How to start a predictive maintenance project- with Tom Jacques
Show notes
Welcome to the Trend Detection podcast, brought to you by Senseye Predictive Maintenance – the platform which enables predictive maintenance at scale across all of your assets, across all of your plants.
What you’ll learn in this episode of Trend Detection:
Why starting with the business problem—not the tech—is crucial for predictive maintenance success
The key differences between predictive maintenance and condition monitoring—and why it matters
How to build a scalable PdM roadmap by starting small, proving value, and selecting the right assets
Common misconceptions and roadblocks that can derail projects before they begin—and how to avoid them
Why cross-functional collaboration and on-site insight make all the difference when launching a PdM initiative
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
It's really important to see the assets. It's also really important to meet the team and learn a little bit more about the culture. So, is, is this a proactive team? Is this a team who really just runs things to failure? Because that, again, has an impact on the business objective. Welcome to the trend detection podcast powered by Siemens, and now your host Nel Sullivan. Hello and welcome to the latest episode of the Trends section podcast. I'm really pleased to be joined once again by my colleague Tom, um, he's fast becoming one of our most regular guests, so welcome back, Tom. Hello Noel. Good to see you and hear you again, um, and today we're talking about, um, a slightly different topic I guess cos we talk about a lot about predict mountains, of course, and the details of the project, but we, we don't often talk about how, you know, the start of a project and actually getting a predict mountains project up and running. Um, so yeah, so that's what we're gonna cover today. Um, but maybe Tom, just for those who maybe haven't, um, listened to you on the podcast before, maybe just a quick intro from your side. That'd be great. Absolutely, yeah. So my name's Tom. Um, my job is solutions engineering for Sensei Predictive maintenance within Siemens. Um, the real role is to try and understand customer problems, um, mainly with a focus on, you know, anything to do with downtime avoidance, um, prevention of certain. Um, mechanical component failures and so on, and really bring this forward to a full scale project, um, and really understand how we can help customers. That, that's what I'm here for. Fantastic, so the the perfect person to interview on this topic then. OK. Um, so let's just start first of all by, um, taking sort of a broader look at things. So, um, we know that predictive maintenance, and you'll know cause there's been so many customers about the predictive maintenance topic. It's getting so much traction. I mean, from your point of view, Tom, what, you know, why, why predictive maintenance in particular? Why is that so in demand at the moment? This is a great question. Yes. So, at the moment, we're kind of in a new phase, I would call it, of, of digitalization, where things like sensors start to become cheaper. The ability to monitor a fully mechanical system or even an electromechanical system is getting easier. And also the monitoring process, the likes of Sensor now coming to the forefront with AI and lots of other features. It's, it's becoming cost effective and simply enough, you know, as a plug and play solution. Um, for want of a better word at this stage, to, to make it work, to pull data from a, from a system, understand how to improve maintenance, make it work for you. Um, and I think just, yeah, that rise of, of scalability, that rise of affordability, um, with the intelligence in the background to make smart decisions. It's really now a reality and it, you know, we've got customers who've been doing this for 5, 10 years, but we've got lots of customers recently really coming to the forefront asking us how we can help them. And, and do you see a shift in sort of mindset away from thinking of predictive maintenance initiative as like a cost on a spreadsheet to an actual investment, like you say, to save money and increase productivity, for example. It definitely is. So you see it as an outgoing, of course, it costs something to install the system to get it up and running, to connect all the different components, set sensitivities of sensors and so on. And also to pull some context in so we can understand what the motors, the pumps, the gearboxes, also non-rotating assets, so things like hydraulic systems or coolers, understand what they're doing. Really, what you need to look at is how does that have a business impact. So where does downtime cost you money? Where does over maintenance cost you money, and sometimes companies don't have an oversight of those things. Um, so it just looks like an outgoing. and also part of my role is to help understand how that's not the case and, and, and really how predictive maintenance can have a really positive impact. So actually, I guess that leads on uh nicely to, you know, this question about, you know, if predictive maintenance is actually right for an organization, because I guess there's many factors to consider. So, you know, what are those factors, Tom, from your point of view? I think start with a business problem. So, not really the technology, really identify high impact areas. So, look to assess major losses when they fail, look to see how leadership buy in and stakeholders are going to look at this, how the IT team looks at, you know, cybersecurity topics. Can we allow data to flow to the cloud? You know, we're seeing more and more customers allowing that to be the case. So we see a positive approach to this. We see a You know, an open mindedness towards digitalization, digital transformation definitely taking place. Um, and then just how do we access the data? So, you know, some of the key factors are, have we got the data? Have we got the sensors that we're going to need? And if you don't, if the customers that we speak to don't have that, that's a part of the scope work that I do. Um, so we can try and build a process, get customers to that stage of, you know, connectivity, pull that data into one central location. And, and of course, we get loads of customers that don't have these things in place. And because there's suddenly a requirement, which there wasn't before, to centralized data, pull it together in some form, you know, that's all part of the project. So we talk about, you know, why is it hard to start a project or how to start a project? You know, that's what I'm here for. That's also what my team is here for. So we all work together. Um, to make sense of it, what's the objective and what are the building blocks we've currently got and what are the things that we currently, we currently lack. So that method to cloud that we might not have in place, or, or, you know, the, the person who's going to react to those alerts from Sensi, for example. And I guess, uh, another thing that you and, and colleagues have to consider as well is you, you must face a lot of misconceptions from customers. And it, and, you know, in a way, you can't blame people because, you know, I know coming from a marketing perspective, the, the amount of tools and products out there saying they do this or that, some doing PDM and not doing it. And, you know, it's a very confusing market out there. So I guess that leads to a lot of, you know, misunderstandings and misconceptions. I think you're right, Nolan. We've seen it. I know you've seen this as well, where we get the wrong sort of requests. It ends up becoming a condition monitoring topic, and PDM for a start is, is much broader than that. It's about really intelligent, um, you know, smart decision making. So the system's able to not only detect that you've crossed over a certain threshold, but also tell you more or less what's going wrong on a broad level with a good understanding of the application and the context of the asset. Um, with its own general intelligence, plus the knowledge of, you know, PDF documents and so on, and give you decision making. So I think some of the misconceptions are, it's basically what we saw 10 years ago. We can add some sensors, we can see when we cross the threshold, but it's so much more than that. And it's very easy when you look at the, you know, the overlevel, the, the, the over, uh, you know, the interface of Sensi. Um, you know, I've seen something like this 10 years ago. I heard that from a customer, at Hanover Messa not long ago. And you have to try and explain that in the background, there's a lot more intelligence than there was in the past, the ability to determine baseline of an asset very, very quickly without models. Um, and I think when it comes to misconceptions, it can be, OK, we just plug it in and it all just works is another one. And I would still say it's plug and play with a little bit of context and understanding of, you know, which sensors do we need, which context data do we need, so that we have that understanding of this particular motor or this particular hoist or cooling system, and we can build that into the software. So as soon as that's in place, which is, I'd still call it plug and play with a little bit of an extra piece of delivery work, then you get the full-fledged predictive maintenance, um. And I think people often think it's expensive and it's not going to pay for itself and what's the value? And, and a big part of it has to be that business understanding, what's the objective we're trying to solve and see perhaps case studies from past, you know, how we've done this with previous customers, um, and really the the proof is in the pudding. We've done this before. And, and actually to, to your latter point, you know, focusing on, because a lot of those points around, you know, it's, it's too expensive and. You know, when will I see a return on investments, actually, a lot of the time it's focusing on, um, the technology aspects of it and not the business outcomes, which you mentioned before, right? So that's part, probably going, oh, you know, do I need this technology rather than saying, what business problem am I trying to solve? I guess that could be a way to look at it. It can be one of the first areas that we've looked to address, because if the business objective is wrong or somehow it's, it's not going to miss, it's not going to align, we can't achieve this certain component, but we can achieve these other areas that you haven't considered. It's making sure we get on the same page as quickly as possible, um, as to what's achievable, where the big wins are going to come from. Um, yeah, no, it's exactly to your point. It's really important to get the business objective, um, in place when we start. And I, I guess to another point you made about, you know, you know, often being compared to condition monitoring. Is there a reason why condition monitoring is often confused with predictive maintenance? Um, I'd say so. The reason, the reason is, and because it's been doing, you know, it's been taking place for, I'd say at least, you know, 10, maybe even 20 years at the scale it is at the moment, where you have maybe some IOT sensors, you bolt them on. There's certain thresholds, maybe there's some, some basic level intelligence in there. Um, that to me is condition monitoring. You're monitoring that sort of condition of the asset. Predictive maintenance then becomes, you know, how do we then make sense of that information? The software within Sensei is able to really enhance that and give you, you know, digital support, give you AI driven decision making. And sometimes it sounds too good to be true, but the more you feed into the software in terms of how you fix problems, how you've solved things, the work that you've done to grease a bearing or do certain work is saved and used as future decision support. And that's a really fundamental difference to, you know, to how predictive maintenance works. Um. Long term, I think we'll see more and more customers start to understand that. The the transition I've already seen it things like the exhibitions we go to, it's quite significant, especially in the last 12 months. The impact it seems to have had great job from the marketing team. I definitely would say that where we've got more customers. Understanding what, what we are, what we can do, and what we bring to the table, um, in a way that, you know, far outclasses the typical condition monitoring as it was, you know, all those years ago. Yeah, well, thank you for the compliment, first of all, Tom, but, um, no problem. Yeah, but I think, I think also the market, you know, the, the, the market's ever changing as well. Uh, but yeah, there, there certainly is that, um, sort of confusion in the market. I mean, I, I've spoken to, um, a few, um, industry analysts recently, and again, that, that's the, the common, you know, when end users approach them, it's always about, you know, what's the differentiator between X product and X product. And I think. Yeah, and I think that's, and it leads back to that point about, you know, is predictive maintenance the right fit for your organization, but if you don't understand what predictive maintenance is because of all this confusion, then it's difficult to get started, right, if that makes sense. It does make sense, and it's good to see that analysts in the industry are seeing the same. Um, and I think just to, just to really reiterate it, in the last 12 months, it's really significant. Um, how many people now know what it is, they see what it can do. I think the, you know, the fact that AI has become a really big buzzword in different circles, it's in the news quite a lot. Um, it gives you more of an understanding of what's actually achievable in an industrial space and maintenance. It's always going to be there until we remove friction from the process, we're always going to have degrading belts, rockets, chains, bearings, you name it. All of these things will always need to be maintained, and it's really important that your process there to, to do that maintenance. I would say nowadays, especially in, in 2025, is digitally driven rather than coming from an old catalog or from an old PDF document that's got a structured process and every six months we do this and this maintenance in a preventative way. We really now move towards, you know, AI driven decision making, which is forward thinking. It's based on the past. Decisions are then extremely intelligent compared to the past. Yeah, I, I know we're gonna come on to these topics, but this is, this is how I see it, massively. No, no, exactly, exactly. And if we, I guess we pull it back to, you know, the start of a predicting maintenance, um, project, but with all this in mind, we've discussed so far, which I think feeds into this question, but what does sort of an ideal roadmap look like for a PDM project? I know there's some context around it, you know, diff customers are all different, right? But if, uh, let's say a typical roadmap might look like. Simple. It is to start small, prove the value, and then scale, and sensesize scales really, really well. It scales because it's got no models in the background, so it works, you know, it learns baseline of any asset that we've never seen before very quickly. But the key to it is to start small and see that value for yourselves, and small to us typically is 50 assets. The reason that's small is because we, um, you know, we look for a process that's enough data, enough machines, looking at enough things that could go wrong. Um, typically in a factory or any environments maintenance, there's lots of things already being well maintained. There's, you know, good quality machinery, which doesn't typically fail. But when it does, it can be catastrophic or you're having to really over maintain stuff. So, if we look at 50 things, it allows us to really show value on at least, you know, 234 of those across the 12 month period. And then you see the value that gives us the ability to prove value. Um, and I guess really that's my job. So we have to make sure we scope the right assets, we look at the right parameters, what sensors, what connectivity do we need? And we've got a whole team of people within Siemens that support this, that support this across, you know, any region of the world. We can, we can make sure this happens. Um, also with partners that are well engaged with our customers. It allows us to really build a relationship and prove that first step, and then, um, you know, really, you know, build confidence, I think is the key with that small scale start and then scale it from there. Yeah, and it, it's, it's really interesting because you use that, the, the phrase sort of small, you know, start small, which small could mean also, you know, in some eyes, let's start with this one really important machine we have here. So what, why is that not sort of a good starting point for predicting maintenance necessarily? That's a good topic. No. Um, sometimes it becomes the machine is, you know, worth €3 million or 3 million pounds. You think, well, actually, this is something that we should monitor. But maybe there's a, there's a whip storage. There's a, there's a back and forth in the maintenance process where that machine is only switched on 2 days a week. You throw a few things through it, it does what it's supposed to do, stamping or pressing or something that's on a larger scale. And actually, if you were to stop it for a few days and run it back up again on a couple of other shifts, it doesn't really impact the process. So you have these big gaps to do maintenance that don't affect the process. Whereas somewhere else, you could have a small conveyor that's only €200. It's really crucial and it's a big bottleneck in the process. And actually, if that thing stops, you lose hours of production time or, you know, depending on the criticality and depending on the output of what the product is. The impact could be huge, even though the asset we choose to monitor is fairly insignificant or fairly, um, you know, cost-effective. It's, it's cheaper compared to the rest. So again, this comes back down to business objective. Do we look to extend the lifespan of the, you know, the 1 million pounds CNC machine, um, or likewise, make sure we don't have a stoppage and a downtime on a, on a, um, you know, a smaller, less significant asset. So, so I guess what we're saying is a really important part of this is asset selection is looking at business outcomes, looking at the right assets, you know, for, you know, to meet, you know, you know, to, to meet those, um, business objectives, I guess that, that's the, yeah, absolutely, and it can make or break a make or break a project because if you, if you select the wrong assets, and that business impact isn't seen for obviously whatever reason. Like sometimes we see this, we see where a customer wants to monitor the biggest, most expensive assets, and that's the asset they would typically do lots of over maintenance on, even if Sun size is in place. And so after 12 months, it never broke down. It never was given the chance to fail or to get close to degrading for Sunite to show the impact. So, yeah, no, absolutely. Asset selection is really, really important. And I guess maybe just, you know, for the audience, um, information like how, how do you go about working with the customer to select the right assets for, for a project on a typical project? What it comes down to is we understand each other, so we have interaction between each other to understand, you know, how can we help? what are the problems? Um, this could be online. And then it leads perhaps to an onsite meeting and somewhere within this, it leads to a scope workshop, which is what we term as the chance to come and have a look at the machines, do a walk around that could be around the mine, it could be around a steel factory, it could be around a chocolate factory or anything. It could be around literally any kind of processes that we're looking at, you know, it goes without saying since size. It's a completely system and industry agnostic. So, we look at anything. And the team, you know, the solutions engineering team have got lots of knowledge, lots of understanding of how mechanical systems fail, what to look out for, what questions to ask, what kind of data to look for, and make recommendations as well, you know, if we need additional vibration sensing or some other monitoring or we need some context from the PLC system. This all leads towards, you know, correct selection of assets. And once we do that, we continue to work together. Uh, it, it becomes a real collaboration. It's not like, here's a box of stuff. …