Predictive maintenance in the metals and mining industry - with Alex Hill

Trend Detection Podcast ·

Show notes

Welcome to the Trend Detection podcast, brought to you by Senseye Predictive Maintenance – the platform, powered by Siemens, which enables predictive maintenance at scale across all of your assets, across all of your plants.

For this session, we were joined by Alex Hill, Global Head of Business Development, Senseye Predictive Maintenance.

For this first episode in our industry focus series, we dive into how predictive maintenance is helping metals and mining manufacturers to create a competitive advantage in their market through IoT and predictive maintenance.

Please subscribe via your selected podcast provider to be notified about future episodes.

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 Alex Hill:

https://www.linkedin.com/in/anshill

Connect with Niall Sullivan:

https://www.linkedin.com/in/niall-sullivan/

Transcript

Welcome to the trend detection podcast powered by Siemens, and now your host, Nel Sullivan. So welcome to this 3rd and final episode in our series looking at some industry highlights related to predictive maintenance. I'm really pleased to invite Alex Hill, our global head of business development at MI Predictive Maintenance here again. Hi Alex. Bit of a mouthful. I know it is. You need to shorten that title, I think, but it's fun. It's fine. But no, we've, we've, um, we've previously talked about automotive industry and the food and beverage industry. But today in our final episode, we're actually going to focus on the heavy industry, so metals, mining, which I know is a, uh, let's say, we've got lots of great use cases in, in that area to, to talk through. But I guess we've started the previous episodes, and it's looking at the challenges again. So it'd be a bit remiss not to, to not do so again. So can we start by looking at the the unique challenges to the, to those industries? I think one of the, one of the primary unique challenges is it's a dirty industry, and I mean that in every sense. So politically, environmentally, uh, technologically, everything is, is, it has some bad connotations around it, and it'd be wrong of us not to address that elephant in the room. But the fact is that our technology can help that industry to be less dirty, to be more sustainable, to be more environmentally friendly, and so there's a huge amount of opportunities to do that. And the companies that operate in that area are actually very committed to making sure that they can operate sustainably sustainably and in an environmentally friendly way and in a way that's respectful to the earth and people, because if they don't, they won't be in business because people care about that kind of thing now and these companies are ultimately run by people. So that's one of the first challenges. How can they operate in a way that is respectful to people and the environment and continue to do that. Then you get into the classical areas of from a machinery point of view, their environment is dirty. It's incredibly hard on the machinery. Things break all the time, there's contamination, there's, it, it's, it's, it's one of the most challenging environments that a machine can be in. And so how can you make sure that you can operate this machinery and make money with it whilst not spending all of your time buying new machinery and maintaining it. So first of all, machinery is quite tough, and you then need to monitor it and have some more sophisticated understanding of where to go and fix and when to fix it at just the right time so that you're not spending too much because. The the margins in actually extracting uh minerals and things might be quite low, so you've got to do it at quite a high volume, which means you need to keep doing things, uh, often through a number of shifts and through a large amount of machinery. And so downtime becomes very important. How can you avoid unplanned downtime, particularly in areas where if you're processing and refining some form of ore and producing steel or aluminum, that's a continuous process, and you have something that is hot and effectively live going through your entire factory. If something fails during that point, you could waste all of that product, damage all of that machinery. And it might take you uh over a week to reheat that material, to reheat the furnaces and things, and, and the energy costs involved in that are, are phenomenal as well. So unplanned downtime through every part of that industry becomes really, really important. I think unplanned downtime, sustainability, safety of course is massively important, making sure that these things are done in a safe manner and making sure that they're not spending money where it doesn't need to be done. And you've also, I mean, I like your, I like your reference to a sort of dirty environment, but it's also, it's, it also requires sort of rugged technology in terms of, I know we're sort of moving away from software to look towards the hardware side, but is that a challenge as well in that industry to use, you know, hardware sensors, etc. which actually can withstand the pressures that you sort of mentioned before can be. The heavy or the mining metals industries, there's actually quite a lot in common in terms of monitoring with oil and gas, where they've got a quite a deep culture and deep understanding of how to use condition monitoring to optimize how they perform their maintenance. So they already understand it quite well. We don't need to explain what condition monitoring is, what predictive maintenance is. They get it already, they know it, they're probably already doing it. The challenge they have is how can they not just be looking at these 10 or 50, 20 critical machines, how can they now be looking at all of the machines on this site, and that could be thousands just in a single site. And that's, that's really challenging to do. The actual sensing equipment itself is often built into the uh the machinery that they're buying, or they're already collecting process data that we can use as condition monitoring data. One of the really good things about the metals and mining industry is they, they centralized on or they, they adopted the idea of collecting interesting and useful data quite some time ago, so that they're decades ahead of some other industries. They almost ubiquitously use a data historian. Which means that you get all kinds of interesting process data being collected because the process needs to be very tightly controlled, particularly when you're you're actually refining something and and putting something out at the end. And that process data can actually be used for condition monitoring, and people don't think about it. When people think about condition monitoring, they think about vibration, going installing and retrofitting lots of things. And yeah, there are certainly some cases where the equipment you install needs to be, needs to withstand very high temperatures, for example, in a pot room where you're melting these, these metals. But there's a different way to do it. You can start to use that process data that's maybe just looking purely at the timing of things and start to use that for content monitoring purposes, because if something is starting to degrade, those timing intervals are going to change, and that gives you that very early warning, essentially for free, that that machine is not behaving quite in the way that you want it to. And that allows you to flag up, this is what deserves my attention, and the other things don't. So we have some fantastic case studies with companies, and there's one in particular with a global aluminum company, and we managed to achieve a full ROI at a site within less than 6 months, and we managed to enable them to reduce their maintenance spending by about 20% without installing any sensors, so they have no specific condition monitoring sensors. They weren't doing anything, but they were collecting process data that happened to be the right kinds of things, current, temperature, torque. Timing, things that are free from a process control point of view, they had already stored it. We really, I don't say we merely connected things and and and it just happened. There's still quite a lot of work to be done in terms of making sure we're looking at the right things, but they had that infrastructure in place already just as a result of, of them being able to do normal business operation, and they could repurpose that investment and infrastructure for predictor maintenance purposes, and that was awesome. And is there a reason why these, you, you mentioned these companies were more, you know, light years ahead of other industries? Is there, is there a reason for that, or is it just based on those, you know, a massive list of challenges you listed at the start that they, you know, they're, you know, they're just looking for solutions and technology seemed like the, the logical, what is so, I'm just trying to dive into that a little bit. I think for one reason or another, there's quite a lot of crossover between oil and gas and the technologies and the approaches and, and mining and metals. And, and heavy industry is not just mining metals, but other things like concrete, pulp and paper, and so on. But for us we do see the most growth and the most adoption in mining and metals specifically, and, and that's, that's because they they actually have that whole culture of safety and appreciating that they don't want to be up against these machines all the time. That they need to be continuously running and that they need to be monitoring and that, that culture is is really carried over from oil and gas. I don't know why. I've not really explored that too much, but it's, it's something interesting to look further into. And I guess another aspects as well, which we haven't touched on in the previous series, but you mentioned health and safety. So I guess the fact that it could be monitored remotely as well, the machinery could be, you don't have to be on site to monitor in these dangerous or, you know, environments, I guess. That's actually a really good point. And one of the things that we focus on quite a lot as a business outcome, the fact that Even before the Coronavirus and the fact that we had to spend 2 years indoors. Maintenance is changing, it's no longer about going up to something and resting your hand on it and understanding what you need to do. The people who can do that are, are quite rare. They don't scale very well, because you can't scale people like that, and, The new maintainers coming in are used to technology. They're used to doing things more remotely, and they have their colleagues who are sitting in an office and being quite relaxed. And do you want to be sending people up against a, a machine that is a noisy, dangerous, difficult environment if you, if you can avoid it? No. So maintenance is becoming more and more remote. The fact is that you can do a lot of your diagnostic and, and planning and preparatory work completely remotely. By looking at the data that's coming through and having a system that guides you as to what you need to go and look at and. The workplace is changing. Mining and metals is still seen or perceived as a dirty industry, but the fact is that they go to extreme lengths to make it safe and clean and and have a minimal impact on the environment and humans, and technology absolutely helps them do that. No, actually, because I know we've done in the past, in the past year or so, done a few podcasts around this topic with, you know, experts in the industry, and from what, what I gather, there's another reason why, you know, technology might be more widespread is that there's a lot of sort of knowledge sharing. I know we've talked on previous episodes about is knowledge shared, but I, I sort of understand that there's, In, in mining or metal, in particular, every industry in particular, it's more not sharing. Is that your sort of experience too? It is, and I think it's because the longevity of these sites is, is measured in decades, not, you know, they're not buying a site and and working at it for six months or something. They're committing to that site, to that area for decades, and so the technology needs to be. Uh, suitable for that and they can grow with that. And then when they have these sites, they've been monitoring for decades and working with them for decades. How can they start to optimize, how can they start to share this learning, make sure that they are getting the product for, for the best price and the best way. And, and so we see a lot of, uh, probably higher than, than most other industries. We see a lot of connected thinking and and knowledge sharing and uh connected tools where you can start to monitor sites across the world. Um, I don't want to say easily, but with, with a relative ease, there's there's quite some um commonality and it might not be exactly the same things. There's at least going to be a commonality, a common language, a common way of doing things. And we're starting to see more and more infrastructure and investment in truly making those tools ubiquitous. Let's use the same instance of something. Uh, previously that wasn't possible because often these assets are very remote, uh, very difficult to get data from them, very difficult to in, in the past, very difficult to, to connect the database that's in Australia with one that's in Ireland, maybe, um. But we've seen that happen now in previous examples where we had a customer in Thailand streaming data to Australia and then processed an island in a data center and and accessed all around the world, and the technology has made that kind of thing possible. And actually it's, it's an interesting point to touch upon because in terms of assets, obviously cent predictive maintenance come on to we say many, any, any type of asset, any type of industry, but it, I guess it's the nuances between the type, what type of assets, particularly in metals and mining are most prevalent and like you say, more mobile assets, I guess is the big difference compared to other industries we work in. Mobile assets is is really interesting and that's, that's been a challenge as well. A lot of these, uh, trucks. You know, the size of a house or a decent sized palace, they already have very sophisticated monitoring, but that monitoring's been really for a, for a dashboard for the driver to say, hey, something's really, really bad, you need to stop, uh, a much more sophisticated version of the check engine light that we all have in our cars. But now, now that we actually can have internet access, 5G, whatever, pretty much everywhere, you have the ability to stream that data out and to centralize the analysis of that. And that gets even more interesting, how can you start to monitor and optimize not just the vehicle, but the whole fleet, your whole usage of the fleet, and, and. Combine that understanding of how the fleet is behaving with how your crusher and how your uh grinders and how your conveyor systems are operating and put that all together, that hasn't been possible before from a connectivity point of view and from a processing point of view because. Since I predict Mason takes that very data-centric point of view, we can start to combine the data from these mobile assets with fixed assets, and, and people always ask us, well, you know, you can do these conveyors and things, but can you do mobile assets? And really it doesn't make any difference for us as long as the data represents the, it starts to represent some sort of degradation or condition of that machinery and a human, if a human were to look at that data and they can't see anything. We won't be able to, it's not magic, but as long as it's valid condition monitoring data, we can automate the analysis of that, and then you can have your mobile machines as well as your fixed plant in the same area, and it gets to be a really nice holistic view. Yeah, I guess the other thing we haven't touched upon in, in the previous episodes either, but one thing we talk about a lot, it's not just the data condition monitoring data and the data from the machines and assets, but it's actually also the human input as well from the the people who know that asset, we work with that asset every day. Could you sort of just explain a little bit how Sensei uses the human input and aspect as well, because I think that's really interesting to cover. So the traditional method is very good, very appropriate in some industries. The problem is, if you want to expand that to 500 machines, 5000, 50,000 across an entire enterprise. You need to be doing a lot of manual human understanding of failure modes and building and maintaining models. That's incredibly expensive. So whatever money you save in the maintenance of something, you're going to be spending tenfold in the actual data analytics, and that's not a viable way to approach things. With Sensii predictive maintenance, we, we, we turn that approach on its head. We accept that the data that we're looking at is going to be poor. It's low context, it's, it's noisy because that's what environment, that's what indus industrial data looks like. Things are never perfect as we want them. So we have to accept that, and that's, that's one of our fundamental principles. Then we have to understand that the, the data we get is actually quite low context as well. So what I mean by that is we don't know everything that the machine is doing. We don't know everything that the company is producing, how they're doing it. We also don't want to know. Some of that's their IP and, and to get them to share all of that, it's, it's just not, not a really viable way of running our business. We want to focus on, on the machine. And then What humans do about that. And so when people talk about predictive maintenance, they get all excited about machines and they focus on machines. We want to focus on what those machines mean to the business. And the business is about humans. So really what we're focusing on is the humans, we're using machine data to help that. So rather than saying that we know everything about this machine, and we're going to uh tell a maintainer that the bearing is spoiling in a particular way and blah blah blah, what the maintainer probably wants to know is the bearing's going bad, I need to do something about it. And the fact is I have all of these other machines with bearings that aren't going bad. So I'm going to focus my attention on this one that actually is, and do something in an efficient manner that saves the company money, rather than having to go and check things all the time and, and, and do things just because the manufacturer says so. So by that, We, our currency of how we work is human attention, not machine health. And that, that flips this whole thing on its, on its head. We're looking at machine data, we're working with machines, we're, we're, we're guiding people about where to pay attention with the machines, but actually what we care about is human attention. That's, that's our currency. We take the data from machines, and we direct human attention to where our algorithms believe it is most warranted. And then we learn from what the human does, because our data is inherently actually quite poor, because we're looking at very simple condition indicators, things like current, torque, temperature, pressure, things that actually by themselves, they're not very interesting. But when you start to look at them over a length of time, and you start to combine them with other things, and you start to combine them with data sets that you feel are also similar, then you start to build quite an interesting picture, then you start to build context that becomes more meaningful. And one of the best ways to build that context is by learning from what the humans do. They, they know this job already. Why sit them down in a, in a room and get them to spell out everything when actually we can learn on the job? Yeah. So that's what we do. As people interact with the application, as people go and perform work, we learn from that as well. What do they do? Why did they do it? That's recorded in the maintenance management system, or within sense I predict the maintenance itself. And we combine those two things, we combine this industrial machine data, as well as this human data, and that helps us build context automatically by people just performing their normal roles, and it helps the algorithms become more and more effective, more and more efficient, and it helps people ultimately zero in on, More and more, what they need to be doing and relate that to a business outcome. So we're matching all of these things in the background. …

Full transcript (WebVTT)