#148 Leveraging Data Analytics and AI-based technologies in buildings

Buildings of Tomorrow ·

Show notes

We experience AI every day, from Siri all the way to AI-powered cameras. Yet when we look at buildings and AI together, there is still a lot of work to do. How can we leverage these capabilities, given that people spend roughly 90% of their time in buildings? Tune in together with Dr. Paul Baumann, Head of Global Product Management Data Analytics at Siemens Smart Infrastructure, and Jon Lester, to find out!

Transcript

Hello and welcome to Buildings of Tomorrow. My name's John Lester and in today's episode we are talking about AI and data analytics in buildings. I'm really excited to be joined today by Paul Bowman, who is the lead for product management for data analytics at Siemens Smart-Infrastructure. Paul, thanks for joining us. Hey John, thanks for having me. Mate, it's an absolute pleasure, and today we are talking about artificial intelligence, we're talking about data analytics and how these two technologies, services, uh, value propositions work together. And we're talking specifically around buildings, you know, we talk often in our industry. Buildings are such an important part of our lives. We spend 90% of our time in buildings. And while we use artificial intelligence based services, we have technology that that leverages these capabilities every day. We sometimes see in the building world that uh that we still have some steps to go and some more opportunities to come. So I'm excited to talk about that with you and get some more details about what are the next steps are. But before we do that, I wanted to start at the start. When we talk about artificial intelligence and data analytics, give us an executive summary. What do we mean in this context? What, what do we talk about when we mention these terms? That's absolutely true, John. Uh, thanks for asking that. I mean, we talk a lot about AI, machine learning, analytics and these topics, and actually, we're all carrying these things around, products like iPhones, uh, smartphones in general, laptops, they all use some sort of data analytics AI to do certain type of services, but what exactly is it? I mean, From the word perspective or what AI is, it's a basically a subset of an idea that a machine can learn and basically imitate or mimic intelligence which we humans take as natural. And one of the subsets of AI in general, the idea is basically machine learning, where machines take data as an input and try to understand the context in which the data has been generated, try to learn all of that. You can imagine like a human being where when we grew up, we observe, um, nature, we observe how our parents behave, our siblings, friends, and take us a data input by our sensors and then try to understand, OK, how do I need to behave? That's a similar concept where AI in particular. Machine learning is, uh, looking at how can we use data and basically, um, try to mimic this behavior, to take decisions, to understand the context, to reason what is happening around us, and then basically to contribute to society by having this kind of, kind of artificial intelligence. I mean, obvious, there are a couple of prominent examples which we see beyond building industries. One of the hot topics currently is really with some kind of self-driving cars, where the promises that AI could help, uh, to save over 1 million deaths annually by, um, releasing the need for people to drive on their own by taking care of this. Today already as the doctors are using AI and machine learning, particularly to understand, uh, to scan the, uh, different MRT scans to understand what are potential things that they might not be observing and to interpret that and help them to come up with the best, um, methods to treat the patients. So at the end of the day, AI and machine learning, in particular, coming into place where people are simply limited into processing a lot of information because again, today, more and more information is getting generated. Um, there are tons of devices that can capture the different sceneries and AI helps you to process the data, but also to detect things that humans cannot simply observe or we're struggling with. So I think from a general perspective, these are the two areas where I see in particularly AI machine learning can do a lot of benefit for the people. And I think you've pointed out already, so how does it link to the buildings and, uh, where do we stand, I think, as in terms of the buildings. Um, I mean, I think that is a big opportunity for us with the AI to enter this market and because um in the past, buildings has not been quite leveraged in this technology and there are a lot of opportunities, I'm more than happy to talk about that today with you. Yeah, bring it on, and, and that's what I'm excited about because, because as you mentioned, you had some great examples where, where other parts of our daily lives or uh you know, other really detailed and specialized industries are using this technology really effectively. But we don't quite see it as predominantly as we would expect in our industry, you know, we talk about something, you know, crunching a lot of data, finding understanding and meaning within data that we can't recognize on the surface from a visual or a human perspective. And, and here the the building industry is ready for this, you know, data we have a lot of, you know, we collect information from everywhere all the time and we're getting better and better at, at doing this, not just collecting but normalizing this and bringing it into a space where we can start to make a positive difference. But where are we up to? because the technology is there, but we don't see it every day yet. From your perspective, what's our status from a building industry side of things? I, I think, uh, the biggest challenge usually in general with AI and particularly with building domain, it's how do I link my needs to this technology. So how can the technology solve my needs? And that is simply because, um, at the end of the day, we need to, we cannot simply apply AI and machine learning and say, OK, let's look at what's going up. We need to understand what is the problem and what to solve. And like you said, um, it, it all starts with data. Yes, we do have data already for decades where specific systems are installed on site, they can collect it to operate and so on and so forth. But the challenges are that in the past, all this information has been isolated. Usually, in buildings, you have multiple systems that have to work like a human being, like a human body. You have multiple organs that need to collaborate, that need to work together with one specific goal. In the past, all these different systems and data has been isolated. They were not or barely communicating with each other. People could not get access to that to really understand, OK, what is happening there. They might have seen a couple of visuals and that's pretty much it. So, and that also is also one of the limiting factors were uh the people were simply struggling to also interpret it with amount of data. I mean, just let us look at an average building. It might have thousands of different data sources and there is no absolute way a human being can process that. And that is where opportunity come into place because we know uh buildings are super great, big, big opportunities as for us as a global player in this domain to um make it more sustainable. I think that is one of the hot topics today. How can we make buildings more sustainable? Because, um, in the past, buildings were built to just rent them out and sell them. Nobody really cared about this topic because simply as the solutions were too expensive. You needed to throw a lot of people on the problem to solve that. And today, with opening up all these data silos, we can finally learn models, learn, uh, machine learning algorithms to solve that for human beings. So sustainability is definitely one of the aspects which you can address. Um, and that is a big one because as we know, 40% of CO2 consumption is directly linked to buildings. So there is an absolutely big market growing up now. Um, the second aspect is simply the entire operation efficiency. So how can I use data and buildings to run it more efficiently? Over the entire life cycle. I mean, buildings typically, they are there for 40, 50 years, and how can I minimize my cost on that? And I think the first, uh, um, aspect is which we see in the industry popping up more and more recently in the last years is the entire aspect of human beings and well-being of people who are inside of the buildings. I think like you said at the beginning, 90% of the time we do spend in buildings. And if you look at the commercial side, we spend, I think, 25% of the time in commercial buildings like offices, like I'm doing today. So it's a big impact on my life, on my health, and if I'm not feeling comfortable sitting 89, 10 hours in the office, it's gonna continuously contribute to my health, and that is, I think, the third key pillar where we need to address. And I do see a lot of opportunities here. I use, uh, artificial intelligence to address that. Understand. So you mentioned a couple of big goals, and what I liked in the early part of your, the last piece of the conversation, you mentioned artificial intelligence, data analytics, or any kind of technology doesn't solve problems for us that we aren't aware of. We, we need to know where we're going, what are we trying to achieve before these technologies offer us any value. And then you mentioned some key topics there, sustainability, uh, operational efficiency, and also health and wellbeing, user, user well-being. That kind of define for us, and we see this pushed by society, we see this pushed in the industry, define for us these problems that we're trying to solve, uh, and then you touched on a couple of these pieces which I want to dig in a little bit more detail is then how do we make it happen because the technology exists and, and we now understand better and and probably driven more by society, have a more aligned goal as an industry where we want to go, sustainability, well-being, operational efficiency. What's the next step? How, how do we get to the next phase where we start to solve these big problems, leveraging this technology? Mhm. Yeah, that's absolutely a good question. I think now, um, in the last years, there's more and more understanding, um, in place how this technology can help people, like I was saying kind of these three key buckets which I do see. Now, if you want to do that, what do we need to do? And for me, the obvious step, like I said in the beginning, it's all about data. So that is the starting point. How do I get data? How do I ensure that I'm not having kind of this isolated view of the data? How do I get the holistic view? And it's not just getting a connection and extracting it, and that is particularly when we deal with machines, they need to also understand what is the meaning behind this data. So, if I'm looking at maybe a graph and I know the context, I can say it's probably a temperature of the room. Machine has no idea about that. So it, we need to be able to not only get the data but also to enrich it with semantic information, with metadata to help a machine to understand, ah, OK, that means that is the temperature, that is the CO2 value, and that is how it is linked together. So really get this holistic understanding and to train, uh, then the technology with this understanding. So that is to me the step number one. The step number 2 is obviously, I mean, at the end of the day when we look at the industry, it's super labor-driven. We hear a lot of uh the uh shortage of workers. Um, just today in the morning, I read an article that's where again, a big shortage of workers in Germany in the construction industry simply because, uh, there's more and more needs to be done and less and less people entering this industry. So how do we scale it up now? And then point number 3, I mean, just look outside of a window and try to find 23, exactly the same copies of the same building. I guess you're going to struggle to do that. So the uniqueness of the buildings and the installation within the buildings, the way how it is used, it, it, it is one of the additional challenges that are contributing to that. So we now need That is exactly where AI comes into place. How do we, by getting this data, by enriching it automatically, how can we now deal with this uniqueness of the buildings and um basically how the buildings are structured? And that is where exactly AI and data analytics are coming into play, especially machine learning, where we can say, hey, instead of I'm going to tell you how you're supposed to run your building, let's put data in and it's going to learn automatically and you don't need to care about. And that is actually the beauty, how we can scale up these aspects and visually connect the technology to the purpose of understanding how do we decarbonize the buildings, how do we run them efficiently, and how do we improve the comfort in these buildings. I understand. Now, you touched on some really key, key topics there around. Getting data, organizing data, but importantly, ensuring that that data is enhanced, metadata, semantic tagging, this kind of stuff, so that the artificial intelligence, the machine learning process understands what this is, because data is without context is useless. Uh, you talked a second about scaling up and really bringing this into the market, getting the right people, and that's a topic I'd like to dive into a little bit more. Uh, and just before doing that, you mentioned also the uniqueness because every building is different somehow, whether it's the function of the building, uh, the expectation or the goals of the operators and the owners of that building, or just the physical space, uh, and geographical location where this building sits. Let's talk a little bit about skills because for a technology like this, there's somehow a a a. A combination of, or, or the marrying of two sides of the coin which are required, you know, we have the skills to deploy the technology itself, and then we have the skills and the domain knowledge that we hold within the industry, and you need both of these to really solve problems and understand problems effectively. How do we get over that challenge, because I think we can safely say, in general, the construction industry is not a fast moving one. It's not, it, it doesn't tend to be an early adopter from a technology perspective. So how do we bring these two sides together, your data scientists, your, your machine learning experts, and the domain experts and those that hold the, the knowledge about the, the problems we're trying to solve? How do we bring them together? How do we get over that hurdle? Yeah. Uh, uh, that's absolutely a great point. And, uh, that is what, uh, basically, my team and I are looking at almost every day because exactly we see, OK, we are better from technology, we can do that. Um, it's existing, we're getting data, but how do we start now to break it down? So what does sustainability mean? Where do we need to improve that? How can we improve? What are the more fine-grained use cases? And like you said, um, the industry itself, it's quite conservative. It's by nature. And, uh, if, uh, I think there was an article on HPR a couple of years ago saying the productivity in this industry in the last 50 years just improved by 1% regardless of all the technology. So I think there is a big opportunity now to accelerate that. And the question is, like you said, how do we do that? On the one hand, obviously, you need to bring these professionals together. You need to get the people who understand the domain, who understand how the heating ventilation cooling systems like uh for the comfort or fighter systems work in a building. So you need to get connection to the people who are uh in the industry for 1020 years who know exactly how it runs. And at the same time, you need to bring people in who know how to use data. On our team, we have people who have studied astrophilology or, I don't know, nuclear power whatsoever. They have no or limited exposure to the building industry, but they know how to crunch data. So how do we bring these two together? Uh, one of the things that we are trying is really kind of, uh, closely working with them to understand what are their pinpoints, to understand where we can run a certain type of short proof of concept implementation, and then simply show to them what is the value. We've been doing that with a couple of uh POCs in the past where we said, hey, let us look at that one specific problem. Let's collect data, let's analyze, let's implement that, and let's measure the impact. And once you start showing that beyond simply, hey, let's do some AI, but really showing, hey, that is how we connect technology with purpose to the actual purpose, so that at the end of the day the users not only see, oh, they're doing some AI magic or machine learning, but really see how it's contributing. I think that is one of the key elements to really make this end to end dot connection. So that both parties can get educated and understanding that the data scientists understand, ah, if I'm trying to optimize now a chiller in the basement, it's going to have a more impact on the energy consumption than if I'm trying to optimize, let's say, a fan in the room. And at the same time then those more people coming from the domain, they understand, OK, if we can also maybe install some additional sensors to measure these things, it would help the other guys to get the data to understand and to optimize that without the need for me sitting in front of a laptop looking at the graphs and trying to figure out what is going wrong. So that's our approach on how we are doing that because like you said, making this connection is crucial, otherwise it's still going to be a buzzword in the room. Yeah, I understand. And hey, not just making, as you say, making that connection is crucial, but also what you just described in, in, in that last answer is also an important thing for us to recognize. All of the topics you mentioned before, uh, you know, user, user wellbeing and health, uh, within the space, uh, operational, uh, efficiency and sustainability, all of these are big topics, these are huge topics made up of many, many small problems. And having that focus on solving individual, you know, connecting those different points, solving one problem with technology, a small problem turns into a bigger solution, turns into a significant impact on these large topics. I think that's an important thing for us to remember as well, right? Like we're not going to sit here and try and solve sustainability. That's not the goal. It's to solve. Many, many of these small problems that are really focused and specific that we can then apply this technology to which gets there. I is that a, a reasonable comment from my perspective that, that, that this is how we go about this as to, to chop and find those areas of highest impact and focus. Absolutely. I mean, as I said at the beginning, there are so many moving pieces in the building. You cannot simply apply one technology and say, hey, we're going to solve it now. You really need to deep dive and understand, OK, what are the small pieces contributing to this overarching goal, and that is exactly where domain is needed, but at the same time, people are needed who are going to tell you, OK, if you want to solve that, that is what I'm going to need for you. And at the end of the day, it asks. To, I'll say, um, coming from a computer science industry, uh, the difference to traditional development where you have your goal, you know how functionality is going to look, you know, the future, you implement, you're done. With AI it's a lot of trying things out. It's a lot of formulating hypothesis, looking at the data and saying, hey, can I use now this information uh to solve this problem? And you might fail 9 out of 10 times. Because that is how this uh discipline works, because at the end of the day, it's really about exploring data, understanding, trying things out, and then basically coming up against that game, that is probably the solution that's going to help you. …

Full transcript (WebVTT)