Predictive Maintenance - Real-world phases + what actually happens after kickoff) - 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.

In this episode, we're joined by Tom Jacques, a Solutions Engineer for Senseye at Siemens, to break down what predictive maintenance looks like in the real world, from kickoff to daily use and scale.

What we cover:

What actually happens during the first 30–60 days of a predictive maintenance project

How proper scoping, asset selection, and data availability set projects up for success

Where projects commonly slow down or stall, including resource constraints and misaligned expectations

How pilots transition into day‑to‑day operational use

What creates real “aha moments” for maintenance teams

Why trust is the key factor in getting teams to act on insights

How Senseye Copilot supports decision‑making without replacing human judgement

What separates pilots that scale successfully from those that remain stuck in PoVs

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

Connect with Tom on LinkedIn here:

https://www.linkedin.com/in/thomas-jacques-22655585/

Transcript

We know in the future, you know, what is indicative of a failure about to happen, and we can, we can look out for it. But really it comes when you've, you've had that first success, you've caught that first breakdown, or you've seen the first instances of, of a, a case being raised in Sensei before the system's failed, and then you realize, you know, this is not just a condition monitoring tool. It's, it sees really deeply into the, the behaviors of your machines and Sensei is really powerful in that sense. 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. I'm, I'm really pleased to be joined once again by my colleague Tom, and he's been on the podcast many a time. Uh, the last time was actually in the studio, but we're, we're remote this time, but really pleased to have you back, Tom. Brilliant. Hi Nia, good afternoon. Good afternoon, and so today we're gonna be talking about predictive maintenance in, in practice, and I guess what we, or what I mean. By that is sort of in the real world, and I know, you know, all the episodes we've done together and, you know, lots of other guests as well, providing their own perspective, and you've got a fantastic one, of course. But yeah, we're gonna go into the detail, each step of the process, um, and the process, of course, starts with a project kickoff. Um, so, if we start there, Tom, so what actually happens, or usually happens in the first sort of 30 to 60 days. And is there sort of a, a difference between what we do and what the customer thinks might happen in that period? Yeah, I think sometimes there is, um, yeah, so I'll gladly go into it. I mean, uh, for anyone who's not listened, um, to me on the podcast before, my name's Tom, um, solutions engineering for Sensei. You know, covering really that first part of a project, so what we call scope when we do a predictive maintenance project, and of course within that you have to understand first of all, you know, what kind of project, um, are you looking for? Is this a downtime avoidance project? Are you perhaps over maintaining the system and that's something you need to look to mitigate, you're spending lots on, on spare parts or you're doing lots of time-based maintenance, perhaps, you know, quite often unnecessarily. Um, but equally you could be GXP compliant and you need to do certain, you know, regu regulatory changes and repairs and so on to keep things operational. Things like food and beverage, you'll need to keep things clean, so there's cleaning windows, those are often also very good maintenance windows. We often see different customers with different requirements. And I think really when we kick a project off, the first thing we need to do is, is just get alignment, make sure we all understand what the objective um is going to be for the project, make sure we're, we're capturing data, make sure we've got the right sensors and things in place and really just, you know, look at the sensors and the assets and the kind of hardware that we've got available to us, if we've got, you know, motors and pumps and gearboxes or, or whatever the assets are that we're looking at, again sens size covering. Rotating, non-rotating assets. I think the first thing is, is really look at what assets we need, look at where the downtime is most impactful or where the, you know, where the, the problems lie, and, and figure out how we're gonna get that data through to, to, uh, the Suni platform, which is all cloud-based, and then how are we gonna see success. Um, so make sure that we're set up in the right way for, for practical delivery of the project. And also just understand if there's, you know, a, a, a historian in place already, the data capture is already there. um, also what the workflow looks like, is there already a CMMS solution in place. Do we plan to integrate those two things and really how is it gonna impact the maintenance team? So Sensei is inherently a maintenance product and that is really the team that's gonna impact the most. It will change the way operations are, are currently run. And to be honest, Noel, we've, we've seen customers where it even might be a, you know, a paper-based process still. We just wanna make sure that the process from whatever is currently there to Sensei is also mapped out and, and the expectations are, are sort of laid out correctly. And in, in those sort of various different sort of phases and parts of the process, is there a place where projects, uh, commonly stall or slow down in particular? Is there something that sort of holds things up? Yeah, I think making sure you've got the right resource, um, and also making sure everybody on the, the Siemens side, also when there's partners involved, the partner side right the way through to the end customer. That the resource is there. Now that could be financial, it could be skill sets, it could be time, you know, sometimes, especially in maintenance teams, you're, you're pressed for time, you've got lots of things to fix and, and keep running. Start in a pilot with a, with a predictive maintenance product sometimes isn't the, the top priority, you know, we see that quite often. And it's making sure that the resources are in place before that then slows you down at a later stage. Really, as well, from my perspective in solutions engineering is that we, we scope the project properly so that we make sure that whatever it is that you're looking to solve, we're going to be able to do that and we're not missing anything. So, if we're going to monitor a certain drive, we know that drive can Output, the values that we need, for example. All of these things need to be thought about, and the better job we do of a scope phase of understanding the assets and really working together with the customer to make sure we get the project right, the better the project tends to run. Um, you know, you could easily halve the length of project or shave many weeks off a project just by, just by doing the scope phase correctly. Uh, I was wondering, because when you say the right resources, I imagine that across a lot of projects, it'd be, you know, similar roles or similar departments, but then. In others, it might be different, whether it's due to culture or the size of organization. And it also just reminded me of, um, our colleague, Natalie, who was on the podcast, we recorded a podcast last week, and she said that actually joining the case review calls was a, the CEO of this particular company, right? Which is, I, I, I still can't believe that, that they're so. Involved in it. I mean, that's quite rare, but it does it vary between different companies, what resources are required, or is it quite sort of standardized, those certain roles which. It can vary. I mean, that example from Natalie, is, is a very good one because it's about stakeholder management. So perhaps from the bottom level, there's a team that want to push this, but you know, in some cases, management don't understand it, they don't see where this is going to benefit the business. So stakeholder management's really important. Make sure the sponsors who might pay for the project. They understand the benefits, how is this gonna pay for itself? What is the objective of the project. Also, we call them pilots, but we're not testing that the, the product works. We know Sensei is a very powerful tool. We know that it can do what it says on the tin, um, for want of a better phrase, but really it's making sure that messaging is passed across to the, to the people that will be involved, um. Yeah, and, and again, that's part of speeding up the project, if you, if you explain all that stuff in the first instance and you get buy-in from the right people. Yeah, absolutely. I guess amongst those different roles as well, I mean, how does, how does that transition between, you know, the initial sort of excitement of getting the project up and running and looking at assets and all that kind of thing to actually handing over to sort of the, the operational side. So we've set everything up now, off you go. It's time to use this day to day, and I know it's not just a seamless handover and that's it. Obviously there's continuing support, but yeah, how, how does that work from a customer perspective? Good question. I think when customers have expectations in the first instance, the goal will of course be, we need a tool that's gonna help us with the maintenance. We, we need a, at the minute we're very reactive, we find ourselves chasing our tail quite a lot, um, especially when you then need spare parts that you might not have in stock or you. You know, for whatever reason, there's there's other delays throughout any maintenance process and to be able to anticipate what's gonna happen is really important. I would say, you know, just being able to determine there's anomalous behavior or trending behavior, you know, moving away from a, a baseline, increases in vibration, increases in temperature and so on. Again, just as one reference point on a motor, uh, that's a, a very typical rotating asset. A large part of industry is kind of built on, on motors, three-phase motors and so forth. But really that the expectation is there. So if, if you say we want to capture failures, OK, but how? So what data have we got available? Do we need to make some data available? Is there a methodology here where we can implement IOT sensors to extract, you know, additional insight, vibration sensors there have been a really, really useful one. You know, new features and functionalities of SenseI where you can look at the spectra in Sensei and get it to really do the analytics for you in the tool rather than having to get uh a manual report done on 100 motors that you've got in place. And then look to the future and you say, well, OK, if we implement it on these assets, you know, we've got 10 of those, but actually, if we implemented it on these conveyors, for example, we've got 1000 of those, perhaps in the logistics facility or other, other customers where you might see big mining operations. Sensor is really scalable, so the tool can learn and And that data can be shared across fleets and groups of assets. I think really when the project starts, looking at how you foresee it moving forward, basically, and there's always the outlay of, well, these assets would be better, but we're gonna have to pull lots of extra data. We're gonna have to really do a hardware install project, you know, to pepper this thing with sensors. But actually over here, we've got loads of data in the drives already in a sense I can leverage. We just need to make sure we can capture that data, pull it from a historian, and really then just leverage that. Start there, start basically where the low hanging fruit is and make sure that the team are, you know, fully on board with that. That's an interesting point about sort of looking to the future, but what I think inspires that is these sort of aha moments, you know, where I'm not saying people will, or maybe you do have some people who are skeptical to start with, but even people who aren't skeptical, they're waiting for that moment where they're like, oh my word, that's amazing, it's completely. This is going to change everything. I need to apply this to every asset or as many assets as we can, but what, what triggers those moments from your perspective, Tom, from your experience? Usually that comes when you've started the project and you start to see success somewhere, or you think if we just, you know, if we just modify the parameters on this slightly, we've just seen a bearing failure or we've, you know, we've just seen a system breakdown because of poor oil quality or something like that in a, in a loop system. As soon as you can start to capture that data and maybe you've just missed the failure, but then of course the work that we do together, you modify those um settings, increase, decrease sensitivities of certain measures and so on, that allows us to really ensure that we can sta you know, effectively stand a much better chance of capturing that failure in the future. And then once we know what that failure looks like, we can match against that. So we know in future, you know, what is indicative of a failure about to happen and we can, we can look out for it, but really it comes when you've, you've had that first success, you've caught that first breakdown, um, or you've seen the first instances of, of a, a case being raised in Sensi before the system has failed, and then you realize, you know, this is not just a condition monitoring tool, it doesn't just. Alert you to a, a, a violated threshold on a measurement of, of temperature, for example, sensor is really powerful in that sense to have the intelligence to see, you know, I really wanna get sort of philosophical about it, but it sees really deeply into the, the behaviors of your machines. And then with copilot, you can start to ask questions to the motors and pumps themselves. It sounds crazy, but really that is as good as I can explain it. It's the same as being able to then speak to the asset and ask it, you know, what happened, you know, how did that happen, why did that happen? and it will then give you all those recommended actions based on we saw this, this was indicative of this type of failure on this asset. If it's, let's say a loop system, you can see that it's based on a high isocode in the oil quality, really degrading, so you can see that isocode's increased. That's then indicative of, you know, you need better filtration, you need, you need to find ways of mitigating that in future, um, so it doesn't happen again. And then it learns from all that stuff that you do, so every time you've fixed a problem, you fill into the tool what you've done to fix it, and that's then used as suggestive action in the future. Um, and pulling data from PDF documents and so on, so it can troubleshoot the system for you as well. But yeah, I think when, when it really comes down to some of the challenges, I think one of the things we have to make sure is that we're, we're not over egging the pudding. We, we really have to make sure that we set the expectations correctly. If it's a small data set or you, you know, you maybe make 100 different products on the same production line, that's gonna take longer to understand. Um, normal operation for each of those different batches or products, and it's, it's essential there that we set expectations correctly, depending on what we're looking at. It's, it's funny, Tom, when you were saying about talking to motors, imagine like, they're, they're like your best friends, right? So you, I don't know why that, that, that appeared in my head when you're using the co-pilot to understand them a lot more. But I mean, it's almost like it reaches out to you and says, I've, I've, I've had a misalignment. I really need you to, you know, come and support me. That that is effectively what it's doing. No, but I, I completely get it. I, I guess what I wanted to sort of add on to that point is also, you, you talked about Sensai and the, we talked about our hard moments, but of course, for an insight to be really useful is someone to actually act on it, right? And so, what usually determines whether maintenance teams act on something, you know, what gives them the urge to do it? You know, we've got lots of examples where Sensai spotted things and people shrug their. Shoulders, something's broken down and then it's, oh, you know, what happened here, you know, those kinds of scenarios. Fundamentally, it comes down to trust. You, if someone who's very knowledgeable and says, you know, I can hear that machine's clunking and making strange sounds, I think that is about to fail, and, you know, it, that is indicative of a bearing that's about to collapse. It's only subtle hints, but that's for decades, of, you know, a very tried and tested method. If, if it starts to sound different or you see very similar. Simple sensors that tell you it's getting hot. As soon as you see a bearing that starts to get hot, you can see there's increases in friction. That bearing's not doing what it should be doing. What we try and do is look out for, um, culmination of different things. You can compile that, you can have correlations between measures and so on, um, control parameters and so on. What the software needs to do once it's all been set up is prove to the customer that it can, it can actually do what we say it can. If it's gonna indicate to you. That there's a failure, that really is true. Um, of course, when you close a case, it gives you the option to say, you know, this was very helpful, excellent, very, very useful, equally no use at all, you know, maybe there's a, a process change that's caused strange behaviors or wildly high ambient temperatures has, you know, indicated that the bearing is hot, but actually it's just a very, very hot day, so making sure we've got other measures in there to pick that stuff up. Um, we've, we've factored in maybe a delta between two temperatures so that we know it's just the difference. Making sure that the system is set up correctly so that we capture, and of course we do this on many thousands of assets, so we know what we're talking about. But making sure we can build the trust because if you have a team of people who don't believe it. The alert will pop up exactly as a condition monitoring software did 10 years ago to say, oh, I think this is a, a failure, you need to go and have a look at it, or you get nuisance alerts where it just keeps bouncing off a limiter to tell you, you need to go and have a look at something. Sometimes previous bad experiences with old software in the quote. Unquote, predictive maintenance space has built that negative, um, negative image of what the software is capable of. That's probably one of the biggest hurdles we have to overcome at Sensei, because Sensei is, as far as I see it, the only tool that can really, really do this with enormous amounts of analytics in the background. Yeah, it's, it's just about building trust. It's making sure that when there's an indication that the team then realize, OK, that's sensor, we know this isn't a false alarm. We know it's based on, you know, good data, good historical information. Uh, we know that it really has seen something. We need to go and do some checks, do those things, and then you think, wow, brilliant, that's just saved us 10 hours of stoppage in whatever facility, a steel plant or a, a, a chocolate making facility, whatever we're looking at. And that, I, I call it confusion, you know, the market confusion, let's call it, where predictive maintenance is applied to so many different products and companies and stuff, so no wonder people are so confused. I guess that, does that sort of lead sometimes, you know, building of trust is one thing, but where you have customers who move from, say, a proof of value position to, you know, using it day to day and operations and scaling and off they go to the ones who get stuck. You know, in this cycle of proof of value, is there something that sort of separates the two, the mindset beyond the market inflating expectations, let's say. I think, I think once it becomes embedded in daily operations, it becomes very easy to trust it. In the first couple of instances, we do see this with, with sort of fledgling customers, they've got their 1st 100 asset project, let's say. And then they get an alert, they'll tell the system, yeah, we've been to have a look at it, we don't think there's a problem, and maybe 2 or 3 weeks down the line, you get what I would class as a catastrophic failure, you know, serious damage in your gearbox or something else goes wrong. And the signs of uh, of that failing and breaking down were highlighted 2 weeks ago, but you don't realize the sensitivities of it. …

Full transcript (WebVTT)