Cloud + On-Premise: Unlocking Predictive Maintenance at Scale - with Richard Ella

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, you will learn:

• Why combining both on-premise and cloud solutions creates more value than choosing just one approach

• How on-premise systems excel at mission-critical control and high-speed analytics while cloud enables scalability and long-term analysis

• The key security considerations when transitioning from operational technology (OT) to information technology (IT)

• Real-world examples of hybrid approaches in predictive maintenance, including vibration monitoring and process data analysis

• Why future-proofing your operations requires flexibility between on-premise and cloud solutions rather than a rigid single approach

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

Look, when it comes to on-prem or cloud, this is not black or white. All right. If, if, if, if you had to ask me what's the best, it's actually the gray. Because uh when you combine the power of on-prem and it's um value add and cloud, you end up with this really powerful solution. 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, today I'm joined by a very familiar face and voice, um, Richard Ella, um, my colleague all the way in Australia. How are you, Richard? Yeah, good day, Nia, uh, from Down Hunter. Um, yeah, pretty good, pretty good. Looks like you've, um, done some damage here. What's going on? Yeah, yeah, yeah, that, well, people won't be able to see this, but I have, um, hurt my, well, no, I've had an operation on my wrist, so, um, I'm doing podcasting one-handed, which is easy as it happens to do. So not going to affect my presenting hopefully today, um. And, uh, and actually, well, today, um, today's subjects, um, it's a slightly different one because, uh, with, um, Sensii and predictive maintenance, we're often talking about cloud and the benefits of clouds, and cloud's great and, and cloud's the be all and end all, but actually, it's not. And we're gonna go into that today and talk a little bit about on-premise and how it doesn't have to be one or the other. One can certainly benefit the other, and you can use both in tandem to get the results that you want at scale, to unlocking the value at scale. Um, so have I summarized that OK, Richard, from your point of view, because I know you sort of brought the topic to the table. So, you know, was there anything else in there you wanted to add to that? No, no, I think, I think you really hit the nail on the head there. Um, you know, when it comes to deployment and when we talk about generating value, um, it's, and, and we start talking about introducing new products. Customers, especially in the industrial space, will always question, why should I do something different? Why should I Uh, move my data, um, into a cloud platform. Why can't I just leverage the large amount of money, uh, I, I've spent on my plant that's got very high speed processes in. I've got my historians. I've got my condition monitoring here. What's the real value of, uh, pushing my data out, um, and, and getting it analyzed and reviewed in a different way. And so. Um, what I've seen, uh, I've been in automation, um, now for over 20 years, and I come from a coalface PLC drive, uh, background, and it is in, it is in your DNA to challenge, uh, uh, why we should do things differently when these people that have been coding for a long time. Have, have spent a lot of uh amount of energy, um, in, in design and deployment and making and making sure that they come up with a program that really looks after the application. And so, so now when we come along and say, listen, we can see things that you couldn't see before and that kind of sometimes gets people on the back foot. Because they were like, well, you know, I know machine, machine very, very well. What are you gonna do that's different to what we do here on prem? And so, that's the topic for today is, um, what do we do on prem? What do we do in cloud, and really what is the best solution moving forward, uh, to make sure that you're getting the most value. Now, that's essentially the topic today. Um, I don't want to give my insights until the, the very end as to what I feel is the best. Um, I'd love to just kind of go through this journey of customers that I've spoken to, challenges that we've faced, and talk about when, um, it was better to stay on-prem and when it was, uh, best to only go cloud or when there was a combination. Yeah, so let's start with that then and break and break this down a little bit. You know, so if we start with on-premise analytics, so, um, in the context of sort of industrial um operations, you know, what, what are the benefits of, you know, on-premise analytics? Yeah, so, um, essentially, this is why factories get built. All right, um, they are meeting the needs of, um, whatever they in the food and beer battery, pharmaceuticals, whatever, um, vertical they're in, um, there are products and PLCs and, and OEM equipment specifically designed to do a function and make the widgets that the customers need. And, um, and automation has been absolutely awesome, um, over the last 20 years in, in how, how much easier it has been to visualize data as opposed to just having blinking LEDs on a template or how, how much easier it has been to actually automate. Uh, code and you didn't have to write code in C++ or CS SEL and, um, it's really become easier at an application level to, um, basically derive process and, uh, rules-based condition monitoring and so. This is um what all of our um customers are doing. That's their focus. Their focus is not to bring their data into cloud to do something else. Their focus is to manufacture widgets and, um, and so this is what we talk about is the evolution um of um Industry 4.0 and especially in the condition monitoring space. So when we talk about on-prem, we are really talking about the physical assets that are on site, be it your motor, be it your, um, you know, your, um, Desal, be it your, um, uh, water cleaning or whatever your application is, um, that is actually where all of the, the, the logic sits and that's, so now we have to come and, and talk about, you know, what can we bring to the table that you're not doing on prem. And it's important. So, what do we do on-prem? Traditionally, you know, everything that's rules-based condition monitoring, uh, if this, when that, those, um, is traditional, uh, edge processing. We have seen the evolution of model-based AI where they've taken. Um, and OEM might know their machines so well that they go and write model-based AI that runs on-prem and is, uh, operating at a high spectrum. OK, so this is looking for, uh, anomalies or deviations at a millisecond time slots, very useful for, uh, monitoring grid. You're looking at frequency or transience in the grid, very handy if you're looking for voltage deviations on drives or um there is definitely use cases for that high speed um AI on-prem and we can kind of go through very, a lot of examples there, but essentially when we talk about and define on-prem. This is what comes out of the box. This is what makes the widgets, um, and we then the next layer up is when you go into the visualization. So, the first layer for usually for factories for visualization, they're going to dashboards and then they'll go into OEE where we're looking at, you know, the optimum equipment efficiency where they get very baseline data from the machines. Um, and then they might go and implement some rules-based condition monitoring on-prem, which might be in a skater, it might be in a In, in, in some kind of overarching controller, but that stays on-prem. And now with, with the power of cloud and the, um, and all of the benefits that come from being able to adapt um and change and um. And, and the way that you are actually processing and monitoring data, cloud is now a very much a big augmentor for what you are doing on-prem. Yeah, so, so I guess the question, Richard, is, where does, you know, on-premise has benefits up to a point and then maybe cloud takes over. So what is that sort of tipping point where. Maybe on premises does enough, but it can't do that next bit, maybe the scale part as an example. Yeah, you see this a lot with, um, you know, global organizations. All right, so it's very important that large companies that have companies in many regions in different countries, they now need to see insights across the board. All right, so they, they need to actually bring their data out of the OTE environment and start bringing it into IT infrastructure. So, this trigger to get out of OT to IT is a huge decision for customers. I can't underestimate or I can't. This is a real challenge for customers as to why should I go into the IT environment? What value are you going to bring there? But to your question is what is the tipping point? The tipping point is when you have run out of value in what you can do on prem. And this is happening now, um, quite a lot. OK, so new machines are coming on board and we've got a lot more data, um, and but to, to really leverage that data on prem is difficult to leverage that data to um really look at uh at different sites and compare and, and really leverage big data. You have to get out of the edge, uh, environment. So the trigger is usually, uh, when a customer is uh getting frustrated because they've got the data, but they're not leveraging um the, the value of it and this is when we step in and we say listen let's. Let's keep doing what you are doing, all right. Um, and supplement it with this. OK. And, and, and this is very much part of a, uh, kind of an industry 4.0 roadmap. You can't go to cloud before you've done Edge. There's, there's no doubt about that. This is, um, you, you don't just want to push all your data to, to a cloud environment and you, you, you haven't, um, pre-process your data. You, you, you're not deriving what you need to on-prem. So firstly, The first thing that keeps everything on-prem is when you've got, you know, mission critical control. You know, if, if that, um, if that motor trips, people's lives and people's lives are going to be at risk within the next 2 minutes, you've got to keep that on prem. You have to. You can't rely on a, you know, a cloud-based system to monitor it, um, and push back. So you, you, unless you've got many different redundancies in there, which we do in cloud, but I always say if you've got a mission critical application, um, and, um, the last, for example, you don't want an emergency stop. Um, going to a machine via the cloud, you know, that's, that's the last thing you want. You, you can, you can have it as a, as a, as a secondary, but, um, you've got these raw protections that sit on-prem, and that's, um, um, that we must never deviate from, and that's the power on-prem. We use the controllers for what they do, the PLC controllers, the drives, um. And the logic and we kind of step up from there into, into the space. But to answer your question, now, what's the trigger is when you've saturated the value, um, and you now either need to go and higher. 20 reliability engineers and data analysts to derive the insights that you feel you should get or you use the power of cloud and Sensei and its AI. Yeah, no, that's interesting. And, yeah, we talked, you talked about that, we'll put very, um, so very well, you know, what's the tipping point. But I guess the next question really is how can the two work in tandem together to produce the results that ultimately we want is scale, high return on investment and quick return on investment. How can, yeah, how does that sort of work? I think the, the first starting point is, um, is understanding what the customer's pain points are, OK? Before you come with a solution, it's very important that, that you listen, um, and you look for that tipping point. There's no point in pushing a customer to cloud if they can still do everything on-prem. They might just need a smaller, uh, edge application or they might need something smaller. So, The first thing that we need to do is sit down with the customer and listen to what those pain points are and then, um, articulate the value of bringing something else to the table that augments what you are doing on prem and makes it a more efficient on-prem. So if we look, here's a, um, if we look at what Sensai does at scale, so. Um, I know we talk about a lot about what Sensi am and how it works, but let's just, uh, define Sensii in this use case today as a digital reliability engineer. If that's its function in this role. All right, so here you have a digital reliability engineer monitoring your assets 24 hours a day, can move from 500 to 5000 assets. Um, with a very simple graphical user interface. So now, what actually happens is you have not changed what you're doing on-prem, but you've gotten the data into Sensii to derive insights that normally the on-prem person would have to do, uh, to go and manually analyze all of this stuff and they'll have to actually, um, they'll actually have to go and, um. You know, look at Excels and look at the on-site historians and derive it what, but as you bring, say Sensai to this picture, to this example, Sensai now tells and suggests to the reliability engineer what to focus on. Uh, OK, and uh what to do. So, the person on site moves away from Um, trying to work out what's wrong to working out to, uh, addressing the actual problems. And so, maybe I can just use another simpler example. This, this we see all the time is, um, for many of our customers, they do vibration, um, analytics. So, for permanent hardwired vibration. Analytics on site, it's very expensive. So, most of our customers only have the hardwired, um, high-speed vibration analytics on their critical assets. Um, and that's usually only about maybe 5 to 10% of their assets at the most. But there's a large amount of assets where they, they can see the value of doing vibration and analytics on it, but, you know, they don't have the budget to go and hardwire everything. So, what they will usually do is either go for every 2 months, they'll go do a manual reading. And they will go and save that data and it'll give them a red or green light, but they'll do a snapshot every 23, maybe 6 months, um, and, and then they'll work off that data, but that's still very reactive. So, what a customer might do then on-prem is that they might go and install some IoT sensors um that can send them alarms when it's over a certain point, but all of that is still reactive. OK. So, if we look at the role of one vibration engineer on site, imagine going to work and having to go and Um, manually test 50 assets, um, the value that you can actually derive out of your own value is very limited. So what, what we suggest is for those critical assets, still stick to on-prem high-speed analytics. Stick to that. We, we, we won't. Uh, we'll always suggest you that, but for the 80% rule, let's go put in an IOT sensor that's not gonna give you your spectrum analysis. It's just gonna give you your average acceleration over a short period of time because all that we really care about is the deviation um from what it has been doing over the last couple of days or weeks. And so, here we have a situation of where. Um, on-prem, we can do this high-speed analytics, um, and we can using spectrum analysis and FFT we can go and work out, is it the bolt, is it a loose flange, do I have a broken rotor bar? We can do that, um, but. That should not be all that we do. We need to do both. We now need to go, OK, that's important, but let's bring 80% of the data into cloud and let cloud go and tell, let Sensai tell me when to go and do those inspections as opposed to a blind 2 month, 3 month snapshot in time as to how that machine is performing. So, um, so this is a very good example, but it can be vibration, can be current, voltage, it can be anything. It depends what you are trying to identify. If you need high-speed, um, analytics, that happens on prem. So, as I said, if it's vibration spectrum analysis, keep that on prem. If you, if you want to look for a monitor, a large drive or uh some application that you need to know in a millisecond range when there's a deviation and then there's an anomaly, keep that on-prem, but that usually is only 8 to 10% of assets, critical assets of the customers. We need to take care of that rest. So, um, so when you bring the data into cloud now this is lower resolution. This is not gonna be your millisecond times since I can do um VA in cloud, uh, it's kind of a new feature, but um by default, I still strongly recommend get a vib uh an IoT vibration sensor in, get a deviation from the norm, let that be your trigger to when to go and interfere. And uh what Sensii is looking at is it's looking at the average deviation over an hour period. Over and then puts that into um its measurements. So, it's really looking for long-term forecasting. On-prem, you're not going to forecast six weeks into the future with 10 milliseconds time slots of data. You know, that, that's, you know, uh, the on-prem analytics is more around immediate uh mission critical stuff. Uh, but, but with Sensai, we look at the, the long, uh, the long term, longer, uh, picture. And when you combine these two, so, so my takeaway here is, um, I will always exhaust everything that a customer is doing on-prem, leverage the, all of the, um, the power of what they've done, and then using some data aggregation, let's get that into Sensai. And then let's um let the two complement each other. And when you bring these two together, when you, uh, and you don't try and, you don't think that um on-prem processing is better or worse, it's just, um, it, it's the combination of these two that you end up with a real powerful synergy there. Yeah, and it, yeah, it's interesting because in, and thanks for giving those sort of examples of when best used on-premise versus clouds. So, what I was wondering is, you know, when you, when you're sort of speaking to different customers and different organizations, is it, does it become a bit more tribal than that? You know, they're not thinking of the nuances of this use case is best for this, whereas this one's best for this. It's a black and white on-premise, uh, you know, or cloud, and that's the, that, and that's the decision. Um, I'm just wondering whether it become, generally it start or maybe starts off being a very black and white type conversation, and I guess to link to that as well, when basically when security and data ownership and things like that come into play at the same time, which adds the complexity to those conversations, I'd imagine. Yeah, definitely. So, um, and these two go hand in hand. So, um, firstly, you, you need to talk about the value, you know, so any customer that wants to take the data out of the. OTE environment, you know, definitely, um, needs to be, uh, rest assured that, you know, their data is going to be uh treated safely, securely, um, and, um, and Sensi does a very good job in demonstrating their security um out of the box. It's very much a part of the onboarding and. That um is certainly one of the biggest challenges customers have is breaking through from OT to IT um is feeling that they, that the data is going and they're losing their IP, uh, there's data sovereignty. There are many factors around cybersecurity. So, um, the only way you get a customer to look at. The benefits of getting the data into the IT environment is by the value, is by demonstrating what we can actually do, um, you know, uh, it's, it's, it's the same thing that kind of um killed CDs. Um, if we, if we look at uh what iTunes did in the very early days where. Um, where they released a digital single where you could just buy one song and you didn't have to buy the whole, the whole CD anymore where you had 15 songs that you didn't really want. Um, and Uh, and, and this is a very good example of where that value far outweighed, um, the, you know, what, what we were doing previously. It's the same in banking. If we look at how we bank today and you look at, you know, how much money you actually have, it's all digital. And um it's because You know, the, the banking sector have basically, um, shown us a better way, an easier way of handling our money. Imagine if we were still trying to keep our money on-prem, you know, keeping it under the, under the bed and, and, and, and doing cash, uh, uh, maybe that, that, that works for some, but, um, when it comes to industrial, we face the same challenges. We, um, and, and, but it's not just about. …

Full transcript (WebVTT)