#102 Data analytics vs. fault detection

Buildings of Tomorrow ·

Show notes

Dear Buildings of Tomorrow friends, this will be the last episode of the podcast until next year. Who else than Tyson Soutter, our data analytics expert, to talk about the difference between data analytics and fault detection? Enjoy this last 2020 episode and have great winter holidays.

Transcript

Hello and welcome to Buildings of Tomorrow. My name is John Lester, and in today's episode we'll be talking about fault detection versus data analytics. I'm joined today by Tyson Suter, the global business development manager for digitalization at Siemens Smart-Infrastructure. Tyson, thank you for joining us. Thanks John. What's the difference between fault detection and data analytics? Yeah, so fault detection we've seen, you know, for a very long time now, it, it looks at point in time data. So, what, what I, what I, what I mean by this is, If something happens and something goes wrong, alarm. Tell me exactly at that moment. OK, this, this, this happened, this isn't what we expect. Send off a flag saying good or bad. So we know what we expect, we check to see if we see it or not, and if we don't. All the alarms go off. Yeah, exactly, like we, someone's swiped their card and the door's supposed to open, because they have access. If that doesn't happen, that's an alarm, or maybe they swipe the card, they're not supposed to go through that door. Uh, it's very simple stuff all the way down to, uh, you know, expected temperatures versus, um, expected movement of people or very, very much binary on off, yes or no. So we preset our expectations, we preset what we would like to see and a, and a sequence of, of events, and if those events don't correlate specifically with what we've defined, we have a fault. Exactly, and. Very valuable as well, but we don't want to underplay how important it is to, to measure expected values and, and see the results. This is the, you know, significant portion of how we run our buildings. So I think analytics and, and, and bringing the two subjects together is. Saying OK fault detection's important, very valuable, we need to do it, we need to do it in a smart way. Analytics is, how can I look at large portions of time or data and say, Could this be better? Is this performance as expected? Is it regressing from what we expect to see? Is it a drift from the normal? Can we predict what should happen and can then we say, oh, we could do better here. So not looking at one specific point in time, but actually looking at the history. And understanding how often something happens or what's regular interval is causing this, which might lead you to more information. Exactly, so you know, predicting er something that's really valuable is pressure leakage, saying oh OK that's, we expect a certain performance in these conditions, with a bit of data and a bit of analysis on past performance, we can know what to expect to see in certain conditions. If it's certain temperature outside and a certain uh uh heat volume within a building, we can we can say OK, we expect it to perform this way. With a huge amount of data we can say great, we know what to expect, it's not doing that anymore, that's an issue. Let's go there before it's 50% uh failed. Let's go at 99% when it starts losing that kind of very valuable refrigerant. Yeah, right, so, so having a better idea of understanding what happens leading up to a fault occurring, so that you might be able to catch or prevent that fault completely. Exactly, and when you bring the two things together, it kind of really enables to run a building efficiently, but we've talked about this in the past, but if you can. Combine those two really valuable methods, and then prioritize what's important. That's where the true value comes from, making sure that you hold people accountable and and close issues off so we can get value in our buildings. Analytics is really nice to have but if your building isn't running effectively at all, Well you need to stick with fault detection, make sure you can get it to a like a baseline, and then look for the improvement. Right, so you need to get to a level, you don't have so many faults that it's impossible to see, like we would say you, you can't see the, the trees for the forest or the forest for the trees, uh, so you have to have an operating building, you know, within a certain reasonable response so that you can then start to apply a, a more considered approach. Yeah, and if we don't have any data prior to connecting, that's going to be the worst day. So we do need to create, Uh, a data set to be able to analyze like what is good. You know, we have other buildings and other equipment but there's many variables in every building, so the most valuable data we can have is from, from your site. So we connect up and say this is day one. We're not gonna be able to predict expected performance for this piece of equipment, but you know, a few months in we're gonna be able to do that. You mentioned before as well responsibilities, so you, you have people operating a building, there's always a human element, not only to bring their expertise, but also on a day to day basis, they have to continue to continually operate and and monitor, and when we see faults, Happen, you know, in, in a certain point in time, often it's the responsibility of one of these people to acknowledge or something like this. Does data analytics give us the power to expand our view beyond acknowledging a single fault? Yeah, so there's also the root cause analysis where you can get rid of duplicate alarms, but, Also just smarter ways to look at the alarm data and say like we don't need to. Acknowledge everything, cos that doesn't solve any problems, but what we can do is say, well, we know this is significantly causing us large issues, these are our critical alarms, solve them straight away, but here's the big cost items, it's either gonna be a big risk in, um, tenant comfort or maybe it's gonna be a risk in energy or performance, asset damage or something else. Exactly, we can say great, let's just prioritize these first and, This stuff down here, not that important, cos there's a lot of alarms that, you know, yeah, it's not working perfectly, but in reality, let's let's worry about that later, let's focus on the big, big ticket items. And that's what you can do with a smart system. You can bring all these things together, put it in the hands of experts and say, great, this is important to us, here's the potential savings, here's what we need to do to fix it, let's get the right people there to do it. Great, thank you very much. Thank you. And thank you for not just listening to this episode, but to all the episodes throughout 2020. It's been a crazy year, but we've very much enjoyed from this perspective bringing you so many exciting topics and so much knowledge from our exciting experts. This is the last episode we're releasing in 2020, so we'll see you early in 2021 with a lot more new episodes with new topics, new experts, and a lot more conversations around this exciting industry that we work in. Be safe and secure, and we'll see you again soon.

Full transcript (WebVTT)