030 - Visibility Prediction Framework with Lukas Arnold

Hello and welcome to the Fire Science Show session 30 wow this goes so quickly. Today. we're celebrating a new achievement by my team and a team of my dear colleague, professor Lucas Arnold from Germany And it's hard to call it the achievement yet, but it's definitely a good step towards a better future. And that is we've got a grant together. We've got the quite a prestigious grant from the Polish national sense of science NCN and the German DFG agency, , to research visibility in smoke. And that makes me so excited because that's a subject that was with me through my whole scientific career. More or less, I've spent so much time investigating how it's being applied in modern fire science engineering and, unfortunately we found more problems than good things about the model that is used today. And yeah, we were thinking about this writing papers about it, discussing it. And from all of these discussions, we came to conclusion. This needs to be done from the start. This needs to be completely redone from the fundamental experiments that led to the emergence of this model. Exactly one year ago, we've spent so much time with Lucas writing the grant proposal. It was one of the most intensive, parts of my scientific career, but, uh, we really got something nice and the reviewer said it is a very nice. The funding agencies said, okay, we'll fund it. And here we are on the brink of new epoch for visibility and smoke modeling. I hope it will be as good as, as we make it look. So in today's episode, we want to celebrate this. We want to share this with you. We want to tell you about. How this grant was born, what were the fundamental issues with visibility and smoke model that we felt that it is necessary to do it from scratch? What are we planning in the experiment and, How this may be useful even beyond what we've applied for and all of these things look very, very interesting. Now, a few things about my guests, Lucas, Arnold from, Juelich Forszungszentrum and, from the University of Wuppertal is actually a physicist that came into the field of fires. So it's much better at understanding the fundamental of physical models that then I am. And that makes me actually very happy in the relation to this grant. Lucas is also known for his , skills and, his activities in the field of computational fire modeling. If you have ever heard about the Juelich Summer School on FDS, he's the creator of that. And I would highly recommend to have your eyes open because I think there may be a new edition over the corner. And, Lucas is also so skilled with, programming and, developing new models and all of them are open source, so absolutely need to check his repository. And this is the skillset. Compliments, , our capabilities at ITB where we focus mostly on experimental fire science. So if I got your attention and you're eager to, learn about how are we going to try it and change the visibility and smoke model, lets spin the intro and lets go welcome to Fire Science Show. I am today with my good friend, Lucas Arnold. Hey Lucas

Hi Wojciech.

And we are here because we got it. Yes, Lucas, we got it.

Yes.

We got, quite a nice grant, from, uh, Polish and German funding agencies. NCN and DFG to. Help us or allow us to pursue a completely new approach for modeling visibility in smoke. And, that makes me very excited. How about you?

Yeah, me too. So this is a really a good achievement for us. it's, we are also a very young group. So this is really, really good. So this will now allow us to do things. Couldn't do before, because well, eventually in the end you have to pay for all the stuff you want to do. even if it's just theory, you have to have people to do it to implement the ideas. And So I'm really much looking forward. And I think the collaboration like having two partners focusing on numerics and experiment is in natural. Good. Combination.

yes, it's going to be very exciting. You're going to learn, or you're going to hear a lot about this project today and, quick announcement where we will be looking for post-docs. So I am going to have a postdoc position for my experimental part. And Lucas is going to have a one or two. How many of you have one

One post-doc the application is on my desk over there. So it's going to be out in the next few days or depending on the internals.

So, if you like what you hear, you can work with us. That's a hint. Okay. So let's move to the topic. The name of our grant is Visibility Prediction Framework. next generation model for visibility in smoke in built environment. And that was the fancy title we, came up with. and, uh, yeah, the idea is there, we need a new generation of visibility in smoke models because from what we think the current models do not really work well. What's your opinion Lucas. Are you happy with existing models?

It probably needs to start, even further out.

Yeah.

I'm a physicist, so I'm more interested in really what's physically going on, but when you are designing buildings and assess life safety, then you often see that visibility is one of the issues. One of tenability criteria does it. And what makes different to other things like gas temperature or radiation is that visibility is something that is way more complex than the other ones. Okay. Yeah. Not talking about the physical details, but in, visibility you have to take care about first of all, how your aerosols , soot, is transported, how has it get generated? What forms does it have? And so on. But on the other hand, what you're interested in is not only the soot concentration or whatever you locally have in your room, but how does this impact the humans? Will I, during a fire, be able to see that sign? And this is where physics is lost or at least this is where humans come into play. Okay. With the, so how does it perceive the side? Is it now? Is it visible? Can they really get the information that the sign wants them to have or not? Therefore you see, this is a very complex topic. We've got different levels with humans and physics and modeling

and light and light interaction, which is also not on

Exactly. Yeah. So there are a lot of things and I think right now, , there's a lot of room for improvement on many scales. We see that we cannot reproduce. light, absorption in room-scale, simulations or together with the experiments that we did. And, when you go back, this, Wojciech your part. I don't want to talk too much about that, but like how many, how do people perceive these things? Everything, probably most of the audience knows Jin's experiments, which are the base for everything right now, uh, regarding visibility, how humans perceive signs and so on, and this is now old.

From my perspective. the reason why in general, my, my scientific career , , went into the route of visibility and smoke because it's , not the first time I'm touching the subject was the fact that every, like every single building we have designs, literally every single one of them. Visibility in smoke would be the first one of tenability criterias to fail. When you're doing science and your visibility goes down to zero, you can still investigate temperatures and it can be interesting and so on. But when you're designing a building and if visibility falls below the tenability criteria, everything else does not matter anymore. Because you're tenability criteria has been breached. Your, , your available safe evacuation time is now defined by the moment when this happened. And this is your engineering. In the real building, your fire safety engineering is often in fact, visibility in smoke engineering. Like you are doing things to, improve this one particular condition that fails the first. So it has this profound meaning for the build environment. This is the fire criterion that defines buildings and. If it is so important to us, if we truly design all of our buildings, , based on this single parameter, I guess we should be doing that correctly because otherwise it would not be a fun story. And the deeper you go into the literature of, how this is being modeled, the. Assumptions approximations. You see the more fundamental magic numbers that come to the equation to find that a yield, this one number that as you said, represents. A hell of a complex physics, a physics of light smoke interaction, physics of smoke movements, smoke production, physics of the human eye. Even because if you go to the roots of it, there was a critical contrast ratio at some point assumed as a value, , which gave birth to all the correlations later on, which we don't even remember anymore. So. for me, I always seen you know, visibility in smoke as on one hand, extremely complex thing to play. And very interesting, scientifically fascinating. And on the other hand, the thing that drives our buildings drives my design. I am working on optimizing this particular parameter, and this is why. When we were discussing this, it was not even that we need to improve the current model. Now we need a completely new generation of the models, completely new generation of the models. Yeah, first, before we talk about what we're gonna try and improving the model, or how are we going to try to rebuild it? Maybe it would be fun for the audience to summarize how exactly visibility is being modelled today. So Lucas may be, you can tell us from physicists and fire engineer perspective. If I do my simulation, I release my smoke in the room. It is there. How do I know? What's the visibility in smoke

So, first of all, when you start saying you release your smoke and move it there, you already have made a lot of decisions, with that. So obviously there are transport processes and so on. The normal ways, how things are transported in fluids, but you have also then defined, how much soot you will induce. So what typically is a soot yield is so how much of your fuel is, kind of transferred into soot but you also have said. What kind of optical properties your particles have. So by this, I mean, what is the particle distribution? What does this shape distribution and so on?

sorry, but, these are hidden within coefficients, right? Like the particulars would be in the specific smoke extinction coefficient, which we date back to George Mulholland's work and we have a value for that. . Lukas Arnold: exactly. So there's one thing about. That you can specify, which is, kind of estimate how the interaction with light will be for, a given wavelength and this value kind of isn't effective parameter to cover everything, all the interactions. So disregard, if the particles change in shape and, If there are other effects that you want to consider different wavelengths and so on and scattering and so on. So this is one parameter to to take it all, which is from an engineering point of view is it's okay to have a single parameter. Okay. You have specified this already. So therefore you have specificity optical properties typically in the light absorption. Okay. So it means like when, light travels through a volume of your fluid, which is eventually containing particles or soot particles, then this will tell you how much does the light get dimmer? Okay. So how much light is absorbed by these particles? So what happens typically when I would look into that camera and if there's smoke in between, the smoke particles would absorb the light, that's coming to the camera and make me, for example, darker, or the whole picture would, would get darker. So you see there was a path there's an observer, like the camera in this case, and me as an object. Therefore the way that the light travels is important, but right now the common assumption right now is a little bit different. So when we go back you have a compartment, you release your smoke and so on. Then, then common way of doing it. Is that you locally defined a visibility. What does it mean? You basically make the following assumption that. at the given position there is a, light absorption coefficient locally, which depends on the density and the parameters you have specified before. There's a single value for that. And then you can use, formulas. So just how light propagates through an absorbing medium, you can use this formula to compute how. light would give him this coefficient would be absorbed, but you now assume that this coefficient is constant for this position for the whole group.. And then you can use empiric correlation, like from Jin to deduce, like, what is your visibility? How far would you be able to perceive the sign. Based on the local value. I think that's the key that the value is local. And I think that's the biggest trap. because, local, it means that, it's a clever way to go from smoke concentration into a value that you can more or less understand, which is the visibility. So if you are standing in an, in a point of your room, you see for a certain distance. And it makes no sense that in front of you, the visibility would be 10 meters and one meter after it would be 20 meters. It's not that standing here, you see four, 10 meters and standing there, you will see four 20. No, it is just a, such a horrendous simplification of the, problem. And actually, when this was developed and we're talking about early seventies, the groundbreaking work of Jin in Japan, when this was developed, there was no CFD there was no two-zone models, even, , there was a single zone model for a room. So such an assumption, you know, that you have one concentration of the smoke in the room. And you plot one value of visibility for the whole room. It actually makes sense. If you assume that the conditioning inside are averaged, when you go into two-zone modeling, it also makes sense that you have a smoke layer that spans over the whole compartment. And the average visibility in that layer is this value. It makes sense, but when you apply the same principle into a volumes of smoke concentration. Like you have different value every 10 centimeters in every cell of your model, you have a different value than applying the same principle for every single, cell in your computer model. And then just plotting a colorful picture with the values in every, in any of them. It truly makes no sense in the. Then, if you apply the condition and we often say that the visibility shall be more than 10 meters, for example, it's that, that one is, is in one of the Polish laws written, If I'm standing here and in front of me, there's a visibility of five meters on a distance of one meter. And then the air is clean. I can see beyond this, a small cloud of smoke. So I see above 10 meters, but, according to my law I have failed my criteria and because the local value has felt below that. So there's an issue, you know, with, making a global, conclusions based on local value.

Yeah, that's true. as I said, that there's missing the, , kind of the path that you're interested in, like from an observer to an object or in our case, from the person inside the room to the object they want to perceive. Okay. So the question is, can I perceive the signs, the exit doors or whatever is open. Or not. And therefore you have to take into account the path, which will in most cases, be traveling through inhomogeneous media and, we'll then, kind of create a image, either dark or blurry of the sign. And this is what currently, well it's not used, in the modelling.

Yeah, for me that, uh, local thing was, the main issue from engineering point of view. But obviously if you look on the subject to the eyes of a scientist, I once have, summarize this in a single sentence that we use, because we basically use Lambers Beer, , correlations. And that correlation is, is valid for homogeneous smoke, that is mainly absorbing and it is valid for a single ray of flights. When we applied this principle for nonhomogeneous. That is a mixture of absorption and scattering, which we didn't know in a very complex, uh, light setting of your scene, where you're measuring that. So you actually break the low on all three levels, whereas you should have not, done that. And that, that creates a whole array of issues. Then let's jump into our work, of the last years where we have, separately came to the same conclusions with different methods. That there's another aspect of this, which is the wavelength sensitivity. And I've been building a multi wavelength densitometers, like the devices that are used to measure the transmittance of the smoke in laboratory. Usually it's being measured in the red light. And that's actually how the value of the smoke efficient 8.7 square meter per gram was devised , for red light. however, the different wavelengths will be absolved in a different manner. So we were building multi waving density meters, and you have invented this, , tomographic led reconstruction method, which is the fanciest term I've ever heard for taking pictures , of LEDs to smoke. But it actually worked. So, so we have founded the Greenlight would be transported through smoke in a completely different way than the red light. And the blue light would be absolved as soon as did you gain the same conclusions?

Yeah. we are currently writing, there will be a new publications in next two weeks, at least submitted, uh, on hopefully also accepted so far we did this proof of concept that we published and now we ever did all the experiments that we, what we did. And, yeah. I agree with you on that. But I would like to find out that our goal is a little bit different probably to yours. So you're focusing on the colors, which is very important. We would like. Point of view or from a CFD modeller would like to have spatially resolved data that we compare with spatially resolved model. Okay. Because otherwise you're still then comparing individual values. Okay. And so the idea is, what I think is working great is to really measure the optical properties spatially resolved without having thousand probes

And you also figure out the distortion and, and stuff like that, which is very helpful.

this is what we work on this exactly. To not only look into absorption, but also on scattering. Okay. So typically forward scattering, this is from a physical point of view. We cannot do any any other kind of, scheduling with a single camera yeah, we look into that and also. Like how, you would expect, from Mie theory to predict, given eight particle size distribution, what the scattering and absorption would be, and for this, we need also, the, ratios of the different ways. So that's why we measured now the spectrum of the LEDs, which are in contrast to yours, which are peaks as in a laser led is a broad spectrum, unfortunately. But nevertheless, you can do some conclusions about the ratios and from based on these ratios, you can deduce what the particular distribution may be.

And obviously that does not exhaust all the research that has been done in our units I had a, I have a great history of collaboration with Gabriele Vigne , who was a guest on the podcast here in episode three. And Gabriele has said lots of nice things about visibility and smoke as well and how we were searching for optimum. Uh, . And recently we've performed a survey of how visibility is being used over the world, which hopefully we will finish processing soon. Ish. I hope.

Um, I'm taking a note. There were some from our last call.

Yes. and, and you, you were working a lot with Duisburg-Essen, , who have, a history of measuring the sizes, the, the shapes of particles. And it's also quite fascinating.

Yeah, exactly. And they're going to also help us out here in this project. They really do the, fine modeling of light particle interaction. So with Mie theory and, as in, not spherical particles and all the, all this stuff.

That's a brief, identifications of the issues with the visibility and smoke modeling while that that's a lot of issues, actually, it's really, it makes sense to re repeat this work, and I'm really happy that we got in chance to do that. Now let's, let's jump into, Visibility Prediction Framework VPF, as we called it what we will try to do. So, , we have subdivided this work into like two grant packages. I can call it like that, I guess. Uh, so what my scientific unit, the ITB we'll be doing will be focused on the experimental research on the visibility. So we're the experimental, partner and Lucas, your team will be doing the software development and computational modeling, which is as exciting, I would say, or maybe even more. And with our powers combined, there will, come and the merger, some really, uh, fun tools to use. First, let me, tell u a bit about the experiments and, uh, I think it's. Very interesting. What we've planned because, you know, in fire science in general, we have this magic numbers that come from ancient knowledge. We have this, variables that no one can trace where the origin from, and it's common over the fire science and in visibility in smoke everything goes back to, to researcher, Tahashida Jin from Japan who was researching this in the early 1970s, who has written some fundamental landmark papers about the perception through shapes to smoke layers. And what, , what did he do? Was he built a small chamber In which he has burned paper and then really smoke. And that's one funny thing that it was mainly white smoke where we clearly apply it to a little dark ish, the sooty smoke. But yeah, that's the first issue with this experiment. He was, having this chamber on the end of the chamber, he was projecting some shapes and, On the other side and on the opposite end of the chamber, there were a window and some mirrors, and he simply asked people to identify whether they can see the shape or not through a layer of smoke that was inside the compartment. And by measuring this light absorption, he could correlate whether a person can see something or not with the amount of smoke that is within the chamber and from this experiment, some really fundamental assumptions for our modeling were born because for some certain ratios of contrast between the lighting of the chamber and the, and how bright the sign was, he was able to show that in some cases, the sign is less visible and we currently refer to them as the, light reflecting signs now. And for some, there was a more clearly visible, which we now refer to as light emitting signs. He found out this correlations for the constant K, aligns this to. And from this simple experiment, we got all the constants for the equations that are being still used up till today. Now a lot has changed since 1970s. We have so many new tools that allowto precisely measure the smoke properties within the chamber. We have so many new ways to generate smoke. Even we have so many new ways to observe with digital cameras and digital imaging. And definitely we have a much easier time projecting signs than using a light projectors and a, a film. We can just use screens into that. We don't have to use mirrors. We can change the shapes and sizes on our screens. So the technology advancements in the last 50 years were immense and we will try to use them in repeating Jin's experiment, but with modern control. And that's like really exciting for me. And, how will you use that in your computational part?

Yeah. So as always, it is important to have a good experimental data, like good. I mean, reproducible things have measured, everything that is, , maybe maybe relevant. So, uh, we will use this data to come up with a. To basically, allow us in the first step to do something like given a smoke distribution, whatever that comes from commonly used CFD, codes, given a smoke distribution, whether the particles are treated as a fluid as are typically done now order via, , a particle representative. That we can create an image of a sign for a given observer at a given position in the room. Okay. So me being here, there is any room smoke distribution in this room, and now there is no exit sign here because my office is way too small for that. Um, but it imagines, it would be one I would like to compute how. The image of, uh, of the sign would look like, okay, from a camera, which maybe easier, but also from, , a human perceive. It, that's definitely a open question, uh, how differences in humans are and so on. But I think a good starting point would be to capture what is, what a camera with. And to do so you have to get the interaction with light, the soot particles with the light, the reflections and so on, so that you, based on these images, you can then do the application of that case. You can deduce from that. Is it still visible or not do like in a given experiment, people. Perceive or recognize the sign or not so that we can use it later on tomorrow, what people would think, whether they would see the sign perceive it or not. And this would then in the end, tell you whether the visibility is okay or not because like in the end, like the length is one thing, but in the end we want to know, are the people able to see the exit or the sign or. Disregarding how far away it is. Okay. And if you automate this for given positions and, come up with a criteria when a sign is perceivable or not, then this will allow you to draw maps or do an analysis for a given fire scenario, time dependent from which, which positions would be critical. Because for example, you don't see any other exit sign or any exit, then you're basically lost.

And the important thing that will come out of my experiments because I obviously don't want to just repeat Jin's experiment. We want to create immense amount of new knowledge within them. First we will invite participants in the study. So we will have real humans, hopefully a diverse group of participants in the experiment, helping us identify whether they can see through layers of smoke or not. But it will not only be a question of, do you see a light at the end of the tunnel, which we will invite them to look at, but we will, , show them pictures of. Evacuation signs with, letters maybe not exit, but we will use words or combinations of numbers. And we will ask them to read them out loud in a way like in an opticians test. But what we will do, we will show them small science, large signs red signs, and green signs. We will show them, , signs that are. Uh, very dim compared to the backgrounds and one that like white and black background with the maximum conscious ratio, we will adjust all these parameters on the fly to try and capture as many data points as humanly possible within the hours that we'll have with our participants. And this would help us really capture this, critical combination of smoke obscuration or smoke parameters and the interesting parameters of the evacuation sign and the environment around it. When the visibility is lost, when you lose clear, grasp of, on what the information is carried by the sign that we show them. And if we can find out this critical. Then we received a completely new threshold condition of assigned distortion, or, dimming of the sign at which the perception is lost. And if you then, , calculate this value numerically for any place within your compartments, based on the light smoke interaction from a certain observer plays in a certain location and he moved up server and you map, what would they see in every single point of your compartment? and we know the critical value from the experiment we are there. We have, we have a true map of visibility. Which means what it says. It is a map of visibility, not a map of fancy recalculated smoke mass densities in, uh, in the room. It will truly be, the map of visibility, but it sounds like quite a computationally expensive. I hope you have some good tricks for that.

I'm not sure how expensive it's going to be. First of all we are, , most of us, are having more resources that we can use. And then , we think, okay. Most probably even this tablet has already a multi-core, uh, uh, thing in there and all in notebooks and workstations. And basically you're also running CFD simulations, which are very, complex and need resources. I think this will not be so expensive, especially, well, it will be costly, but not the post-processing would take longer than the

Mm.

simulations. Okay. And especially when so what we plan to do in a first approach, maybe you come up with something clever and the end is ray Okay. So basically to create an image of something you can trade the individual photons, if you want to on their path to the camera or to the observer, you, there are a lot of tricks how to kind of reduce the amount of work you need to do because you, for example, are interested only in the sign okay. I don't care like how the wall looks like. I just need the design we will use the benefits that comes from gaming. Okay. This is what your game engine does for you when you play a fancy retracing game. So we can use these techniques which, are existing and we will see how we can adopt them. , So that's something that, we will definitely investigate, right now. These effects, like also absorption, so are also integrated in like tracing capabilities. But, , it is, this is what our experience showed us is hard to come up with your own distribution of optical properties. Okay. So these things run like with a given fixed property or so, so this is what we see in our VR implementations. And so on when we have inhomogeneous media. but yeah, so we can build, on top of many things and eventually, if this runs on GPU's, then this would make things way more faster, and this is the goal, but we will see we need to get the physics and so on correctly, and then optimizing goats. That's the second part. But first I want to reproduce exactly what you will do in the. We want to have the same pictures. This is the goal. And then making it fast and run it on your computer. Yeah, that's the second step,

For me, that second step is quite important because. I mean, if we are investing so much time in changing the world, it would be quite nice if the world would actually appreciate the change and go for that, because we're not the first ones to try and great things with visibility in smoke. And there were many smart people before us who were trying to do that. And many, courageous solutions were developed. They just never clicked. And I think.

Yeah.

One thing that was very important for us writing the proposals that we need to couple this to existing, common models. Like we need. To have this couple seamlessly with FDS we need to have this coupled with ANSYS. Not that it has to be a part of it is not, that's not the point we, but we need to have a very easy way where literally a person presses one button and then. the simulation is processed and you just obtain the new, new results may be maybe it will one day be introduced into smokeview or something that would make my day. But, we really want this to be user-friendly and You made a very clear statement that if you work on it it has to be open source, which I appreciate a lot because I think that's a way where community would be able to participate in it. And, , actually, maybe at some point, extend our work beyond, uh, what we have originally seen.

Yeah. you know, I'm a big fan of, processing data with python and we do a lot of teaching about that and, we have our summer school. We also do that. So we've got nice tools to post-process these data. What are these things will make it into smoke view or not. They will be in the end or in not only at the end, all the, during the old process will be available for everybody. And we will make it easy to use if you've got the very basic knowledge of python which is not a click but like a copy and paste a command. And here we go. So we we've, , post-processing FDS data. We've got a very nice tool. Uh, one of our team members, young has written for reading in all different kinds of FDS stuff. It just one line, and then you can access everything. So, and then if you S interest in special analysis, I think it's may not be one click. But it would be one or two lines on a common client or whatever you're using to come up with such maps and so on. So that's something that I, uh, I see is, is a realistic. And then in the end, it's also like regarding the computing time, it does, it depends like how often do you want this analysis to be done? Like every 10 times a second, which probably doesn't make sense. Like every five minutes. Yeah. Or one minute or so this reduces also the competing costs dramatically. I'm very optimistic that it's going to be usable.

We, we really want to do a breakthrough experiments that have never been attempted with level of fidelity that has never been achieved in this type of modeling. We want to determine the smoke parameters on the scale that has never been achieved in fire science, never. push our numerical modeling capabilities to the edge, trying , to really replicate images, taken through smoke with, , our model, not replicate model them and, reached the point where the model can really predict, the images and one way or another, how close we will get to this ambitious goal. I don't know. I hope as far as, as humanly possible and we'll do our best, but now , let's think about the impact this New model could have for the whole field and, beyond buildings, because obviously the tool will be able to, to map visibility in buildings and that's number one goal, because that is something, the fire safety engineering needs

Yeah, if I may add one thing, , to the buildings, so what we, plan here, what we are also seeing right now that we cannot reproduce the experiments, simple compartment, risque compartment fires with heptane. We cannot reproduce them regarding visibility. With the current models. So I hope, and we, in our case, it's FDS overestimates, I mean, it's not default of FDS it's about which parameters you put in, but if you use the commonly used parameters in there, you way over estimate the, the smoke effect. So you have way lower visibility. Okay. So. Experiments which have controlled boundary conditions where we measure particle size distribution in the compartment, not only at the source and so on. I hope that we will also be able to contribute to better predictions , in the building, which I would assume will lead to a more realistic or lower, uh, light absorption or higher visible. Okay, so that I think this would be also a huge contribution to make these predictions better. or closer to reality,

I think it was, Ulf Wickstrom. who once said that we, it was not related to visibility, but to structural fire engineering, but it was something like we should stop calling wrong assumptions, the safety factors. And, uh, and there's the same, there's the same thing in here. We should stop, claiming that the model is completely wrong, but at least it gives us safe, uh, results. Now it doesn't, it gives you a false image of the building and you cannot engineer with a false image. When you have correct image, you can work an engineer and and that's the

And then you can add your safety factors.

And then you can add the safety factors, but you control them. That's the test, the thing that's, that's the beautiful thing. And maybe let's go further on into buildings, you know, If you have, a model today, if you model a building today, please tell me what size of the sign have you put in your simulation? None of the, because you're not modeling that. What was the brightness of your sign while you were limited to non illuminated or illuminated Was it on the white wall or on black wall? Because that kind of matters, right? Was it at the height of seven meters or at the height of two meters or maybe on the floor? these are the things. That if someone asks you. Okay, can you please, can you CFD tell me if I move the sign from five meters to two meters, will my visibility improve? Well, that's an question you cannot answer today. Not with the visibility in smoke model. You can do it by, uh, engineering, the smoke layer and so on, but not with the visibility and smoke model itself. And with VPF, that will be your outputs. Like the map of visibility will be intinisically connected with the size of the sign, the color of the sign, the location of the science versus observer, the ambient lighting conditions in the compartment. And. This will help us like engineer evacuation paths, like when you go. And my favorite example is Barcelona airport go to Barcelona airport or Google it and see this insanely huge human sized evacuation signs that are there. I mean, that is a beautiful solution for fire safe building to have the signs that you can see from a hundred meters and tell what it means that sign. But today's with the tools we have. I mean, you can engineer it, but not with this tools, these tools don't allow you that.

Yeah. So I think there would be many contribution for the buildings. Uh, yeah.

okay, I'll start the second, the second one, that that's really fascinating to me. And that was also a reason why I like doing this now. We are at the beginning of the age of virtual and augmented reality. And there will be a growing need for representing smoke in these environments. And today we're not representing it, correct. It is not the smoke. I mean, it looks like a smoke, but the physics of it are completely off and I think you mentioned gaming and I think in gaming, it's also similar issue, right?

It needs to look

And this looked good. Yeah. But it's the visibility through that cloud , is not a representation of the reality. And I think if we make this model work and we accomplish all of this experiments and numerical modeling and we find new efficient and clever ways to model this light and smoke interaction. Yeah. Our colleagues who are using virtual reality for their experiments on human behavior. For example, maybe they could benefit from this better or truer. If there is such a word exists, a representation of, of smoke.

Yeah. So, we have now also a, , model a, or sorry, a software package or tool that allows you to load FDS data into the Unreal engine. Great. Okay. So this is like now the real numbers. But as you, as you can imagine, it's just it's the soot density basically. Okay. It doesn't tell you like, uh, how will you perceive something? Obviously you can put it in numbers. And so, okay. So the first step to get this number into the engine, which was really. , yeah, not that easy is there so that you have the physical data. So in combination, what we get with properly modeling the light interaction, it would be great for people who then do, like how do people orient themselves and so on to use the physically based, maybe realistic, , properties that people can either perceive the sign or they cannot, or. Kind of properly do it. So, and not guessing how it is because this will influence their decisions. Okay. So, this all is dependent on a good representation of the reality. And so I hope that this will be a spinoff of that with VR providing a platform for this. Human in a, in fire, , studies, with a physically based, approach. So as far as I know, there is a lot of people working on that, which, as far as I know, sorry, if I overlooked something, did they using kind of this gaming or prescripted, smoke and not let the CFD based.

I've seen an approach when I visited a University of Greenwich and I saw the research when they were also trying to process that. I mean, they're, they're great at augmented and virtual reality. And, uh, I was so impressed by the research, but it was. Also, they've mentioned this challenge, that to get a realistic image of smoke and you have to resort some, some tricks because a good model does not really exist. And last but not least, we are fascinated by visibility. We are fire engineers and, it's something very obvious in, in this field of science that this needs research. And we obviously went from this direction, but to just measure the optical properties of smoke That is produced in a flaming fires. That is, the measurements that are transient. You know, how the properties change with the age of the smoke to try and capture some soot deposition on the walls, maybe to really see the effects of particle distribution on the visibility or on the optical properties of the smoke. And I mean the visibility properties or optical properties narrowed down to the visible spectrum of the light, which is the most interesting for us and the smoke narrowed down to what is produced in fire. I'm not really interested in tobacco smoke and I'm not interested in sand storms. I'm interested in, in smoke, that is an effect of flaming or smouldering combustion. And I'm not interested , in microwaves or ultra violet. I'm very interested in, in, in a visible light spectrum. all of these things have been researched by other researchers in combustion field in the fields of, of atmospheric science. And, , these people have done a great job, but they've never really focused this narrow, , part of the light spectrum. And then the narrow part of the, of the particle distributions, which we want to study , in the best way we can, like, we literally have planned every. Possible a measurement technique we know to, to try and get this visual ones, the led tomography that you've invented multi wavelength spectroscopy that we've got we've gravimetric sampling, these, From your colleagues from Juelich uh, the, um, ways to measure the particle distribution. So there's so many tools to be used in this one experiment to really precisely capture the, , particle distribution and the optical properties that on alone, I hope this will be already a great contribution , to the science, not only fire science

And if I may add something, we are speaking mostly about suit particles, but what about.

Yeah,

And stuff like that. Okay. Which,

complicate the Grant's Lucas noooo

no, no, no, no, no. Now we're looking here into the extension of no, but I mean, there are people who say, well, one of the issues is visibility and mist okay. If you, well predicting these things is also helpful for new new measures. a watermist makes situation worse or would it make it better? Well, the visibility will not get better, but, , you know, how bad is the limiting of the visibility due to , mist and so on?

So many pathways open and I really hope, we will get there first. Let's start with the recruiting process. So if you like what you've heard and the, you fancy a PhD or you fancy a post-doc, but I think I will have a PhD student position as well in the grant. So

we will decide on a person

Yeah, we will, we will decide that on a person basis. So if you would like to work with this team and change the world with us, you are more than welcome, to do that, to look up for the calls for them recruitments and, yeah.Any final words Lukas?.

No, it's gonna be a fun project beside the only outcome is for the economic grade to your youth. And it's going to be a great Polish, German, , collaboration. So maybe we should have done this podcast in Polish

Or German

after two minutes, I would be already gone probably, but maybe I can also improve my Polish by

That's so so great. So yeah. And thank you so much to, NCN and the DFG for believing in our research for evaluating this. so highly, and we really appreciate, all the reviewer's comments we got. I mean, for the first time, I really had a good experience with grant reviews and stuff like that. It was so nice. And, so on the point I really enjoyed this process and. I'm really grateful for, for the funding agencies to believe that this is an important topic that may actually change the part in the field of science. And, I hope we will achieve that. And no matter, no matter how it ends, it's going to be fun. And you will hear a lot more about that.

Oh,

Okay, thank you, Lucas, for, for coming to the show and yeah, let's go back to the work. There's so much to be done. Cheers,

that's true. Okay, bye. See you soon.

And that's it. I hope you've liked what you've heard. And, yeah, we're a bit crazy with developing all of this and planning such as an intense research over the three years. But on the other hand, it looks manageable, achievable. And I've said that many times, if we only do half of the things that we want to do, it's still going to be a breakthrough. And if we do all of them is going to be a paradigm shift into discipline. I have never been more excited about the carrying a research item in my life than now. So I'm very happy to do that. One thing that we will need, and we will need that. It's the clever people who will do it with us. There are PhD positions in my part and the experimental part, there will be a position in Lucas's part, which is aimed as a post-doc, but I am pretty sure you can negotiate to turn it into a PhD research as well. And I will have a full, a post-doc position for three years as well in the grant. So there's so much work to be done with us. And trust me on that. it's going to be, uh, one of the top grants in the field of fire science in the year. So you want to be a part of this team. Keep your eyes open on, the announcements. I'll be posting that on my Twitter, on my LinkedIn. I'll, I'll keep you updated when the call is. And, I guess that's it for the episode. I'm not going to add much to what has been said already. in the show notes, I'll pop you some links to interesting resources on visibility in smoke, both from my team and from Lucas's team. , including some YouTube videos where on the history and, principles of the visibility in smoke. Are discussed and some really nice papers on the recent advancements in measuring smoke properties that both my team and Lukas' team have done. Anyways. Thank you for listening as usual. Next episode, we'll be waiting for you here next Wednesday. So let's stay connected. , share the episodes with friends, please. Please do that. Please help me sharing the knowledge about the podcast and yeah, see you around!.