we can factor that in as context, but then from there we see, we see a, a, a baseline, we see a, a fingerprint is what we call it, and that's something that I think is sometimes a misconception, you know, how can you possibly have a model for. Um, all of my 100 different asset types. Um, if you need to, you can speak to, like you mentioned, Saxon Mill, Blue Scope Steel, cust customers like this who have many, many, many different asset types. And the same is true in, in pharmaceutical and automotive and food and beverage, and, uh, it's nice now. I think we have a lot of customers with, with the, the, the use cases, with the case studies that we've got. …

Full transcript (WebVTT)