134 - Fire Fundamentals pt 5 - The Evacuation Equation with David Purser

Hello everybody, welcome to the Fire Science Show, welcome to the year 2024. I hope you will have a brilliant year and the podcast will be an important part of that year. For the first episode of this year, I hopefully have a real banger. I have once again invited Professor David Perser, this last year's recipient of Howard Ammons Award from the FSS for his lifetime achievements, and Professor Perser was here already multiple times we have been discussing how he has revolutionized the field of toxicology and toxicity in fires and our understanding on how smoke affects human beings and how do we measure smoke. In today's episode I wanted to reach to the other side of his career where he has also kind of revolutionized how we consider the timeline of fires and how we consider the evacuation processes. For most of the fire engineers, considering evacuation processes in fire is a very simple thing. You calculate the available safe evacuation time, you calculate the required safe evacuation time and you got it. But if you think about it, it was not known for the only history of fire science and it had to be invented at some point, and Professor Perser had significant contributions in establishing this and also establishing in stuff that we know as evacuation timeline the plot in British Standard that everyone knows. That's the cover of this episode. It's something that David has participated in creating and I'm super thrilled to have him in the show discussing how this came to life. It was supposed to be a simple interview episode, but I've kind of upgraded it into the fundamentals of fire science, because that's the purpose of the series you learn the fundamentals from the most knowledgeable people in the world of fire science, and I would doubt that there's anyone more knowledgeable about the stuff that we're talking about today than David Perser. So I guess that that's enough, and I hope you join me and David in this very interesting conversation on the safety equation. So let's spin the intro and jump into the episode. Welcome to the fire science show. My name is Vojci Wigzynski and I will be your host. Fire Science Show is brought to you in collaboration with OFR consultants, a multi-world winning independent consultancy dedicated to addressing fire safety challenges. Ofr is the UK's leading fire risk consultancy. Its globally established team has developed a reputation for preeminent fire engineering expertise, with colleagues working across the world to help protect people, property and the environment. After one year of working together with OFR to deliver this podcast to you, we have extended our collaboration for the entire year 2024 and we are very happy to deliver this hopefully useful and interesting content on all ends of fire science to you, just like in the previous year. Over the course of the year, you will also learn about some of the OFR projects, involvements and opportunities that are related to their company. An exciting year for all to come, and, on behalf of myself and OFR, we are wishing you all the best for the 2024. And now let's have a great start into the year with this interview with David Perser. Hello everybody, I'm here today, once again joined by Professor David Perser. Hello, david, it's great to have you back in the show. Hi there, I hope you had a great time in Japan where you were receiving your big career award for your research, and I would love to follow up. We've done so many episodes already on smoke toxicity and how smoke affects people, how fires produce smoke, how we can measure that with devices and all the chemistry that's behind it. One thing we have not covered yet is the other end of the fire safety, which is the human part in the engineering, and I also know that you have been very involved in the research related to the evacuation processes. So first, to give a little structure to our discussion. We often build fire safety within a framework of available and require safe evacuation times ACET, r-set. So let's try and introduce the listeners to the concept so we're on the same page, and then we will go into details of what goes into the human part of the equation.

Yeah, ok, so I think of ultimately all survival infos is all time based, the time based processes. Now a two parallel process is going on in any fire incident. One of them is the available safest sec eight time, which is the basically the fire growth pattern. So at a certain time a fire ignites, it takes a certain time when it's growing slowly and possibly in the enclosure of origin. Then typically you get a phase when it spreads very quickly and at a certain period of time the conditions become untenable in various spaces within the building, called environment. And so we need to do calculations, as we spoke about last time, which we talk about, how long you've got from the ignition of the fire to when the conditions in the occupied spaces and escape routes become untenable, because that's the point of which people who were trying to evacuate or trying to shelter in the fire will become incapacitated, and once they're incapacitated they're likely to die. So we need to make sure everybody's in a safe place or out of the building before we get to that limit of ASAP. But in parallel with that timeline we've got the what the occupants are doing, what people are doing, and during the 90s and 2000s, early 2000s, I did a lot of work to try and characterize this and quantify it in a way that engineers could put into calculations. So colleagues like Jonathan Syme, who had done a lot of work on he was more of a sociologist and psychologist, he did a lot of work on human behavior and he kind of opened our eyes as engineers to the fact that human behavior was an important and complicated component of a scape thumb. It wasn't just the physical time it took to run through the escape route, you had some, and what people were doing and why they were doing it. And I kind of picked up on that and I said, okay, how can we put that in a way that we can actually quantify it for engineering calculations? And to do that, what I came up with was a concept to design behavioral scenarios. You know that if you look at what happens to people in any one type of building or case, you find they went through a series of stages between the time that the fire ignited. So the common starting point for our set is that it is ignition time it's very important and then the various things they were doing. And I came up with this kind of diagram which you find in the British Standard on PS7974-36. I commend you to have a look at and I've been published in a few papers as well where we broke down the escape process into stages and so it's fairly straightforward, obvious thing. So the first stage is the time from ignition to when the fire is detected. Now we normally think of a simple process where the fire is detected, a small fire, is picked up by some automatic detection system and then immediately a loud alarm spreads throughout the building, everybody runs out. But when we look at real scenarios we find it's not really as simple as that, that often a fire will start and it probably will be picked up quite early, either by a person or by some kind of detection system. But then there's a period of time before all that information is spread to the rest of the occupant population who are affected. And a common situation, as I think is very important, is in sort of large public buildings where the occupants are not sleeping in accommodation, where if often if there's a detection event by an automatic system, that information will be spread to somebody in security who will then have to take some other subsequent action or not. What we often find in, particularly when things go wrong, is that although somebody becomes aware of a fight at very early stage, then the delays creep in. While they go to checkbox, going on, various behaviors are engaged in by those individuals. None of the regular occupants yet know there's a fire, but security might send someone to go and have a look and then they might report to their boss, who then phones his boss, who then has to make a decision. Go now another look, see if it's really a detection. So you get a kind of a multiple detection loop going on.

I would second that based on my experience in engineering buildings and the running experiments in buildings. There are like technical aspects to that because you have some delay times built into your detection. You want to minimize the false alarms in your building so you cannot just immediately evacuate everyone the moment a sensor sounds in your building. You have to confirm that it's a real threat. So usually that takes time. There are human component to that. As you mentioned, someone might need to seek approval for launching evacuation of a large building or stopping the whatever production is happening. Imagine fire in an iron mill. You cannot just shut down the iron mill with a button because there was a fire in it, right? So this time can range from almost immediate, like the moment the fire happened it's detected. It can be awfully long time when multiple levels get into that and at the same time the timeline of our assets. The fire is growing, usually in an exponential manner. So it started as a tiny fire. It could be a few hundred kilowatts by now. It can be a few megawatts by now. It can be approaching even a flash over by now if you have a small compartment and a large amount of plastic material in that compartment. So it's the time zero before people know that there's a fire and the time where we start our analysis are two different points in time right.

Absolutely. Yeah, that's right. And I mean I talk about this sort of golden period when usually after some ignition, that things are progressing quite slowly at first and there's plenty of time to take action if it's taken early. I mean, I think it's human nature we always underestimate threats. Potential threats have become serious at an early stage. You can see it in the response to COVID, you can. You know all these kind of things and because I appreciate this dilemma that people find themselves in, this depends very much on the type of occupancy. But if you think of a classic thing like an airport, you get a report of a small fire in the baggage area or you've got to immediately evacuate that airport, keep all those planes in the sky that are coming and going. It's a big decision to make. So obviously the management on the scene reluctant to make a big disruption unless they really have to. But we have to find some way of factoring these issues into our design and assessment processes. We have to and I think we need to really review these things and look at well, exactly how should we be responding in these situations? And I think it's very important that the fire safety management, the active on-site fire safety management, should practice different scenarios, up to the point where they would evacuate, not necessarily involving an evacuation. But you know, you should say here you are, you get all your crew in, they're all there, and suddenly you have a false fire in a certain area. And then you say to them OK, someone, the detector's gone off, what are you going to do? And then, three or four minutes later, you say OK, now it's a megawatt, what are you going to do now? What have you done up to this point? And how do we balance these conflicting things? Obviously, you don't want to evacuate a building for a very small incident, like a piece of toast going off in the restaurant. You need to have some way of getting there. But on the other hand, what I see when I look at incidents now, of course, the incidents I'm looking at are all ones that have gone badly wrong. So it's a subset of the total set of incidents, many which may have ended up as a trivial result. But when the serious things do happen and you get multiple deaths, you can nearly always see there's been this chapter of early detection and then subsequent delays before warning has been given to the people who have mostly affected the occupants. And you know I gave a few examples in my talk in Tokyo about Düsseldorf Airport and the Channel Tunnel and the Mont Blanc Tunnel to transport related ones. But we've also had quite a few building related ones Bradford Stadium, martins-pencers' Department Store. There are lots of. There's a common theme to these things.

Yeah, fantastic, we'll go there. I'd just like to close one thing, so let's at least briefly mention what's in the 7976.6. What are the parts of the time? We have briefly covered the detection time, but there are other components to the time. So let's just mention what the other components are, and then we can come back to discussing them.

I'm not going to do that anyway, but we've got sidetracked, yeah, ok. So the first thing you've got is detection. So how long will detection be before some kind of alarm is given to somebody? And then you've got what I call alarm time or warning time, which is once a person could be somebody in security has detected the fire. How long will it be before a general evacuation warning is given to the affected occupants? Now this may not be everybody in a building, but affected past the building or the whole building. So that's the warning time. Now everybody's been warned. Now we get into pre-travel time or pre-movement time. So when I did a lot of this work in the 80s and 90s, we've realized that there was no real data for covering this kind of behavior in pre-travel time. So we'll try to do different experiments. These are experiments in different types of occupancies, where we bug the place with cameras and then sound of the alarms unannounced, and we looked at how people behave and so we tried to measure a pre-travel time distributions for occupied spaces for different types of occupants with different types of characteristics. So my plan here was because there's something you can't calculate from any kind of theoretical basis. You can't calculate how long someone's going to take to respond. All you can do is set up a case and observe it, collect the data. So I was very keen that we should try to assemble for the engineering community a database of pre-travel times and occupants is defined in terms of the main characteristics and affected those pre-travel times. Ok, we can talk about that in a minute, so we'll get your pre-travel time distribution. Then, once people have so if you say you have 50 people in an occupied space and you sound a warning, then there's a sort of pattern that follows First of all, nothing happens. Everybody carries on doing what they were doing to start with for a certain period of time, which is variable depending on the situation, and then individuals, one by one, will realise there's something serious going on. They've got to stop what they were doing and they've got to do something else. They've got to react to this emergency situation. So then they go into a new kind of behaviour, but they don't hit the exit. What they go into then is their response time, and in the response behaviours what they do is the sort of typical things like collecting the kids, looking for belongings, wayfinding and looking for information and thinking about what have I got to do, and then, once they've gone through that phase, they think I've got to get out, I'm going to the exit, I'm going now, and then they go into their travel phase. So those are the sorts of phases and behaviours involved in pre-travel time which we've now got quite a good database of and I refer to that in my talk and so you find it in the handbook and various publications. So that's where we need a lot more, but it's coming on. Then people go into the travel phase. Now, once they go into the travel phase, we develop cellular automata models and things. There are a number of these around, but we developed our own in-house model called GridFlow where we look to the relationship between distributions of pre-travel times and travel times for various types of occupancies single occupancies and multiple occupancies, like blocks, multi-story blocks and this kind of thing and basically what we found was that the time it takes to empty a space is bounded by two sets. So if you have a crowded space, then the time to empty. Any crowded space is limited by the time it takes the first few people to start to move, so, like the first percentile who start to head to the exit, so it's the pre-travel time of the first percentile and then they form the queues at the exits and then the rest of the time is a simple flow calculation, depending on the exit width, aggregate exit width and the number of occupants having to get through that exit.

At that point the response of individual people in that venue does not matter that much, because they eventually all would be moving towards the exit and they would enter the queue. And the time is just the amount, how quickly the queue can Get through the exits.

Yeah, that's right. Now the other, the other sort of bounded situation, is where you've got the same space but it's sparsely occupied and then there's no queuing. So the time then depends on the pre-travel time of the people to move, and that's also another little fact there, which is they're walking time to the exit. But that's a small component, usually very small. The actual travel distance is not very important, it's the time before they start to move and then whether or not there's a queue at the exit. So in the sparsely occupied case, what we find is that people, it all depends on the last few people to decide to move. And what we also found was that you almost always have a log, normal distribution of starting times in any occupied space. So what you need to know is when will the sort of 99th percentile group decide to start to move, because they were, with the one, the last ones to go to an exit?

So when they reached the exit, no one is there because there was no queue.

Exactly, they just walk through.

It doesn't matter In the end they could be the only person in the building and they decided, like the last person to decide, to evacuate is the one that's going to give you the maximum value of assets that are required to save a vacation you might want to allow for that.

But there is a bit of a distinction, because if you think about the crowded case, then there's a physical limitation all the time it takes to clear that space, whereas with the sparsely occupied space these occupants always have what I call the means to escape. In other words, if they walk to a door, there's nobody stopping them going through it, they just walk through it. So if conditions started to deteriorate in the sparsely occupied space, people should have a better opportunity to escape. The big problem arises when you get rapid deterioration of conditions in a crowded space think station nightclub when you have a huge number of people trying to get through limited exit capacity in a very short time available for them to do it. The station club was a classic. I calculated for that that it was about 90 seconds from ignition to loss of tenability and in fact it's interesting because the occupants from the videos that we had started to evacuate that space but before the alarm went off and the alarm went off in a timely manner.

So there are pre-evacuation distribution, everything was. It was pretty good, pretty good.

Not too bad, but what was the big problem was there was totally inadequate flow capacity in the exits of the people who had to try and get through it. And what happened at the station nightclub was that they actually kind of created their own exits by smashing the windows and in fact, if they hadn't smashed the windows and just relied on the physical exit provision, there would have been more deaths there than there actually were in these. So the combination of pre-travel times and exit flow capacity are very important in emptying a large occupied space. Then, if you get to the final stage of evacuation, then you've got where you have multiple streams of people converging, for example coming off different floors into a stair, and then you've got to also think about the congestion in the stair and the flow capacity of the stair. You can do fairly simple calculations on this, but that's where the models come in very handily, because it's quite complicated to look at all the relationships between them. So those are the different stages. We've got the detection stage, the warning stage, the pre-travel stage and then the travel stage, and they're all basically additive. And so in order to compare ASET with ASET, and if you're doing a proper analysis for your case, it's very important that you address each of these parameters in the ASET timeline adequately and continue with design calculation or whatever you're doing.

Fantastic. I mean, everyone knows your diagram. By the way, it doesn't have to be introduced. Everyone is using that diagram. And if you don't know which diagram we're talking about, oh, you know very well which diagram we're talking about. I just have to point it out to you. You will see back, then I'll make it a cover of the episode, so you know. Anyway, as you said, a lot of psychological, physiological things that happen with humans, how they move, how they act, that has to go into some sort of engineering assumption, because today I have to engineer my building. This approach is often criticized because it doesn't reflect reality, but I don't think in many models in fire safety, engineering reflect reality. It's not the point. The point is to have a useful model to allow, as engineering, buildings at the decent level of safety, not to be a realistic reflection of what real fires are. When you were creating this framework and we did not just create the framework in the published document there are times attached to those assumptions based on occupancy. How did you figure out what is appropriate time for a hotel, for a simple occupancy multi-dwelling? This sounds quite difficult and there's a big responsibility in those numbers because they are used by people to craft buildings.

Yes, they were put into some extent as examples and there were lots of caveats in there to say, really, as an engineer, if you're doing a specific building, you want to try and obtain at least that table. Particularly is to do with pre-travel time distributions and in that table who you're referring to, I put in estimates of first percentile and sort of 99th percentile pre-travel times and the idea is that you would use those numbers or distributions related to them, together with the travel distributions, to work out the times you'd expect to clear those spaces. And what those were based upon was at the time but hopefully better now a limited number of these experiments where we went to different types of occupancies. We put cameras up and we evacuated them with no warning. That was the whole point, so we could measure people's reaction times. Now, obviously, when you do that, there's quite a time One thing to do in evacuation drill, where everybody knows the bell goes off and you get out, which many buildings do but it's quite a different thing, particularly in those days when cameras and things weren't so good, to measure reaction times of people. It's a lot of work to set those up. So it's based on a smallish number that we had at the time and, of course, the ones that are easy to do, I think, like university classrooms and occupants or offices and things like that, you can really find a tame population to test. But we did manage to do, and then we did one evacuation where we evacuated a hospital outpatient department, for example, so that one source of data is where we actually set up to measure these. We did a lot of work with the University of Ulster where we did some department stores and I also looked at data from real incidents and what I really wanted the gold standard I wanted was where we got CCTV of actual incidents showing the whole process. But at the time they were very, very rare. There weren't many of them. There was more of them around now, so it was a limited database. But one of the things I was particularly concerned about was what are the main qualitative variables that we can identify that have a big quantitative effect on people's behavior, and so I came up with this kind of plan that we would have something called a design behavioral scenario, which is a bit like a design fire scenario. I think there's quite a good parallel, because you think of a design fire scenario, you come up with a sort of set of fires that you expect might happen in a certain occupancy. But you know they're sort of semi theoretical, but there are so many variables involved you can't actually predict with detail how a fire will go, and this is something analogous to this. But you can get some kind of feel for what you might be likely to get. So what we came up with was this plan, and we defined these scenarios not in terms of the building itself but in terms of the nature of the occupants, right, and so the first category was buildings where people are awake. So are they awake or are they asleep? There's a big distinction. So people are awake and they're familiar with the building. So what? That scenario number one, scenario A, type A, type B is people are awake but they're unfamiliar with the building. So type A would be something like an office or a factory or warehouse or something like that. We just got the employees. Now what are the characteristics of that kind of a scenario? Well, often people are scattered amongst multiple enclosures, like in a big office building. So that's a bit of a challenge for how you manage and understand how they will react to that. But one characteristic they generally have is the density of people is quite low in an office, okay, and therefore, going back to the trouble side of things, queuing at the exits is unlikely to be a serious issue in that kind of a scenario set. So what you're really focusing on, then, is warnings and getting people to start to move their pre-travel time aspects. Now, how do you quantify those? Well, you can measure them, but then you have to define the things that are affecting the thing you're measuring it. So I then say, well, what the main things are. Determined how people respond in those little situations are the extent to which they're familiar with the building and its systems, how well trained they are, how well managed they are and what kind of alarm system they have warning system.

So are they okay, they're familiarity.

We're only looking at the pre-travel time now, so we're assuming some kind of warning has been given. We wanted like some sort of estimate on how long it's going to be before they're going to respond to that warning and start to move the building. And so what we find is that if they're well trained, if they're familiar with the place and you have a good culture, you only need a sounder because they know what the alarm system is. The louder it is, the better, and then you can get very good compliance and you can get very short pre-travel times. If, on the other hand, you have that same basic scenario but people of the quality of the management, the fire safety management is poor, you don't have regular drills, you don't insist, you don't have floor wardens, you don't have people who are compliant and used to responding well, then we can find that these times can get extended. Or if you have people there who have been unfamiliar with the building and don't know the systems, we have limited database, very limited database, but we do have some examples of long pre-travel times where people didn't react to the alarms because they didn't really believe they were serious or things like that. But, for example, we had a big database from our own institute, bre. We lost the buildings in there not people working in fire, people working all sorts of building related things and we had regular drills and we had a database of those and I plotted that database and I found that basically everybody was out in the building within a couple of minutes in every case, because we had a good culture. So you could show that you could get good compliance in those situations. But sometimes you could get bad compliance. So that gave you a kind of feel for what you could expect in that kind of occupancy. The next one is Awaken, unfamiliar, and this is basically things like shops, stalls, things like that. And the thing about that is you might you like to have fewer spaces. If you think of a department store or something it's usually a few fairly large enclosures but or an assembly space, perhaps in a hotel, then one of the distinctions there is people aren't familiar with the systems, so when you sell the alarm you don't know what they're going to do. But we'll come back to that. The other thing is they're likely quite often going to be high density occupants. You think of a shop at Christmas coming up soon or something like that. You could get a high density of people, and so there are going to be situations where well, the nightclub we spoke of earlier, for example then you're going to get situations where you've got to get people out quickly and you have to be very careful then about your travel time, exit capacity, because it could have a big influence on the time. But you've also got to consider your pre-travels all right, and what we find in those kinds of occupancies is that if you have people who are a good management culture and some of the examples that we did experimentally were like this, where you had well-trained staff and when I talk about management, I'm not just talking about the management of the building and its systems all the time, I'm talking about the active management during an incident. This is what's so important. So you have an incident, you have your staff trained and you train them to immediately switch people out, and there's quite some nice examples of this. The first one, experimentally, was Gila and Prule's work in the Tyno-Weir Metro that she did for her PhD, where they found that if they made an alarm just as an alarm signal on the platform, you know people were very slow to respond. If they had a voice alarm system on the platform, they got a better response. But they had much, much better response when the voice alarm was backed up by a sweep of the staff in uniform, so they'd get out, get off and we've got to leave. And so by looking at these sort of scenarios, you can see that if you have all these things in place of the management perspective, you can get a rapid and efficient and a short pre-movement time distribution. If you don't have those, you just have some kind of buzzer going off or alarm going off then you're going to find that you're going to get a much longer pre-distribution. So this was the basis that I sort of started to put together this guidance. But there was a big caveat there to say you know, there's some of these scenarios, we don't have very good databases and we need to get better databases. And if you're a designer, think about these issues and come up with. If you're a performance-based designer, what database do you have? Supposing you're designing a store for a chain, you have 300 other stores and you're doing number 301, then you should say to them okay, what kind of response do you get when you have evacuations for your other 300 stores? We can use that as a basis for our calculations, you know. So that's the kind of approach I was taking.

But a lot of those information would come from non-fire drills or unannounced fire drills. So as real as it can be, but no fire component to those In the real fires. You will also have some visible cues of the fire happening in the building. You can see the smoke, you can perhaps see the flames, you can see the fire trucks outside of your building, so you can get more cues than just the alarm. Does this impact the evacuation time? Does it shorten it longer? It?

Yeah, that's right. So what we're talking about is years ago we used to talk about alarms and then the behaviorists, we supposed to say well, it's not alarms, it's different sorts of cues, which an alarm is one of the cues that you might get. I think here we want to distinguish a little bit between maybe some of these sort of public building type situations and on the residential side, which we haven't spoken about yet. But I think if you think of a fire in an office, multi-story office building or maybe in one room in a multi-complex theater complex, cinema complex or something like that, I think it's reasonable to estimate that most of the people in many of the enclosures within that space during the early stages of the fire may not have any other cues other than the alarm if it sounded for them. So it's reasonable in a design context to assume that people only the only cue they have is the alarm system or the management intervention and that most people in many real scenarios that will be their only immediate situation. If you get to a point where in those kind of scenarios the airports, the theaters and things like that where people are seeing something other than the occupied if the fire starts in an occupied enclosure, of course, then you do get these things immediately, then it's been too long before you've already gone past a point where you should have done something. But I mean, I guess station die club is a good example where people were responding to the cue of the rapidly growing flames on the stage and the ceiling jet before the alarm went off. I thought that was quite interesting, telling in that particular case, a big problem we had in that and we're doing. I think there's another issue here, which is that in some of these occupied spaces we're getting these very, very rapid fire growth currents, particularly in these kind of nightclubs. There's been a lot of them all over the world, particularly in Europe, over the last few years. We seem to keep repeating this problem where we're getting rapidly growing fires. But when you think of the longer going fire particularly and Grenfell was a good example of this, yes, then people are influenced by a whole range of cues. I mean in the domestic situation as well. You might smell something and I did talk about this in my lecture that the main reason, for example at Grenfell, that many people in the flat six, which is the flat column of flats that were first affected, how did they become aware of the fire. Well, actually, in most cases it was because the kitchen or hallway smoke or heat detector went off, but quite a proportion of them it was because they smelled smoke or they heard a sound. One person said she saw the reflection of the flames on the windows of the school. On the other side of the courtyard the flames were reflected off another building. So there were a whole range of cues. The people who were in the other flats, who were slower to become aware or be affected, they were picking up on all these kinds of different cues and a lot of these cues wasn't fact and being called up by Rosesh who were outside the tower or another. So this kind of thing.

Grenfell is a very direct example that just knowing that there is a fire is not enough for you to evacuate, because these people had conflicting information from even from authorities telling them to stay put in their buildings, calling fire service and being told that you were supposed to stay in your flat and at this point it's very obvious there is a fire in your building. You are absolutely unaware of the severity of the situation. You just know that the fire is there and your decision of evacuation now is not just the knowledge that the fire exists and that it is in your building, in fact, but you get a much bigger amount of conflicting information. If I could like, before we go into the Grenfell because it's a very interesting case study and many that you mentioned before Mont Blanc and other fires I think it would be great to go through them individually. So if we are in a scenario where people would receive the information about the fire from some sort of automated system that delivers the warning to them and the fire is not there and they evacuate and it happens multiple times how that affects their ability to evacuate when the fire happens. I'm simply talking about the issue of false alarm in the buildings and to what extent the false alarms may underpin the entire safety strategy of a building. It's very interesting to me.

Yeah, now, it's a question people always ask me and I don't have any kind of quantified study of this. All I can say is that I think we need to make a big distinction between dwellings and other kinds of buildings. Okay, so I can't remember. Nearly all the research we've ever done is on what we call kind of public buildings, buildings where there is an alarm system affecting the whole building and an evacuation strategy and 24 hour on site fire safety management. So all sorts of things shops, airports, whatever transport. So in those sort of situations we usually get pretty good compliance and there you would have regular expect to have regular drills and people talk about what's a reasonable frequency of evacuations and I would say once or twice a year is reasonable. Generally people will tolerate that and you get pretty good compliance in most cases. There are examples like I think the sort of case that people bring up is student halls of residence where the students are always setting the alarms off and you might all hostels and things. They're there. It's a bit more challenging to get good compliance. I think the way around that is to insist on good compliance and have a good management there you need a good management. So if you have floor wardens somebody who are volunteer there's not, there's not anybody who's paid to do this Somebody who responsible will take on responsibility in case of an incident, getting everybody moving. That's very effective and you can get around these kind of problems. Obviously, if you have an alarm system that's going off for a while, then you have this kind of cry wolf problem possibly turning up. Then you need to address the system in place.

But generally speaking, you can get good compliance if you approach it the right way it's almost like an enforcement almost, of the evacuation, because if you have wardens that are responsible for evacuation, that means they're obligated to have everyone obliged with the commands. So they kind of enforce the evacuation on people because they know they will be compliant if the evacuation does not happen. They failed their job. So you're almost. And in this case perhaps the role of the false alarm is lesser, because every time an alarm sounds, that person is enforcing others to follow. And in dwelling systems do you? Dwelling is completely different, yeah.

I think let's talk about dwelling. Is this what I brought up to do with Grantsville really? But big problem with dwellings. We don't study dwellings that much because there's no alarm system, there's no requirement for evacuations from dwellings under control, and so nearly all the funding and all the research that's being done when I look back on it was on what I might call loosely public buildings but think about. Certainly in the UK and I think it's probably pretty well the same worldwide is that a power from things like certain student flats and places like that where they do have alarm systems and we can get some data from those, the majority of dwellings. The requirement in the UK is that you have for all buildings, all occupied spaces, that you have to have some system of detection and early warning of form by means of escape capable of being safely and effectively used at all material tallings. That's the performance-based mission statements for any building in the UK. The way that's addressed in something like a blocks of flats is that each flat is required to have a detection alarm system, a smoke alarm in other words, and you have to make sure that the means of escape from that flat are adequate. So you don't normally have the kitchen cooker near the front door, this kind of stuff, and that's really all. But once you get out the flat you then have to be in a sterile communal space which is protected, fully protected, and leading to a protected stair if it's a multi-storey building, which gives you a solution to the final exit. So the concept is, if there's a fire in your flat, you should be warned, you should be able to escape quickly into the communal space. The fire door of your front door shuts behind you and so the communal space remains pretty well smoke-free, maybe with a bit of extraction or ventilation, and you then go through another fire door into a totally protected stair and you can get out. All your neighbours should not normally need to escape. That's the assumption, because you have far-resistant constructs of the flat and all the other spaces. So the concept is that if there's a fire in any one flat, it will be confined to that flat, both the smoke and the fire. Therefore, in nearly all cases and this is pretty well true in most cases you won't need anybody else to get out. If your neighbours do get out, they should be able to go through a fairly sterile common lobby to the stair to get out.

Well multi-layers. So eventually the communal space could be in some extent smoke lodged, but then the protected stair should not be at all right.

But the bottom line is that the assumption is that nearly everybody else, apart from those in the first affected flat, would not normally need to escape. If they do decide they need to escape, they should always have a clear escape route out. That's the concept, that's the kind of platonic ideal of how a block of flats works, because in practice it doesn't always work like that and I've come across a number of cases over the years where somebody has a buffering, a curly one flat and the occupants of escapes. They've left the front door of the flat open, the communal space that can contaminate it, and occupants in other flats they've been in deaths in other flats. It's not a completely unknown phenomenon. Usually it's one's and two's.

So it was not even the fire spread to surround the compartments, it was just the doors being open.

Well, I'm thinking of one particular, but again, there's no systematic study of this. But one particular case I'm thinking of some years ago was drawn to my attention just as an individual case. I remember the story was that there was a fire in this. It was about a four or five-story block and they're on the top floor, you know, somewhere up the tunnel, not tower up the upper floor and a fire occurred in this flat and the occupants still run out and for some reason the front door didn't shut behind them and seal the flat. It was quite a big fire in that flat. The occupants of the flat next door decided to stay there because they thought that's what they were supposed to do remain in place, you know, which is the default situation, and they did so. But then the fire in the flat next door to them spread out the windows and started to come in, break into their flat. The smokes came into their flat and when they tried to escape, by this time the common lobby was very badly smoked long, so they were trapped in the flat and they died there. This was some case that occurred more than 10, 15 years ago. So these sort of scenarios aren't totally unknown, but they're pretty rare because, you know normally things thick fires. A small point of flat origin.

But in our language rare doesn't mean never.

And the example of the Grenfell is a horrible example. But then when we talk about I mean, we went into this situation because you're talking about fusion thing so the key thing about Grenfell was and other similar instance was the occupants then there's no common alarm system, there's no on site fire safety management, there's no intention to have any kind of warning of occupants other than in the flat of origin by their own alarm going off. So there's no real plan B to get people out if things start to go pear shaped. And unless or until there's some form of fire service intervention perhaps or something like that, and I think this is a, I think this is an issue that we really need to address. There should be a member giving lectures on this about the time of the World Trade Center after that and say well, you know, in a multi-story building, occupants deserve to be kept informed of a developing scenario. There should be some way of communicating with them and they should always have the ability to make their own decision about whether or not, with up to date information, they need to leave or wish to leave, and they should be able to if they want to. And this kind of rather failed at Grenfell because, although when people were aware from a very early stage because they most people on most floors realized what the fire was going on, but obviously they didn't know what was happening below them. They didn't know if it was safe to go down the stair. They were in a very poor information situation. I feel very, very sorry for the situation they found themselves in. But it's interesting that you know you've got these kind of two or three cohorts. You've got a group of people who picked up on early queues, decided to go. They went to the early stage, before the lobbies and stair became smoke logs, and they all got on. A large number of people got out without any great difficulty. The problem arose with the ones who stayed after about 1.20 to 1.30 am, who found their lobby smoke log, and then they had this very difficult decision do I go or do I stay? Well, having decided that they had to stay the first time round, they had to keep revisiting that decision over the next hour or two as the conditions gradually changed. And to start with, of course, they were in a clear flat. It's no far outside their flat. The smoke in the lobby is much safer to stay where you are. But then of course, the forest spread all the way round and then eventually came outside their flat, breaking into their flat from the outside, and so they had a different behavioral scenario to contend with. Now the stimulus to escape is much greater and so they're more pressurized, if you like, to make the decision to try to go across that lobby and into the stair. So they have to keep revisiting that decision do I go or do I stay? Depending on the developing cues and information and conditions they were experiencing. That situation was changing continuously. Do we?

I'm trying to put the ground full situation within the pre-evacuation time concept. I'm contemplating in my head so, does the pre-evacuation let's say behavioral scenario approach, does it work in here or not? You say that there were people who would take a decision to evacuate and then eventually return, revisit that decision. Well, technically there are pre-evacuation time. If we could close it, it ended the moment they made the first decision. Everything that happened later was, I wonder, like okay, if I was engineering a building like that, not knowing the Grenfell example, would I be able, given the tools I have, to figure out that the scenario like this exists, or this is a failure that had to be overlooked by any engineer that would be doing? I'm contemplating that. I don't know the answer.

Right. Well, I would say it's a question of we got to break it down into our other set diagram, right. So we got the two stages. We got the warning time to warning and then we've got the pre-travel time. Now the people at Grenfell were getting these. They weren't getting any warnings. All they were getting was mixed cues because there was no provision to give them a warning. There was one subgroup who did get a warning and ironically they were the ones in the flat six column which was the first flats to be affected by the fire, breaking into the kitchen on each floor over a 10-minute period going up the tower. As I said in my Tokyo lecture, I was able to we know exactly when the fire broke in at each floor level and from my study of the witness statements and the evacuation times, cameras in the basement, the ground floor and various other sources, I was able to establish that those people responded very quickly. They did get a warning because their smoke alarms went off and when their smoke alarms went off, they very quickly got up and went to have a look in those kitchens and they saw the flames outside. They realized if the series.

As in individual detection in their flats.

Yes, because they were the only group who did get a warning, a classic warning, from their smoke detection system. All the rest were just accumulating cues but they weren't giving any form of warning. Now there were one or two floors where you could see a huge difference. I think one of them was the 13th floor where somebody was coming home having been out late and he got into the elevator and he went up and he got to the fire floor at a very early stage and saw there was smoke and realized there was a fire on the fourth floor and his family were on the 13th, I think it was. So he went into the stair, got out the lift, went through his thin smoke at that time in the lobby on the fourth floor, got into the stair, went up to the 13th floor, got his family, said we've got to get out, we've got to get out, and then he went and knocked on the doors of all these neighbors and the six flats on that floor and strongly encouraged them to get out. And so they all got a timely warning and then they all left very quickly and survived and that there were a number of other floors where something similar occurred, where one person recognized at an early stage there was a dangerous situation developing and took the trouble to communicate with all the neighbors and get them all moving, and with a good outcome, whereas other floors, where that didn't happen, then you were more likely to get people remaining, and until conditions became untenable. But the big failure to me was that there was no evacuation warning given for an hour or so a couple of hours, as we know, so it wasn't. Until conditions became extreme people were formally advised to evacuate.

I'm just going through the book show me the bodies by Pete Absent. It's absolutely horrible lecture for any fire safety engineer because the description of the process of the fire itself and all the systematic failures that surrounded it is just mind blowing to us. One thing that is also very surprising and disturbing people got conflicting information. It was not simply yes, no decision. Should they evacuate, should they not evacuate? They were receiving increasing amount of cues from the fire and from what's happening in the building. They see the amount of firefighters outside. They have their relatives calling them from other countries in the middle of the night telling them how severe the situation is because they see it on the news. And yet at the same time they're being given information through a phone, through the services, that you should stay at your home and even the moment that state put policy is lifted for the building, which is somewhere 1.30 am or something, it takes time until this knowledge propagates through the people who are giving this information to the occupants. So Grenfell is a very, very, very complex case systematic problems that all came together in one building in one horrible fire. But seeing your lecture from Japan, given the toxicology aspect of how people get incapacitated and incapacitated by smoke and all the stuff that we've discussed previously, that there was a large group of people who even got the information very late but still would be able to do it right to make the risk 50 or so people who and I'll just say they were people who led their flat before conditions in their flats became extreme.

So although some of them had been in their flats, depending on where they were around the tunnel, for some hours, an hour or so, the conditions that they, and there was smoke penetration of the flats from the lobbies throughout. But by opening windows away from the fire and by shutting interior doors and things like that, they were able to minimize their exposure for an hour or so. And I calculated that and from their descriptions, from witness statements so these are survivors, from their witness statements of the condition they were in. They were alert, they were on the phone, they were lucid, they were making sensible decisions, they weren't being overcome by carbon oxide, et cetera and the smoke. They were in a good condition until the flames got outside their flat and started to break in through the windows of the room they were sheltering in. If they left immediately then or before that time, and if they were able to cross the lobby without in a few seconds, without inhaling a lot of smoke, then the conditions in the stair, although they were very hazardous, were such that it was possible and they did. They all descend right from the top floor to the bottom floor and escape. So you could say that that stair performed reasonably well in terms of its intended purpose, which was to be a relatively safe space for quite a long period of time. It was a dangerous situation but they did survive, and the ones the ones. There were some people who did collapse and die in the lobbies or the stair, and from my investigation I'm convinced that they were all people who stayed too long in the in the flat after the flat was penetrated, before they attempted to get across the lobby and into the stair, and some of them collapsed almost immediately in the lobby right outside their room, others almost immediately after entering the stair. All I can say for certain is they collapsed within a few steps of the landing in the stair and therefore I'm saying they must have. It must have happened very quickly after they entered the stair, so they didn't have time. I don't believe they had time to inhale much in the stair before they collapsed.

I find it very difficult to talk about Grenville.

It's a Well if we want to address this from a design perspective or more or less. I mean questions I asked myself going forward is is it reasonable to maintain things as we are, so that we have this basic assumption? All you need to do is have far-resistant construction and means of warning and escape from the individual flat? Should there be some means and some protocol for having some kind of management of an incident and communication system whereby you could communicate with all the individuals, collectively or individually, should conditions deteriorate and encourage them to evacuate? And I think we should. I think we should have some kind of protocol for deciding when would you then operate that? Do you have to have a Colossierge on site all the time? Who would be responsible for that, or should it be something that the fire service have access to? You wouldn't want everybody having access to it, because it would be easy then to have lots of fossil arms in a block of flats which you don't want. Maybe it's a key system that emergency services would have access. There are various protocols we could adopt.

Okay, so thank you, david. Thank you very much. It's always a pleasure to talk with you, even if we're talking about such difficult subjects as case studies and the real fires in which we observe different human behavior. I hope there's a lesson to be learned by the engineers. Of course, we don't have answers to every fire and there will be fires that will come to our surprise, but also learning from those incidents is the best thing we can do now to improve our engineering for the future.

Perhaps you have a closing message or lesson to engineers that comes up of this case studies and this Tragical Fires we've just discussed as a general thought, Well, I think, from a design perspective for engineers, please be aware that there are these different stages in the evacuation response we really do need to address each one of those terms in the R-set equation. If you like to have a good design perspective, we can do performance-based calculations, but be aware, of course, of some of these parameters where we have limited limits of our knowledge. I think we do have enough understanding now to improve our designs, particularly improve our management or protocols for these kind of situations, and think about how we incorporate the behavior response of the management into the overall design process and how we might improve the training and protocols for fire safety management globally during incidents. I think that's going to be very important going on. This is this pre-warning issue that we were talking about, but I think we are. I mean, historically, what we did was do travel time calculations and still a lot of design is based solely on travel and flow. I think we need to be really aware of the importance of these earlier behavioral phenomena and how big an effect they can have on the outcome of different types of scenario, and I think I'll leave it with that comment.

Thank you. Thank you very much, and that's it. If you noted that the interview ended a little abruptly, it's not because I had to stop David, it's because we kept going for another hour and recorded one more podcast episode. I'm not sure if it's going to get published next week. I definitely need some work and perhaps I will use up in different material. But there is much more stuff recorded with David Go deeper into the Grenfell case study, but also touching the other case studies mentioned in the podcast the retirement house, the Mont Blanc tunnel fire, Very interesting discussion, and I will definitely share that up with you. I hope you've enjoyed this episode. It started with a very basic explanation of what the evacuation time is and how we define the timeline in the fires and ended up with a very brutal case study in which this approach was definitely not enough and all the things that made it not being enough. I purposefully made it like this. There is a lot of controversy for a CEDR set approach in the fire science and I also have been criticizing this approach in the past and I still am. For some of the projects it doesn't make sense to do it like this. For most of them it does. I hope you can make out of this episode some basic understanding of what things take place in the evacuation process, what things influence the evacuation process, what are the important things that happen during the evacuation process and, based on that, improve your fireside engineering. That's the purpose of the Fundamental Fire Science series. I hope it was interesting to you to hear all of this from the person who pretty much defined all of this. It was a men's pressure for me to have David Purser once again in the show and I promise to bring David again to you because this is some knowledge we really need to get through. The fire science community. And while we're at it, the community. There is a community webpage of the Fire Science show. If you go to communityfirecineshowcom, you will enter the Fire Science Show community where you can meet other people alike you fans of the podcast, enthusiasts of fire science, fire safety engineers, my craft people who simply make their living out of making buildings and stuff fire safe and love to talk about it. And if we can make a great community out there, it will be even better for 2024 to share our knowledge, experiences and thoughts with each other, and I highly encourage you to join the committee Once again. You can find it at communityfirecineshowcom. See you there and see you here again on next Wednesday. Thank you, this was the Fire Science Show. Thank you for listening and see you soon.