Facade fires and AI with Matt Bonner

In the pursuit to change the building industry into something more environmentally friendly, and trying to make our buildings green, and I don't mean the color, make them more net efficient, make them consume carbon not released that while being built. With all these efforts, I think we sometimes lose the sight on the fire safety of them. And some technologies introduced to improve the efficiency or other parameters of our buildings, maybe unintentionally changed the fire properties or buildings, leading to some really dangerous designs that end up with bad fires. And, yeah, I think you can see that the most in the facade fire region, where it's obvious that we've seen more facade fires in the first first pages of the of the newsletters in the last decade, than in the whole length of time that I remember, it's obvious that there are more facade fires, and they're bigger than ever. And they're going to be concerned. And that's the topic of today's episode with Dr. Matthew Bonner from Imperial College London. Matt has spent a significant amount of time researching facade fires. And trying to learn from what we have at the moment. And that was results of experiments, tests done at different commercial labs using AI and some clever processing, to filter out new knowledge from already existing data. And all of that allowed us to identify where the issues may be with the facade systems. And by knowing where the issues are, we can start working on solutions. And that's what we're going to talk today. Touching subjects such as fire testing, using data to understand problems, there will be a lot of talking about complexity of facade fires, and a little bit of AI in there. So the research of Matthew is definitely exciting and interesting, and I hope you will really enjoy the chat I had with him about that. And yeah, without further do, that's been the intro and get into the interview. Welcome to the fire science show. My name is Wojciech Wegrzynski and I will be your host. Hello, everyone. I'm here today with Dr. Matt Bonner from the Imperial College London.

Hi, everyone.

Hey, Matt, thanks for taking the invite. So my friend Matt is an expert on facade fires. Matt, can you tell me how it all started?

Yeah, sure. Thank you for introducing me. Thanks for inviting me on the show. Very excited to be in the top first three, three people. That's great. Yeah, so actually, it's quite interesting how it started. So my background. I did my undergraduate my master's degree, or my integrated master's degree at UCL. In theoretical physics. My research project, my master's research project was on like high, high energy, laser physics. And yeah, when I applied for PhD, I was kind of just, I just realized, I'd like to research but I found that the theoretical physics was too theoretical. It's like, there's what why am I even doing this? What's the point? And so I was looking for PhDs that had some sort of application to society, and Guillermo who advertised this project as like, facade fires and AI. And I was like, oh, artificial intelligence. I've done some of that. That sounds cool. And like, facade fires. I don't know what you know, I've learned nothing about fire and physics. I want to learn more about fire. So I applied, it turns out the reason I think they don't teach you fire in physics is because it's way too complicated, you know, is is not the realm of physicists. It's an applied problem. So just sort of started and, and then that after eight months, Grenfell happened and suddenly, what started off as a project that was kind of a minor thing that no one was really paying attention to. Everyone was suddenly asking me a lot of questions.

Yeah, it's kind of shameful. it sometimes takes a huge tragedy like Grenfell to bring attention to important subjects in our field. And you mentioned that you jumped into the fire from physics. And I know a lot of people who actually did the jump, and they've all ended up as great fire scientists. So I think you're in quite a good company.

Yeah. Well, I actually got very excited. I know Dougal Drysdale. So obviously, you start off, Guillermo just hands you the "Introduction to fire dynamics" book. And like, oh, he came from physics. And then also, he has a paper on like, the application of dynamical systems theory to doing flashover. And that was what my master's project had been using. And I was like, Oh, this is really related. But then I realized, like, actually, there's, it turns out, it's not just dynamical systems theory. It's also really complex fire dynamics, which took me a lot longer to pick up, I think, Oh,

And what you've picked up the facade yeah. systems in fire, I don't think these are simple systems, either. So you actually had a paper about this, your first

Yeah, that was actually really interesting. paper, and I'll drop the links to the papers in the show notes. For anyone who wants to read up a bit. In that paper, you've pretty well nailed the complexity of the facade systems and, and the multi parametric issues around it. Maybe you can tell the listeners a bit about that. Starting on that paper, was basically because when I started my thesis, like I said, I didn't know anything about fire, I didn't know anything about facades, and trying to learn about it was really, really challenging, because there wasn't really a guide out there that said. This is what facades and fire are. So I was the PhD was funded, in part by ARUP, who had the foresight to be seeing this as kind of an issue. So a lot of people were interested in it in the fire team. But they all had their language about facades that they were sort of talking about, they had some ideas. They also had a facade engineering team, which had their own labels and ideas. And different people in the teams had different cut offs of like, Oh, this is a rain screen. This is a curtain wall system. This is a, you know, all these different types of facade systems. And then in the literature, there were other ones. And I was like, how do you crack in, it turns out, like, they're actually way more complex, putting them into categories doesn't work super well. So I just, I was just trying my best to like, add some sort of categories, and basically use what little literature had been done in the past, to see how you could put it together into some sort of coherent narrative, which is writing that literature view was as much for me to try and clarify my, all these disparate things that were coming at me to clarify into a story that was as much of the purpose of it as much as helping other people in the same situation.

And you're definitely not alone in being confused by jargon. And I was as well especially in faacades when the solutions that you meet at your home, are regional based on your local climate. So these things may be completely different when you go to Middle East UK, Poland or any other part of the world. I was also wondering, because UK has a history of energy efficiency, introducing new energy certifications. And and I think that this aspect of fighting for the energy efficiency of the building envelope was a significant driver towards the development of the new new types of facades. So I was wondering how this aspect did affect your research?

Yeah. Well, I'm certainly I'm not like an expert on the history of energy usage, or in buildings or anything like that. But definitely, the big theme of like, when doing this facades research of like, why a lot of the components have changed, and why there's a lot more plastics being used. A big part of that is energy efficiency. Because they want the facades to be more thermally efficient, so that you don't need to use as much energy heat in the building. And also, similarly, when we talk about cavities and stuff, and self ventilating facades, that's a big part of zero energy buildings and zero carbon buildings and all this. And they're really cool technology. But the problem is, it's new. And it's using people just try and focus on that problem and be like, ah, we can add this component this will solve it. This was sort of a lot of those new components are plastic because plastic so great. It's basically a magical material. But it also means if you're not thinking about the five properties at the same time, you can get into a situation where your facade is suddenly much more flammable than you intended it to be when you were putting it together.

Oh, yeah. And the the market is expanding and there are new solutions pushed all the time when you start considering the flammability of these things or fire properties in general, it becomes very, very awkward. And this this complexity, I mean, it's, it seems very challenging to find what is the best solution to the whole equation of the efficiency? I mean, to what extent you can worsen the fire properties. So you can justify the increase in in the, let's say, net emission cuts or energy efficiency of the building. Because the bottom line is if the building burns down, and the facade of the building burns down, it was not that sustainable in the end, right,

For sure? Yeah. And you're asking about, like, how can you find that point? Yeah, that is the big question. And one thing that I throughout my PhD, my postdoc, like coming into this field, the complexity of the problem is mind boggling. Sometimes I talk to people who are like, Oh, you know, we've pretty much sovled flame spread on a vertical surface or something. Now, we're just worried about this. And I'm like, have we that's complex enough, like fire is a phenomena is one of these, you know, complex phenomena, where the small scales affect the large scales on the large scales affect the small scales. In fact, just fluids have that property, before you even add all the thermo fluids and pyrolysis...

And a bit of turbulence in the middle.

Yeah. And then when people say, like, Oh, you need to be holistic, and think about evacuation or this stuff, you're just adding more and more layers of complexity, that all feed back to each other. And, uh, you know, complexity science as a field is very new. People trying to get their heads around emergent behavior, complex systems, all of this, we haven't really done that as a society yet. And I think like, fire, fire is still really sort of struggling in that regard. Because it's a very new field itself, is working with fluid, you know, all these sort of old calculus equations, which I think of, you know, really, really useful and really powerful. But it's not the only tool in the toolbox of complexity science. I don't know a lot of the other tools, but one of the tools I was sort of looking at is trying to utilize more sort of big data analysis, you know, that's something which people have been using to try and tackle a lot of the complexity of economics and medicine, all this sort of thing of like, okay, we can't tell exactly what's going on. Or like, we know what's happening at the small scale, but we don't necessarily know how that extrapolates to larger scales, and all this sort of thing. But we do have lots and lots of examples. And maybe we can like fit a model to that, of course, that has its own challenges as my, as I found out during my PhD, but that was where I was starting out. I was like, okay, that's one way that people use in these other fields to tackle complexity, maybe we can do the same fine.

I always found the idea astounding, you definitely changed my mind about using this powerful data that might lie somewhere in the fire testing lab, to actually try to seek some new relations seek new knowledge. So in our research, we focused to utilize this, this data from fire testing. And while doing that I've met I've met some comments that the data from fire testing is not does not have the sufficient fidelity for doing that, or that data will be inevitably biased towards some solutions. Yeah. Namely, the solutions that pass the test, because in a way the test drives the the the outcomes. So what was your take on that? And

yeah, I mean, that's a great, great question. And I have I've had similar discussions with people before. And I think it's a it's a really valid concern, because it is true, that like, these tests are low, low fidelity, they don't, they don't take a lot of precise information. You don't get a lot of precise information on the products and materials you use quite often because manufacturers often don't give out that information. Or maybe it's not been tested. I mean, in my work, it was sort of quite opportunistic. It's trying to learn what you can learn from what you have. But also, I think there's this whole element between what are we looking at what currently engineers do fire engineers doing when they like have to make these assessments, they're also kind of looking at like, Oh, I have this product in this location, this thing in this location, quite often, what they're looking for is they're just looking for some insight into, oh, if I tested this setup, or if I tested this setup with these kinds of products, you know, what would the outcome be across different tests? You know, the point is currently, the only way of testing that is by looking at a test and you can't like make a new one. You can't learn from all the previous ones that have happened. Whereas, you know, it may be that at the moment with this kind of low fidelity, you can't say, Oh, my precise facade with these precise materials will reach a high flame height of this high on the building. And you'll have this much time. But what I can tell you is, oh, it'll do this badly in this test. And then from that, from learning from the past, or predicting how it's going to behave in these tests, you can already make some judgments. And you can make some judgments before you go on, you know, it's sort of a tool to assist engineers, it's not a tool to tell you, your facade is safe, or your facade is not safe using this data is, you know, assisting people giving people more information with which to make judgments, because as I say, it's still super complex. And we're not at the point where our tools can be used without, without sort of human judgment.

But yeah, coming back to the test method, how do you think that the set limits the capability to do science? Well, with that,

I think what became interesting, when we did our own experiments, the temperature measurements, useful as they were, were actually much less useful than just the video footage of the experiments. Because looking at the whole thing, lets you spot all these things where like you say, because fire is so complex, if you've just got temperature measurements at one location, you can try and build around that location. But it's not a sort of, you know, fire isn't a homogenous thing. It's quite in homogenous, and therefore, having sort of that overall view is very important to know how something did in the test. And yeah, and why also, I think what you're saying about regulations, that's why it's important not to just look at the, they give a sort of pass fail criteria for standardization. But one of the things that in my thesis was that I didn't want the analysis to be on the pass fail criteria, because you say they're very dependent on the regulations, whereas measurements during the test, they're still going to be biased by the country, and by the way, people build stuff and all of this sort of thing, and dependent on how the test is run. But the between these tests, they should be consistent, you know, they're their objective measurements, as opposed to just oh, this is related to this is related to regulations, you know, a flame height is a flame is a change of gravity, but that's true.

But when you say it like that, I mean, it's kind of it's kind of obvious, and to believe that you can place a single thermocouple and capture the complexity of flame behavior inside the cavity??? ...And yet we just place a thermocouple and "oh yeah, that's the cavity temperature"

Well, it's quite similar with that, with the Dalmarnock. Right, where they had all the thermocouples in the mock up compartment. First of all, it was very homogenous, right. But also, when they were trying to get people to predictively model, how that was going to end up, you know, it was really, really hard to model it, even the average temperature behavior, or the fire growth, because there's so many different factors to keep into account, I think like that same complexity comes in just sticking some thermocouples in there, it's not gonna tell you a lot, what it can help to do is to give you further information, on addition to the camera, once you could also have visual behavior to back it up the temperatures, then it's possible to see like, Oh, that's probably why that's got a spike there, or why that doesn't have the spike there. You know, whereas when you're just seeing all these spikes out of nowhere, it's like, well, that could be for any number of reasons. And, you know, the fire could be doing any number of things at this point. Yep.

One thing that is in here, I always have the trouble, because many people usually try to flatten this whole discussion into a material issue. Like, I mean, your facade is combustible, so it's gonna burn down. And I think I understand that the complexity is not only at the material, like, I can probably design a very non combustible facade that would be very dangerous with the knowledge I have. And if I can do it by intention, there's someone who can do it by accident. Right. And, I mean, what's your take on the material versus system?

Yeah, I mean, I did. To be fair, in my thesis, I do mention a lot of sort of a lot of the things I'm changing has to do with the materials and the facade. And I think the reason is, and the reason we often talk about materials is because they do obviously make a big difference. If you don't have any materials that can support flame spread, or support combustion is going to make it much harder to design a dangerous facade. You know, lightly say, there are ways to do it, but it does make it more difficult. And I think the thing is what we're talking about before with the complexity of it from, from our point of view from like the, you know, from people who just want fire safety, it's the obvious solution is to kind of, well, let's just remove those variables. And then because I can't, I can't understand Currently, the problem is too complex for me to say that's definitely safe, or definitely not safe, or even what scenarios it's safe in and what scenarios is not safe. And we're not really confident in that, I think, at the moment, and that's why people want to keep it a material issue. However, obviously, these materials have huge advantages, in a lot of, in a lot of areas. And a lot of them are important, like energy, efficiency, sustainability, and all this sort of stuff. But also just in terms of, for a lot of architects, it's just really important to make beautiful buildings. And I think that is an important part of human life is that we want to be in, we don't want all our buildings to be ugly, or like just functional. So that's good, too. The thing is, obviously, that's not worth paying the price of someone's life for. And that's where we come into this thing of how can we use these materials safely? It's a really, once you start thinking about things that like the heat scale of the entirety of humanity, it gets very, very difficult to, to work out those decisions on when you can use stuff or not. And I think that complexity is why you go down to just like, okay, just don't use combustible materials. And we can feel pretty sure that we'll be safe. But obviously, it's a very conservative risk.

Not sure if you can call it conservative, like, maybe maybe you can in terms of facades and combustible materials... What's your take on the direction? Where are we heading? Because I mean, this all started, I know, when, maybe 20 years ago, when we started to introduce this complex facade systems. And nowadays, like every building, you see has this complex facade, like they truly change the system, the way how the facades are built, the way how building envelope is constructed, you can see the buildings growing, and they immediately have the facade put on place. And it's usually a very complex system that are brought from a factory and being just assembled on the site. So it takes a lot of complexity from the from the assembly, point of view, and the and the number of this, these solutions is growing exponentially. And I can imagine in countries like, like UK or in the Middle East, like possibly all new buildings will have this complex facades. So the problem is yet unsolved. So where are we heading? are we creating more trouble while doing that? Or is there any way we can limit that while doing that?

Well, that's a really good question. And I certainly am going to preface this with that. I don't know the answer. But I guess some things I've been saying, I've been thinking about doing it that sort of come into my head is, like you say, I don't think people are gonna, I don't think the industry is just going to stand by. And when I say the industry, I just mean construction in general around the world is not going to suddenly stop using any combustible material, because they are reusable. And you know, we want to start using wood and we want to start using building more and more interesting structures.

So no, no, we will also not start removing cavities from facades and yeah, when the renovation wave hits, I think double skin facades will be one of the ways to truly improve the energy efficiency of existing building. So only thing these will grow. So what to do.

Yeah, and I think the thing is, the number of these facade fires are growing over time. But there's been so far like, there have been some really disastrous fires of Grenfell being the most noticeable example. But it's usually not just the facade in that issue. There's like a number of other fire issues in that building, as well. And I think what you also notice is a lot of those buildings are buildings that were built quite cheaply, and with a lot of cost cutting and stuff involved along the way. Maybe that's a personal judgment, I can't see that's true all the time. Sure, you can find exceptions, all of that sort of thing. But I do think in the future, if we like, you know, maybe the time to sort of think of stuff at the material level and just say don't use combustibles is when you're already trying to cut costs elsewhere. It's like, Okay, if you if you're not paying attention to the level of complexity involved in this, we need to think of stuff at a simpler level. It just like don't use combustible materials. But if you're going to spend a lot of time making sure everything's packed corrugated fabricated perfectly in all the cavity barriers in place and everything's been put together to a really, really high standard, then maybe you can, and also that the other layers of fire protection are backing up. So you've got plenty of evacuation, you've got suppression, all this stuff, then maybe you can, you know, accept the increased rate,

I think you touched a very difficult topic for for fire safety, because even though there's usually one, one person that pays the bill for building the building, or refurbishing the building, the work is usually sliced down into packages, mechanical and building envelope package structural package, you usually have different stakeholders doing different things, and I'm afraid you would more often rather than finding people who would take additional precautions because others failed, I think you'd rather find like, okay, let's let's, we can not do that. So so so others can, can do something more, and these people will not like to do something more. So it's very difficult to organize the safety holistically for the whole building. And I think we had quite good success with with, with codes and standards being put in place, as a matter of presenting the solutions that work or are acceptable, we recognize that the codes and standards fail every now and then. And they definitely cannot predict the future and proactively prevent from disasters in future, because we don't know what will happen. So in your work, you take in the data from our lab, and you've magically found some relations between them assign this new indexes, found unbiased methods. I was already very impressed with that. But when I took your PhD in my hand, I was really surprised you've taken it further and introduced a new player, namely the machine learning and can tell me a bit more what the machine has found in, in my tests that I have missed.

Yeah, I mean, I think this is where things get really tricky, right with the machine learning stuff is because So basically, we trained a sort of neural network on this is this, my thesis will be out soon, very hopefully,

if it's before the podcast, I'll drop a link. And if it's later, I'll update. So keep your eyes on the shownotes.

We trained like so. So the inputs for the model was sort of like these different materials in the facade and the properties, those materials, so like the thermal conductivity of the cladding, the heat of combustion of the cladding. And similarly for the insulation, or, I mean, they weren't all also the type of facade they weren't a rainscreen facade, certainly were ETICS facades, some of them with sandwich panel facades. And they had like, and also like, say, cavity width was something included or whether or not there was a cavity, these are all fed into the model. And also importantly, because the there were also different standards involved, like there wasn't just data from your lab, there was also some data from other labs. And so we also had an input as to what kind of test was used? Was it a Polish test, British test, or the American test? and feeding all this into the model? It could then try to predict what was the maximum flame height in the test, and what was the peak temperature measures during the test, because each of these standards kind of measures, the overlap of what they measured was the peak flame height, and the temperature. Obviously, it's a bit weird, because like the peak flame height, and the Polish one is always kept the similarly with the, you know, the tallest ones, that the British one can go higher than the other ones. But that's why that that standard was an input, you know, to these things. And the the output the predict the prediction of this on blind data, the prediction of this model was like about 80% 85%, sort of accurate. And the thing is, part of that accuracy, for sure is going to be the fact that we used the subset of data we had was it small by machine learning standards, it's like only a few couple hundred examples. And that for machine learning is not a huge amount. And what that means is that this model might perform really well on the kind of data we have that might not generalize very well. But the good thing about these kind of models is you can just feed it more and more data to make it more general. As for what it was capturing. I think the thing is exactly what we're saying there about all these elements that we don't think about manufacturer, they might be biases, certain materials might be correlated with certain performance or certain, you know, cavities might be like, Oh, that's a cavity width they use when they're experimenting rather than a cavity with they use on a real building, something like that. These are all things that might pick up, as well as just the fire behavior. And that's the thing. It's not like, it is a black box, it's not going to tell you why it predicts stuff. Well, but the idea is, if you make the model, if you make the data large enough, and include enough examples, you've then got something that can very robustly tell you how it's going to perform in these different tests, when you give it a facade. And that's kind of what you want to be able to do you want to be able to learn from past results, you know, which is something I think that currently, we can't do

very well, I mean, you're British, and I don't want to torture you. But if I had the choice between the desktop study and the machine learning, I would pick the machine learning for, for some reasons. I mean, the weakness is obviously it cannot predict seeing things that it has not seen ever, well, maybe, maybe it can figure out some relations on a big enough data set. But for example, if you train it only on ETICS facades, and then you show it a cavity it will not know what what's happening, and it will have very difficult time predicting. But then again, I think it's a very, the power in it lies with with the fact that you're taking the intermediate results of the of the tests, not the pass fail criteria itself. I mean, if I ask if I if I make a survey in my lab, with 10 peoples who test these facades, so they're actually well trained biological machine learning algorithms, and I show them like you look at this facade, will it pass or fail, they can pretty accurately tell what will be the outcome perceived outcome of the test. But if I asked them, What will be the peak excess temperature, now way. And then and this, this, these methods are can really can really shine? Let me move back a bit, because we've mentioned that you've mentioned about this criteria that are not the pass fail criteria. I think you've mentioned the peak excess temperature and flame heights. Could you tell us a bit more how this were conceived and introduced in your studies?

Yeah, so in my studies, we needed sort of ways to take the results from these tests into a way that could be easily analyzed across a large number of examples. So the thing about a temperature time curve is that it's got a lot, it's very data heavy. So you know, there's a lot of trends that you could pick out of that. And you've got to find a way of collapsing that into something that can be compared across a lot of things. And this is something that's quite commonly done, actually, you notice that in like the euroclass standards they have, what is it FIGRA and SMOGRA, and all this sort of thing. And there's similar ideas of like, a value to represent what a curve looks like. And so the ones we looked at, we looked at peak flame height, because that is something that all the tests are looking at, anyway, at least the ones that we in all the standards, we are looking at, kind of their key criteria was how high the flame when that was one of their key things they were measuring. So that was very easy to use. And the other thing, they were all taking temperatures. So we looked at the peak temperature to start off with because that seemed like sort of the most critical, but How high does the temperature get? seem to us one of the most important values when looking at

local local maximum temperature or during the test?

Exactly, yeah. And then when the reason we added this excess, rather than just saying peak temperature is because obviously all of these tests are made as a obviously, all of these different standards are measuring temperature at a different height, and also from a different fire source to begin with. So the idea of this excess temperature was to work out, oh, if we if we take the base value. So excess temperature is usually used related to ambient temperature, like room temperature. In this case, we're saying like, oh, if the ambient temperature is like the the temperature where you just did a test without a facade where it was just the fire, what would be that ambient temperature? And so the excess is like how high did it go above that temperature during the test. That was the that was the idea of what we were trying to get. Obviously, it's not like it's an indicative variable, rather than like a precise variable. It wouldn't be something you would measure in a lab to be like, tell you something about the material behavior, but it's a good one for like analyzing all this kind of relative behavior, or at least we hope it's a good one, analyzing all this relative behavior with Both of these facades, so Which ones did better or worse, you know, sort of being able to rank that quantitatively?

Yeah, I think, I think that was one of the better ideas in the research to, to figure out if there are these different ways you can quantify this result results, because just comparing the temperatures doesn't say anything, like, if you say, in Polish method, the temperature was 400 degrees, but in NFPA, it was 600. We have no idea if there's the method facade, or or, or the way how it was measured. And standardizing this against the biases of the method is quite interesting. And actually, when when, when you've done that in the database paper, or looking through the facades, and using these measures to, to to filter out the, let's say, dangerous ones, or are to filter out which which would have this higher perceived flammability index, I think it was called by back then, it was quite successful, because we've picked up the ACPs and HPLs. I mean, even if you haven't had no idea about the structure of these materials, from from this database, you could see this ones, for some reason, are the dangerous ones. And in reality, in the real world, we also observe that the same is the same types of facades, we see in the news, when we're thrilled with the fact that there's, there's there has been a huge fire. We also did a funny thing, when we, when we finished our research, we realized that there was a lot of bias in this data set introduced by the fact that it was based on the Polish method. It's a representation of Polish facades, in a way, right, which are biased, by the way, how this method is used in Poland for thirty years. And we found this blank spots on some solutions that could potentially be used elsewhere in the world, like UK or Middle East, and we've actually built them and we've burned them, which has been a great experience. And I wanted to ask you, like we've burned like 20 facades, ventilated and unventilated with different types of claddings HPL, Class B class C, we had two types of ACP panels, we had some very simple setups with like, just mineral wool covered with non combustible board, and we had this crazy ventilated facades that were perceived as very dangerous. And what I mean, it's obviously very interesting to discuss the interesting differences between the tests, but I was wondering if there was anything that really surprised you while doing this, this 20 experiments like what was the shocker for you? I have one.

Yeah. Okay, I'm interested to hear you on. But for me, I think the biggest surprise was just the effect of removing the cavity. So I think you're saying we use ventilation, ventilated facades. But I think what's interesting, what was unique about these experiments, was that we didn't just try real facades that were ventilated. And like some different styles of facades, we took the same ventilated facade, and just put the cladding directly against the insulation. And so there wasn't a cavity anymore.

And it was the same material composition, right? It was the same

Yeah, same material composition. And it makes such a huge difference. Yeah, to the fire behavior, like it really slowed down. The flame spread. In some cases, in like, one of the cases it actually allowed the facade to self extinguish. Whereas before, it was, like, really, really on fire. And then in this one, it just, it went out basically, or like it went to smoldering, at least. That's the contrast between that behavior was really surprising.

For me, the biggest surprise was the fires that we've made were not really large. I mean, they were when you observe them, they were huge. They were scary. We've, I think the one of the first ones, we just extinguish very soon after ignition because we were scared It was such a huge fire. But as we grown to be familiar with these fires, and we turned off this bias, thinking about what fire is, what is a big fire. I've come to realization that this these fires are not that big. Obviously the minius of our hood is that we do not have oxygen calorimetry in it, which is a shame we could not quantify that. But from the perceived dose of heat, accumulated by by my skin. I can I can tell you, I mean, sophisticated biological heat flux measurements. But the fires were not huge. I mean, I could approach the facade from fairly close distance, I could observe it from fairly close distance. And I've done, I've done many of these experiments in my lab in this hood. And I've used gas burners with known mass flow rates. So actually, we done that on the second day of our experiments where we played with the cavities. And we've run these experiments with half megawatt one megawatt, two megawatts of fire in a very similar setting. And I remember when we did this one and a half megawatt, I really felt it from, like, 10 meters, that there's a one and a half megawatt. Yeah, well, when we were burning facades, I mean, these fires were huge in when you looked at them from the front. But I don't think that the heat release rate was that tremendously big. And we'll jump into the visual method. But I think you actually quantify that and it correlates to my the feeling of my skin. And what I want to say, I mean, from one side, it's really interesting and exciting that these fires are smaller than you would expect, which I would expect. So that makes me happier. Because if a fire is smaller than maybe it's less strength, but on the other hand, the temperatures that we've measured the damage done to the facade, the scale of destruction brought by this fairly little fires, was astounding, like, if you consider that it was just, let's say, 500 kilowatts or 300 kilowatts that destroyed this big sample, great temperatures in cavity in range of 1000 centigrade. And we had flame extensions for like two and a half meter above that. That's kind of frightening, you know, because, because this, it means it's, it doesn't have to be huge fire, to be huge for the facade. That's my point. You don't have to have 20 megawatts, venting on your facade to create conditions that are immediately dangerous to the facade itself. So that was kind of my shocker that I mean, these fires were not as big in terms of heat release or resize. I mean, they were two dimensional in the end, in a in a way.

Yeah, I agree with that. The reason they build these facades is because they're very thin and lightweight. And it did mean that the fire was kind of restricted to these kind of very thin regions, they can have burning, which is why from the front, if you just looked at that, you're like, Wow, that's a big fire. From the side. It's like, Oh, that's not such a big fire.

From my perspective, it was like from the front, it was like, I think we should extinguish that. And from side it was, can I stick my iPhone inside the cavity and survive? Which I kind of did? And yeah,

not during the test, we will have the experimental results thing.

Now, but but there's the thing is, like, kind of, it's kind of two dimensional, which is, in a way fascinating, because that's another five behavior you would not normally consider, like, you're not learned about this in school.

What was interesting, the three dimensional though, which wasn't really planned for the experiments, because we haven't really talked about how to make the facades really flushed like you wouldn't, you might have a normal test, you know, or you might in a real building, you'd have the insulation, the cladding, and the fixing, and the external wall, as like a flush piece of material. In this, we've accidentally created a channel between like, the, it's hard to explain without a photo, but it's like, we had the brick wall, and then the cladding. And then we had sort of this side channel, where there was like, the insulation. And we thought, that's fine, because that's just the side of the facade, but during the test, because like, the wind blows the fire about in a couple of the tests, that channel ignited, and it just went right up, because it was in the channel. It really, it was unexpected flame spread behavior. That was that was quite three dimensional, because it was in a level that we hadn't really considered as like, we were imagining this sort of 1d problem. What do

we what you mentioned are the vertical like, separations we've introduced by by building the structure of the facade.

Yes, yeah. Yeah, exactly.

That was fascinating to see how it's separated into sections that could burn rather independently from each other, often simultaneously, but there were some differences. Between that that was definitely very, very interesting. And we've also used like combustible, non combustible insulation materials. And I mean, there was there were these combinations that included combustible materials, which, I mean, looking at the whole project didn't add that badly. Right. And we also had ventilated facades with non combustible panels. And that led to significant damage of the panels in the experiment. And I mean, it may not be that dangerous in terms of spread of fire, but the damage to the facade did exist in this test. So, so I think it was another statement of complexity versus materials. In experiment itself.

There. Yeah. The debris from the we are using some import, it was basically spalling. Right?

Throwing a lot of it was raining.

Yeah. Which would be interesting. Whether that would be like, it's the sort of thing we would it be a risk, or, I mean, they wouldn't use just a cement board, I imagine sort of rainscreen cladding. But yeah, I there was that sort of complexity, and also the complexity of like, in one of the tests, I guess in a lot in the test was sort of flammable cladding, it was kind of like the insulation, the cladding, it was hard to separate, you know, where one's burning and one's not or anything like that, that was part of the difficulty of analyzing these results. But in the test, where we have the cement board, and then foam insulation, there that you were getting this sort of smoldering going on in the in the cavity, right. And it was slow, but it was like, definitely spreading the smoldering through the facade, it was burning for all, you know, how we left it for however long it was 45 minutes, and it was just still still going. And interestingly, it was then igniting at the top of the cavity like so as soon as all this gas met some oxygen at the top, it was like, then catching fire. So it's definitely hot enough to be burning in there. But there wasn't enough oxygen getting, which is interesting.

There's definitely a lot of interesting things happening when you burn this. This thing's in full scale, and you always learn something new and somewhat unexpected. I've also really enjoyed the way how the data processing was was performed. I found the method of estimating the heat release rate from cameras, quite impressive and and starting to, to be honest, and I think the audience would love to hear about that idea. Because it's, it's it's really powerful. You've called it the visual heat release rate. And what you did is like you've estimated the value of the heat release rate based on the volume calculations on the flame. Can you say something more about that?

Yeah. So we could the visual firepower. And the idea, I think it was when we were just in the fire dynamics textbook they're mentioning, or you know, volume is correlated with heat release rate in a turbulent fire. And then, again, was like, Oh, you could investigate that. So I did look into some more research. And, yeah, there's this paper by DeRis and and Orloff, where they're sort of going like, Oh, well, in a pool fire. If you do Froude modeling, and you correlate like the flame height to the diameter, and then assume it's this shape, then it will, the volume will be proportional to the heat release released rate. But then there's also been there was then some work done in 2003. And 2005, by these guys, Mason and Stratton and Michael Spearpoint, I think was overseeing it, trying to do it with cameras. And basically trying to do the same thing of like estimating volume, use that to estimate heat release rate. And we're like, oh, let's try and do that. That sounds really useful. But I've like the kind of the work they've done was very sort of the beginning of it. And they hadn't provided code that I could use. And the methods they had were like, way too, Time. Time slow for like it was it would have taken me weeks to run experiments just using the sort of explanation they had given. Obviously, they might have done some stuff when they actually did it themselves. I don't know, I could, I didn't ask them. So just sort of developed this sort of different method for but inspired by that To try and measure this release rate, or estimate it by by measuring the volume of the flame. So by using these two orthogonal cameras to estimate the flame surface, averaged over time, and I think that's the key thing in all of these examples. It's a it's a time averaged volume and the reason for that and I was talking about this with me. And because we're talking about it, because the thing is the reason why volume is correlated with flame height is to be honest, a little bit of other than this Froude modeling, explanation. There's one other paper on a theoretical explanation, which to be honest, I didn't understand, because it was very high level fluid mechanics. You don't get that, as a physicist, we don't do a lot of fluid mechanics, he was talking about, like, Rayleigh-Taylor instability criterion or something. But essentially, the heat release rate should be more to do with the surface area of the flame. Right? That's where the reactions happening. The problem is, how do you work out the hot? Like, where's the surface area of the flame? And so I think that's the key thing is if you time average it over, like a long enough period of time, that's going to even for a turbulent flame, that random behavior is kind of going to average over to the volume that you see is like that flaming region

that makes sense. And they should nature is very efficient and nature would not create the shape with very bad surface to volume ratio, it will, it will tend to self optically eyes, in a way. And would you say that the surface where the burning is happening? That's obviously true. I know that because I've sticked my iPhone into inside of the flame, and it was hollow inside. Yeah, I confirmed that by by by throwing an iPhone through a campfire in the slow mo. So I think it's very fascinating, because you're pretty much taking a fairly easy thing to set up before the experiment, and I mean, visuals, cameras, and can create a powerful output, which is some sort of evolution of the heat release rate over time. And by simple I mean, it's it's very robust it to create a good measurement of temperature, you need to know, where do you expect your flows to happen. And believe me, countless times, I've faced the issue of trying to measure a fire phenomenon just to figure out it happens 20 centimeters to the left, and my, my measurements need to be repeated once again. And with the with the video that covers the larger part of your, of your assembly, it's robust to do that. That's that that's really powerful. If we have some few minutes left, I really wanted to ask you about your future plans. Because I mean, the PhD is great and everything, but what's the next step in in the facade research? Are you switching? Or you're switching back to physics? Because the fire is not working?

Yeah, no, we, we did. We, we've got a project to work on at the moment. This, which is why I'm still at Imperial. That's what's got this postdoc funding to do this project. It's essentially, we call the project BRIGID, which is after the Celtic goddess of fire. God, so fire, that wasn't my choice this time. That, but it's a bit it's a fun name. And the idea is it's like to compare different, different standards. So what we're talking earlier about, like the fact that all these standards, all these facade, test standards, use different kinds of fires and different facade setups, and different failure criteria, and all this sort of thing. It's a bit of a minefield, in terms of like, if you ask in trying to ask why people do things their way, everyone's got a different answer. And you know, a lot of people just think, Oh, well, if we do my test, the facade is safe. If I pass this test, it's safe. If I don't, if I, you know, or other people are thinking, Well, why do I have to use this test when this one's cheaper? You know, there's all these people people have a lot of opinions, but there isn't a lot of data available. To actually back up these opinions. There's not really any experimental series, comparing the same facades across different test methods. And so that's what we're doing. We're doing five different facades across five different test methods. And we're just testing the same facades, same setup on these five different methods. And we're just going to see what happens. We're going to take measurements with our cameras again, we're going to add thermocouples. And we're just going to, you know, try and look at the not other pass fail criteria. I mean, we might do some analysis of it, but like, really with we're just trying to see is the behavior due These tests consistently rank, you know, these facades. So if we have facades at different levels of, or expected levels of flammability, do they come up with similar similar ranks across all these different standards? Or do they have completely different behavior in these different scenarios?

Is Polish included? I'm scared now I'm scared now. Because our it's actually it's at the same time scary and fascinating, because this is scary, because it's, our method is probably the smallest one. And I'm not certainly sure how the scale will affect the outcome. Like if it's possible for a facade to pass our tests, because it was not big enough to like, combust and, and and fail, maybe larger facade of the same material has this critical mass, let's see, if he could he can say that. Maybe it will behave differently, on the other hand, is kind of fascinating. If all of them are ranking the things in the same manner, then why? Why use nine meter? test? Yeah, do that with it with two meter? And or maybe you could use a smaller facade to the screen the solutions in a way that I mean, it's a fascinating idea, and I hope it will be a great piece of science. It's certainly very interesting.

Yeah, at the very least, I think what we hope from the outcome is, it will at least allowed because I think there's a lot of discussions going on now, definitely in the UK, about what to do about facade tests, what to do about facades, in general, what, you know, and then harmonized in the European standard, all this sort of thing. I'm hoping at least the experiments help to contribute to that discussion with some, you know, with some data that isn't opinionated. That, you know, we don't have an agenda behind the data. We just want to see what happens.

And new designed, European methods are not in in the in the

well, no, I, because I think it's not been fully not 100% released yet. Right. I think it's still discussed at the moment, but it's heavily based on the British and the German methods. And both of those methods are included. Okay.

The European one is definitely under development. And, I mean, it's a great idea to have a harmonized standard for the for the facades. Okay, man, I think I've reached the limits of your, of your generosity with time and it. I still have some questions left on my list. But let's let's leave them for a follow up interview. There will be a part two, definitely. And when it when when you burn this, all of these facades, I will definitely love to welcome you back in the show to to discuss if, again, there was something shocking that you have found. Yeah, if people want to learn more about your research and your research group, what where should they seek information?

As a great question. If you want to look up our research group, Imperial haze lab, you can find us online. We've got a https://www.imperial.ac.uk/hazelab. I think it's the web page. But you can just Google it. That works too. If you want to follow my research in particular, I've got LinkedIn, Twitter, research, gay, that's MattBonnerFire on all of those things, or https://www.linkedin.com/in/matthewbonnerfire/ on LinkedIn,

I'll put it in the show notes so people can track you.

But yeah, you can find my links there. And if you want specific stuff, yeah, feel free to email me. If you just Google Matthew Bonner @ Imperial College, you can find my contact details. And like at Imperial, like my Imperial email, and I'm happy to answer questions, I'd love to love to. Or if you've got suggestions, please throw them in.

Fantastic. Thank you. Thank you so much for agreeing to do that. I hope you enjoyed. I've enjoyed it a lot. And I'm sure the audience will, will really like what you had to say on the facade fires. And yeah, all the best man. Thanks.

Yeah. Thanks. Thanks for having me. It's been great fun.

Yeah, I hope you enjoyed this episode. And Matt is a great guy, and I really like working with him and discussing facade problems with him. In this episode, there's so many different aspects of the safety of facades from the way how we test the facades, through seeking knowledge inside some tests that have been already done. touching so many aspects of complexity of facades. And the the back door of, the kitchen of how we are doing the facade tests. And actually how this, the way how we do them, is reflected in the structure of modern facades, that that's actually astounding. And there will be a follow up with Matt definitely, as he just told you, the Imperial is planning a new series of experiments with five types of facades, and the good news is that we may be doing with them. So that makes me very happy. I'm always looking forward to to run some large scale experiments with interesting people. So yeah, thank you for being here. And listening to this episode. After the first day, or the first week, the podcast has aired it. I was astounded with the reception of of the podcast, amount of views and everything. And I'm really thankful to every one of you listening and I truly hope that as we switch from the hype into the routine, I'll be able to consistently provide you with great contents every week. Oh, and one more thing. For all new episodes, I'm gonna start making your LinkedIn threads, in which we'll be able to discuss the contents of the episode. And if you have any questions to the guests, I'll encourage them to answer you there. And I'll pop the link to the thread in the shownotes. So it's always easily accessible and maybe this way we can continue talking, despite the episode is ended. So thanks for being here and see you on the next Wednesday. This was the fire science show. Thank you for listening and see you soon.