119 - Different scales needed to model fires with Lukas Arnold

I've invited a guest Professor , lucas Arnold from University of Uppertal and the Ulish Forchance Centre , who is developing interesting approaches to modeling simple materials and , in future , more complicated materials for the fire safety in the railway industry .

In this episode we're going to talk about what it takes to truly model the fire behavior of a material and how these approaches , the properties of materials the things that really matter change when you think about different scales at which you model this behavior . Very interesting , perhaps .

On the difficult side , the episode is certainly not easy , despite the fact that I'm having a fire laboratory . Even for me it was difficult at some parts because we're going quite deep into the modeling aspects , but I've decided we need to show it like it is , because people need to understand how challenging and difficult this approach to truly model fire is .

So I hope you'll find it inspiring , interesting and if you happen to be a researcher who is also working on modeling fires . I am absolutely sure this episode will be a gold mine of knowledge and inspiration for you , so , without further ado , let's spin the intro and jump into the episode . Welcome to the Firescience Show .

My name is Vojci Wigzynski and I will be your host . Firescience Show is produced in partnership with Offer Consultants , a multi-award winning independent consultancy dedicated to addressing fire safety challenges . Offer is the UK's leading fire risk consultancy with the globally established team .

I've just learned that Offer won the Small and Medium Enterprise of the Year award in the 2023 Engineering Talent Award . I've met a lot of Offer's talent and I understand why this award went to them . So huge congratulations to the Offer very well deserved award .

And if you are a talent that would like to join such an esteemed team , offer is always looking to hear from industry professionals who would like to collaborate . Get in touch at OfferConsultantscom . Hello everybody , welcome to the Firescience Show . I'm joined today by Professor Lukas Armold . Hello , lukas , hi Vojci , good to see you again in the show .

We've talked a lot about modeling stuff in computers and I know that you have some very nice interesting developments in the field of modeling fires related to railway fire safety in your project basket . I love the name , by the way .

So I would love to pick your brain on how are we modeling fire , spread fires , fire kinetics , and what does that mean for fire safety engineering overall ? I guess we'll take a deep dive into modeling , but first let's hear about the context . So why your research interests have focused on modeling how things burn in the railway context ?

So the idea in the basket project is to move forward to the field of fire spread modeling and the context , as you said , is the trains , and this is due to the fact that we have in Germany funding options to do research , safety research that needs to benefit the society .

So in our case it's to make trains safer or do the process more efficient to make trains safer . So that's why we teamed up with two companies , one BCL from Leipzig and two Zuit Rail . So the guys that we work with are situating in Berlin and I think they main headquarters in Munich and they both work exactly on that field .

So they work on the safety of trains means regional , national trains , but also trams and so everything , and the intercity trains as well , or just original ? So , in general , I would say they work on all these kinds of trains and they run simulation , so that's why we teamed up .

So they use , of course , different approaches to predict fires and especially fire spread in trains . So they are very interested in having new models or even faster models to come up with , design fires that then can be used to alter and station safety and so on .

So we as researchers and by we I mean and this is Fabian Brünschtl from the University of Hoppetal and Alexander Bay from Aachen and me from Ulrich and from Aachen that we come up with new approaches , new experiments to help these guys to work on that . Of course , it's always everything that we do is open access , so we're not doing this work only for them .

We team up to the process , but everything is pretty available .

The question is how well it is , and the role of scientist and engineer is to judge on that . I mean , that's the idea of an endeavor that you have also had in your podcast about the MacFP where people team up , many people team up to understand fire growth , fire spread of DMMA . So it means without going to need it .

Obviously it's a challenge , so you cannot just use these parameters and it's going to work , and there are really various reasons for that that it's not working .

So what I meant in my head was you look up literature values , as you said , put it into your fire model like FDS , do a flames spread and then compare it to an experiment that you have modeled , a real scale experiment , and then you would compare different properties , each real-estrates mass loss rates maybe , and some velocity fields and whatnot , or temperatures of

the solid . And I would say , in general , that's not going to work this way , but you need to check and the challenge here is and this is also what , in my impression , we are also facing that we need to be able to model the real scale also . So what we focus on is mostly the paralysis , so the solid phase or the fire propagation process .

But once you have a large scale experiment , the gas phase is also very important . So what is the relative fraction of your fuel that you don't know ? So you come up with surrogates . Even if we may name them methane or whatever , this is in reality probably just a surrogate . So we don't know these properties .

And in a large fire you need to do the interaction of the gas phase and your solid and the overall geometry right to judge whether your predictions based on your material parameters are correct or not .

So everything can match the experimental data , maybe due to the fact that your material parameters are wrong and your numerical model is also not correct , and then by chance it works out . I think that's a challenge , and it's a challenge to run large scale experiments in a lab .

So therefore , I think it is very , very important to have reproductions of experiments . I mean this is the goal also of MacFB , where different laboratories tested , or XMI , the same material . That's very good , so it works out nicely on the microgram scale , but when it comes to the real scale experiments , that's something that we need to do as well .

Okay , so you know different labs , different environmental conditions and so on , because then it's still hard to judge , like , is my heat release rate curve off because something happened in the experiment or is it off because my parameters ?

are wrong and therefore I think we need to have a more extensive experimental database on real scale and even do the same experiments . It seemed boring and I know it's hard to sell to people giving out money , but it would be , in my opinion , a huge value to have the same experiments run by different groups .

You know , because we are very based in the paradigm of design fires .

You take a design fire , you put it inside a room compartment , a railway cart even , and if you have a design fire that has predetermined the release of heat , release of smoke , release of toxic pollutants , there is nothing in kinetics to model in that you simply want to release this amount of Outcomes of the chemical reactions into your environment .

Though I also working in this field especially is when you work in experiments , in large-scale experiments , you really see the variability of those fires . Like if you burn Crips under the hood , Freestanding on the ground , you have 20 cribs . You burn them the same way . They're gonna burn more or less in the same way .

There will be some discrepancies between one test and another , but in general it's gonna be the same fire .

But if you place it in a small room , in large room , in a train , against the wall , it's gonna be completely different fires and I feel that these design fire paradigm Becomes at some point very hard to justify in , especially in small spaces where the flows will very much Define where the fire can be and how can it grow .

So if you had a model that , instead of placing A heat source inside a mock-up train , if you could model luggage seats , floor cables , whatever is the that's combustible , trigger the fire and see how it spreads through the cart , giving you an outcome in terms of the design fire , that would be a huge , huge improvement , because there's so many factors that that

affect

that .

However , I think with increased computing capabilities and so on , we can move forward to tackle the fact that the design Fire is really just an approximation a crude one , I think .

You , that you can come up with material parameters I mean , that's a big topic , how to determine them at all that you come up quickly with the material from just that you need and that you will gain the opportunities that you outlined already .

The design fire will be different if you place your higher Somewhere else , but especially when you want to investigate what are the safety measures , like sprinklers or water mist or whatever , without predicting the outcome of it impact , so that you can evaluate different designs , different setups .

But not then you need to have a model that really interact with its environment . A design fire is basically a fixed book , so , like a screenplay , it's already written down what will happen . In the other case , there was interaction , okay , so that you can really evaluate how good is my safety measure performing or not .

And then the question is but your smoke is like 200 degrees and yeah , because there's like nothing to cool it . That's how it ends up . Now people would take a design fire and then put sprinklers in the model to cool down the fire and and for me this is like artificial because there is interaction with the fuel , with the fire .

Like you either are solving fire physics or you are Substituting this fire with a surrogate design fire with the all consequences of that substitution . Like it's a process , it's a test . I usually brought Comparison to fire testing , fire resistance testing .

Everyone knows that the standard fire test is not a real representation of a fire , but it doesn't stop us from comparing hundreds of different types of solutions with known benchmark to have some insight in how they behave in fires . We know it's not the reality , but it's still useful .

In the same way , you know CFD or any other analysis with the design fire is useful because you can benchmark your building against this test , which is the design fire . But you're nowhere close solving physics , you're nowhere close to simulating a real fire .

Even you're literally testing a building or its systems against a known , a mission of heat and smoke and having the ability to Really account for the spread and the even more Fantastic things that you mentioned suppression reactions between those perhaps even more complicated flame spread , perhaps transition Into , flash over , or so some more complex phenomena .

I think this , this would be a whole new world for fire engineering . I wonder how many people would like it , because the work gives a living good for many people and it's quite convenient . But no , for me as a scientist .

That would be fascinating to have Models for which I could really trust and say that I'm really sure that what I'm modeling is close to reality .

I think that hand calculations are great , okay , so everybody understands it , but you need to know the limits , as with all the models , and if you think that none of the models applicable but the super extensive one and it's worse for a project , then I hope that there will be eventually , at one time , a tool that you can use to do that and you can derive

with this one many , some sub models , okay , so that you don't have to run the complex things all over again .

Okay , having that data , you can investigate the process and the heat as the heat of combustion , for example .

Okay , so these devices don't Provide you the values for your reaction rates it's , for example , the mass loss rate but you can then apply models to find out what would be the reaction rates , for example the a's and ease of an Arrhenius approach to prescribe what the apparatus did measure .

And , tga , perhaps to some extent you can have a sanity check on what temperature specific things are happening . Because , if I currently Understand the method , you do some stuff along a Temperature growth . So if you know that some stuff is happening at 200 and your model would show it at 100 , it's also like a sanity check . It's it's not gonna happen .

However , I find this bounce interesting and useful as bounds . They're not answers to questions . You know why and how things are progressing . That that's how I understand these methods , because I really like working on this micro scale . It's convenient .

So this is already a simple setup , so it's a zero-dimensional thing , basically . Hmm , so , having been the kinetics of your paralysis , you can move forward to the next scale , which would be , for example , the , a classification apparatus or a conch or meter .

A Gathification apparatus is taking samples of roughly just order of magnitude 10 by 10 centimeters or 10 centimeters let's stick with one direction and what it does is it irradiates the sample , like the conch or meter that Probably many people know the sample in a inert atmosphere , so mostly nitrogen .

So what you measure is there are no flames , obviously because there's no oxygen . What you measure is a mass loss of your as it can suffice . You can do the similar thing , not the same thing , with a conch or meter where obviously there is oxygen and Will have then a fire , so that you can measure the mass loss but also the energy release .

And these are the , the quantities that these apparatus provide to you be safe others , but the main ones is , for example , the mass and the energy release rate . And then you do the same thing , that you come up with your fire model . Now it needs to be more complex because you have the heat transfer in the sample .

You've got the porouses with the parameters that you may have estimated from the micro scale Experiments , and if you do a conch or meter , you eventually have to at least my opinion Include the gas phase so that you have also the heat feedback from from the flame . And then you use this model to come up with model parameters for the thermophysical parameters .

So if everything works out , you would have then a set of parameters that you would need to prescribe with flame spread . So all means , the porouses , parameters and Heat conduction , heat capacity and so on of of the sample . Of course these numbers are not just numbers , but that the heat capacity and heat conduction and so on of a sample is temperature .

Okay , so these are not individual values but actually functions of temperature . So we typically approximate them with a few points . But you need to take that into account and get it . Assume that the material keeps its properties .

So if you could , you know , just run a cone , test on a material and then have a complete no knowledge on on the behavior of the material , as at this scale you , as you mentioned , having intercount some heat transfer and some other material specific properties , it would be fantastic just run the cone and , and you know , you have a working CFD model , but but

I don't feel that's reality , though I know a lot of people are trying that .

I think it's not easy , and maybe we are not there yet , but maybe that's doable .

So there is , you know , a relation overall , if you can predict the SBI , it kind of means you predicted the flash over in room corner tests . So it is in a way , prediction of the very large scale behavior with a very small method , because you perhaps get most of the important elements out of it .

But again , if you're talking about cone , you're not just talking about take one sample , 50 kilowatts , burn it down . You know everything . It's more than that . It's multiple repetition . It would be multiple radiations , right ?

But right now we use one or few of them to come up with the material parameters , and what needs to know is that with this approach I mean the model of the cone , kilometer and the sample . It's very simple and we cannot do the assumptions that we did on the small scale like that . Everything is basically zero dimensional , so it's like a super small sample .

So if you go to the eventually simplest material PMMA that we hope that is the easiest to grab and if you have a look on that , maybe you can add it into your podcast you can see bubbles . You cannot see them , but you can hear them . Okay , so there is a bubbling layer on top of your sample where it gets irradiated and you hear it cooking .

That's something that we don't have in our model , that there was a bubble layer on top which obviously has different material properties than the solid , for obvious reasons , but it strongly interacts with the heat transfer and mass transfer processes and relations also .

So we have a set of parameters that describes this process , but we have to take care about saying , like these are the parameters that are the properties of the PMMA , no , it's the of the system itself , like having this layer of bubbles on top of that if you think about wood or timber that creates tracks and so on .

That's something you cannot kind of cover , not yet at least . So the set of parameters that you have is an effective value that represents these behaviors , which is in a sense , fine , because this is what's going to happen later on and it depends , as always , on the scale that you look at .

Okay , so we've got also other processes where we know , if we zoom in , they are completely different , but we come up with a model to describe what is happening . Okay , so it's fine to have these parameters , even if we know that they are not the true one . A simple example for that view , as a CFD person know , is turbulent model .

So the turbulent viscosity well , that's an artificial parameter . Okay , that we , if you would zoom in and really resolve the eddies and the dissipation of the eddies , go to the Kolmoro scale and make it even finer than then , you wouldn't need it .

But you say , okay , I cannot go down to that and I describe the effects of this dissipation and so on via a larger scale process viscosity in that case to describe what is going on on the smaller scales .

And you can go into large eddy simulation where you assume that for large eddies you solve them , for small eddies it's average . And you could go DNS , where you just model everything up to the tiniest little element that's physically possible . In the end you hope that the result of all of them would be the same .

The computational cost is drastically different at these levels . Is it the same with the pyrolysis and the process that you're describing here ?

Now we are zooming out , so we take the system with this , for example , bubbles and so on as a system and figure out how does the system react to heat transfer and other things .

And what I often say here is that , yeah , these parameters may be grid dependent that we come up with , because eventually , the model that we use well , the grid resolution is part of the model itself . Okay , so this is really the whole setup and , yes , this is not nice .

I would also like to have something that is not eventually not dependent on grid parameters . But if you think about what we are using in turbulence , again about the NES and Spargramski , then you have also the filter width , which is the grid resolution part of the constant to compute the turbulent viscosity .

So I mean , maybe it's not nice , but it's how things work that you have really grid dependent parameters .

But you're not solving for that . Cool , I have a follow up for that . I know in your project you also expect to do intermediate scale with like a kilogram size samples , 50 centimeter tall ones , and the real scale which I would assume like is like real objects or parts of the trace .

If you already learned that much in the small scale and if the challenges are in making those models zoom out , you won't learn more about chemistries or kinetics from larger models what you're learning from them then .

Because in a cone parameter you have no flame spread . So this system is so strongly driven by the heating element . So I mean you're putting a lot of energy on the temple , okay , so there is no kind of dynamics given by the sample to impact the flame spread or whatever . It's a strongly biased , if you want to a strong driven .

So if we have flame spread or flame growth over a small sample , then we can investigate computer models , for example also the sensitivity of these parameters . And the important point for me here and this is what one of our PG students is doing , tasia is to investigate the sensitivity of the parameters at the different scales and how this transfers .

So imagine , let's pick one of the no , we will not name it , so that no one would say the one of the parameters that we have that we talk about is very important for flame spread . Okay , anyway , I don't want to name something so that no one would see me on that .

So if there is one that is very important , but it's important only for flame spread itself , then maybe it is having little to no impact on the in the cone kilometer .

Okay , so it means that when we derived the parameter from the cone kilometer or the other scales , this parameter , when you vary it in order to find the its current value , will always give you good results .

Okay , you move it by a factor of two and the results is still good because this parameter has little impact on the outcome of your heat rays rate in the cone kilometer . Okay , so this one will become very unsharp in the prediction on this scale . So you go with something that is kind of unsharp , if you know what I mean .

So its value is kind of ambiguous within different grade , specific range , and moves to the flame spread experiment and the role of that experimenter becomes now significant .

Okay , and the question is is this happening ? And if if it is , how can we find out which of the parameters at last give are important to do that , and how can we design or modify our approach to address this parameter ?

Okay , is there a way , with the existing experiments to become sensitive in our approach for this parameter , and I think that we are working on that and I think , with different optimization techniques , that we really watch out for all parameters and find also cost function to address them , that we will be able to do that .

But like the question is well , is it really so that we can kind of pinpoint or narrow down the parameters at all scales that good as it's needed for the later real-scale predictions ?

In the full scale , if things happen faster , you have larger part of your material burning , which means much higher heat release rate , which means much stronger feedback is , you know , like Self-fulfilling prophecy , leading to Instabilities , flashovers and stuff like that , the the world of fire . So in the full scale , this feedbacks start to be very interesting .

And now coming back to what you've learned at smaller scales , even if you did cone at the best , you've done it at maybe three heat fluxes . Perhaps you've done only 25 and 50 , and the feedback will be whole range of heat fluxes from zero to Whatever number you would have , from a , from a flame directly impinging on on the surface .

So there's material properties figured out with cone at these two levels . So with what confidence they give you answers across the heat flux scales You'd find in real fires when the flim suit is actually spreading ?

And so what we typically do is that we develop a model based on one heat flux and then use the other one for testing . Okay so , because in the ideal case , you , once you I mean your material parameters should not be dependent on the Heating rate of your heater .

You know that's Not , that's not good , but you can , with this one , estimate how well or not you can get there . So that's already a good , a good thing . If this matches nice , and what we'll needs to keep in mind is also that we are speaking in the conglomerator about the heating flux Provided by the heating element , but there was also the flame .

Okay , so it's not that the sample will see only 50 kilowatts , but maybe it's seeing 60 plus 25 , or I don't know how much , but it's gonna be definitely more than that , because there was still the claim that is close to the surface , so it provides also a lot of thermal radiation to the , to the sample , so you already have a wide range of Radiation , but

still , in the real case , as you say , you will have different Properties .

And the second part of the question is can you do it with literally a very simple equations that are already built in the softwares , or you would have to like take additional software outside of FDS ?

You know I'm thinking about the , for example , the difference between Pyrolysis models in FDS and GPYRO , which is a super sophisticated tool for modeling pyrolysis that exists as a separate package .

Okay so like , let's say , it's easy to predict a pyrolysis . You can know the parameters because the Arrangement approach is well working out nicely and the E-Conduction equation is also not so difficult to solve . That's what I'm not sure . But the question here is we want eventually to predict reality , and so we need to put in the right numbers for the models .

So what you would need to come up with is what you all don't need to do when you do a design , for that's a benefit of it . There has been no Interaction between the solid and the gas phase . You just predict how much mass is used .

So it means in order to have gas released , you would have to do pyrolysis , which means then you would have to define what kind of reactions you would want to have . So let's say there's just one for simplicity . But all what we say is you would have to multiply it with the number of reactions .

So then you would need to specify the Arrhenius parameters , so activation , energy , exponential factor , order of reaction , and you would have to specify what kind of gas is released , okay , and how . What is the heat of combustion ? What is the heat of reaction for the pyrolysis ?

So it means that you have to come up with the density , the heat conductivity , the heat capacity , and all of them are in one can provide just single values , but , as already said , you would have to come up with temperature functions of these quantities . And beside that there was the interaction between the solid and the gas phase , especially for radiation .

So you would have to also come up with emissivity and , I think , the material properties like density , heat conduction , heat capacity . That's something you would need to come up , of course , also for every layer of a material . So if you think about cables and other composites , then you would have to do that for every layer that you have .

So I think then you would be good to go . So that's a bunch of parameters and that's the challenge in all of this because , as I already said , you don't have an experiment per parameter so that you run this experiment and the experiment itself tells you this value for this parameter .

So we have always this indirect measurement , so we measure an integral property , whatever it is heat rays rate , for example and then deduce based on that . Given a model , what are the model parameters to get there ? So that's the challenge in this .

Whether it's going to have any practical applications in engineering or not , I understand that , but it's not going to get easier with PMMA , to be honest .

I also think this is a part of your MAKFP experience , right Exactly ?

These experiments are part of the FDS validation guide and of the MAKFP part for the solid phase . And , yes , we work on the prediction of the heat rays , rays and radiative heat fluxes on the walls and so on . And , if I may advertise , this year we just published a paper on the full process .

What I was just saying from the micro scale , I think it's named from the micro to the real scale Fantastic .

So what we want to do in MAKFP like this was I think we extended our view on that is to see , like , how well do the material parameters behave ? Okay , so this is what we all said . Okay , everybody comes up with material parameters and then we test them on different scales and also go to the real scale .

The problem at the real scale is that there are more effects that we would have to cover . Okay , so there seems to be a slattering of the flames that you would have to include into your model as well . They may be in homogeneities in the burner itself , so that they send . I think it was sent burner anyway .

This doesn't seem to be so homogeneous as you assume it in your models . So you're starting with a model that may already come with deviations by itself , and then it's hard to judge . How well do your material parameters now perform ? You know what I mean . So there are some other uncertainties in the setup and also the interaction of the gas base .

So what kind of fuel do you release from your sample and what is the relative fraction of that which has a huge impact ? So it's because it's not as in the gas burner , it's not methane or whatever that you know , but it's something different .

Yeah , so I think the next steps are definitely to combine also the gas phase , make Fp efforts , because fire , as you know , is a combination of all these things , and I think we all know that we shouldn't be so naive that we can separate things we need in fire , what distinguishes us from combustion community and so on , that we need to take care about

everything , and I think this is the point where we need to work together , also with people investigating the gas phase and the real scale , because this is now you know . Then you get more uncertainties . Then , for example , the gasification apparatus , whereas there is no combustion .

That's fine , but well , zooming out means you have to take more processes into account .

You know because we're not building buildings out of MMA you know , and I would expect , even at the simple in homogeneous materials like textile covered foam materials like the ones that insulate my children from the podcast audience , or upholstered chairs , up to very complicated materials where you have multiple different plastics altogether .

You've mentioned research on cables before . There's also complex material Like what are our chances against those complex materials with the flagellates ?

So you say this is a complex material , so many layers , many different layers , but I want to have a model that presents its interaction with being heated up with fire eventually , so the outcome to predict how does it release mass , and what kind of gases are released combustible , non combustible and treat it as a one surrogate model for the material .

Okay , so I think this is how we do it right now with many things , and I think that you can also work with that , probably with composites .

But this may not be true , okay , because you may have things that really significantly change the structure in the process of heating up and there may be also time dependent things going on , and then maybe these things don't work out .

The other approach that you're trying to move into is to look really into the materials , but then things will become less applicable for kind of everyday engineering applications but be more for the kind of research part , at least at the moment at least , where you would really model what is going on in the solid phase , and of course then we have to come up

with different models , so it's then at the S will not be sufficient , but the idea is to team up with people really do material science so that they can really go to really low scales .

I mean for homogeneous materials , maybe you can still stick to that , but if you think about timber or paper or other things that have small structures , to really work on that scale and then move up and up in scale , because that's something we cannot work and hire on that scale , but all the instructions that you learned on the small scale to move it up to

bigger scale so that you have also then affected models , but those that are based on different scale investigations .

? They're having great efforts on on this colorimetry database . That was that went through an AI and AI now can predict to some extent how big fires are and how things can burn . So how do you feel ? I mean , for me , the immediate problem is that it doesn't take into account the context of the building , which means again , we're in the design .

Fire versus flame spread discussion .

So what new things can we learn from that that we're not already part of the training ? So I cannot judge on that . Me as a geotechnical physicist and now also becoming part of physics , we've got good laws like equations to describe these things . And in other disciplines , when you go to material science , this is what other people do .

They can really do the interaction of small now we're talking about really small structures inside of the material and we can also learn from these communities a lot . And so this would be also an approach .

And if the model , the AI model , never saw water mist , will it now predict how the couch , if you stick to that will , the heat release rate of the couch will change if at one point the water mist system is turned on or off ? So now I think that's another toolbox . Right now we stick to the classical approach .

However , in the BISC project , one of the goals is to actually come up with an AI model , but to predict the material parameters , and it's a different approach . The prediction itself is still done by in our case it's FDS but the way of finding material parameters .

So , for example , we have an in-house code that's also pretty available , which is called property , to do the inverse modeling . Now we use a genetic algorithms for that , but the idea is to do the optimization process with an AI model . So basically , the AI model doesn't predict anything .

Well , we predict the material parameters , but then you check it and if they set your experiment data , then it's the same as the classical optimization which do .

But of course we hope that it's going to be way faster , because if we go back to BISC the engineering companies that work with us the main challenge is to find the material parameters , the actual simulations . That's fine , but we need to find the material parameters .

This is super costly and they want to have a quick way to come up with the material parameters . And that's where our AI model kicks in , hopefully to provide a good estimate , maybe as a pre-processor for a kind of classical optimizer to come up quicker with the material parameters that they can use for their playing strategy model .

So there are different ways to use AI , of course nowadays , but in that case we wouldn't use it for prediction , but only as an optimizer .

It's not as easy as it looks and I hope this discussion gives them another depth into what does it take to model things and how much effort it is . And I'm really happy with the VmAC FP .

I mean I'm really looking forward to Japan and it's going to be incredible hope to bring up this research and those experiments to the audience of Fire Shines show as well soon . So thanks , man , for being here and all the best . See you in Japan , man . Thank you very much .

It's not so easy but definitely worth it and definitely needled for the evolution and future in fire safety engineering .

But that in mind , when someone approaches you claiming that they've just model flame spread over complex material in a compartment fire and they've done it at 20 centimeter measure resolution and in one day , that's not really the thing that we are looking for . So hopefully the world of fire modeling is a little bit closer to you now .

And , yeah , let's cross our fingers for the developments in basket project and developments of Lucas team and let's see where the world of modeling goes .

And the next chance to update on what's happening around this . And I've assessed Congress in Japan where there will be MacFP workshops .

I've covered MacFP with our note through there in the podcast previously , so so my hopes for this workshop are very high and I hope we will see a lot of interesting data that will help scientists develop their models in the future . That's it for today . Thank you for listening and see you here again next Wednesday . Bye .