173 - Pathway to scalable fire CFD

I want to talk today about scalability of CFD simulation , that is , the computational fluid dynamics simulations in fire , or in other words , the future of CFD simulation . That is , the computational fluid dynamic simulations in FHIR , or in other words , the future of CFD simulations in FHIR .

Or maybe even one could say how could simulations in future , and maybe even today , look like instead of what they are ?

I've got inspired to talk about this during our summer school that we've organized at the ITB together with Zach Frisby , and during that summer school we had excellent speakers in the house , including Professor Jose Dorero , who said something along the lines like he would prefer not to do CFD at all , and that was something targeted at the young fire safety

professionals who were learning the performance-based design , and I kind of understand where Jose got to this point . I mean , the context was that a good scientist , good engineer , can figure most of the stuff using fundamentals . It's more about the engineer than the tool .

Many times when we use CFD , it's just an overkill , it's unnecessary , and I agree with that , and I agree with that that such a use of CFD is very unnecessary . However , I still believe the CFD could be used differently , and I know that because we've used the CFD differently in our research projects .

And this use of CFD is something that I want to cover in this podcast episode . The timing is good because just tomorrow I'm going to Ireland for an IFE conference where I will talk about scalability of CFD and I will show some practical developments from our office .

Next week , I'm going to New Zealand where I'm gonna present the case study of a car park and how we've processed this , with some new concepts related to processing CFD . It all ties back to this idea of scalable CFD and this episode is kind of a context for those two talks .

So , yeah , that's the context of today's talk and I really , really hope you will enjoy it . Whether you are a CFD user or not , I think this may be interesting to you . Let's spin the intro and jump into the episode . Welcome to the Firesize Show . My name is Vojtěch Vyjgřínský and I will be your host .

This podcast is brought to you in collaboration with OFR Consultants . 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 environment .

Established in the UK in 2016 as a startup business of two highly experienced fire engineering consultants , the business has grown phenomenally in just seven years , with offices across the country in seven locations , from Edinburgh to Bath , and now employing more than a hundred professionals .

Colleagues are on a mission to continually explore the challenges that fire creates for clients and society , applying the best research , experience and diligence for effective , tailored fire safety solutions . In 2024 , ofr will grow its team once more and is always keen to hear from industry professionals who would like to collaborate on fire safety futures .

This year , get in touch at ofrconsultantscom . Okay , scalability of CFD . You know the background . You know where the idea came from . Let's figure out how CFD could be done in a way that would make us proud , maybe even could make Hosea proud . So let's first think about what CFD computational fluid dynamics tools are for fire safety today .

I guess I should start with the very , very beginning . So what's CFD ?

Cfd computational fluid dynamics is a way of solving momentum pressure , mass , energy balances in finite spaces that we've subdivided into smaller chunks that we call cells or elements , and basically we solve a lot of physics for every little cell within our model to figure out where the mass and energy are going within our buildings , because we're used to simulate

buildings . It's a very clever tool based on late 19th century mathematics . Our modern computers allowed us to solve for that . And here we are . It's used all over disciplines . You know , airplanes fly because of CFD , the wind turbines are able to generate energy because of CFD , the combustors and engines work because they're designed with CFD .

Formula One teams win their championships because of CFD , and so on , so on . There's so many uses of CFD because it's a really , really powerful modeling tool , a tool that allows you to model physics , to capture physics . You don't care about phenomena which we have in fire , like flame spread and trainment and so on . I mean you care .

This is your output of your simulation , if you've done it correctly , but it's not an input . The input is your energy , is your chemistry , and then the solver shall solve everything on its own . That would be the perfect CFD shall solve everything on its own . That would be the perfect CFD . However , not that simple .

Of course we need to do a lot of simplifications . I guess I'm not going to go over and over CFD in this episode . I had five fundamentals on it . I had a really good episode with Wolfram Jan , one of the early episodes of the podcast . He's my dear friend and we've talked about the use of CFD with him many , many times .

I've had a really good episode with Jason Floyd about mesh convergence . So there's a lot of resources in the Fire Science Show which I will link in the show notes .

So if you are new to that tool , if you're new to the computational fluid dynamics and in our fire space , many people would simply say they , they use fire dynamic simulator , fds , which is in a way in our space , synonymous with cfd . Fds is a cfd solver .

So if you would like to learn more basics , there are resources for you in the fire science show and you can find them . Now the question was how do we ? We use CFD in today's world , in today's engineering , and I've just had an episode with Jonathan Hodges from Jensen Hughes a few weeks ago and we've kind of discussed that .

We've discussed that there are simulations for fire effects , like the smoke spread that tell us how smoke spreads within the buildings and they help us design things like smoke control , for example .

And you could technically do simulations in which you try to predict some of the fire spread , how the fire would spread , how the fire would grow in your building very challenging but maybe possible . I mean you have to listen to that episode to get a more complete opinion on that .

Anyway , with those concepts , the way how CFD is used today , I would call it explanatory mode . So we use CFD to explain things . We use CFD to explain if a smoke control system is working . We use CFD to analyze and explain if there is sufficient time for escape from people in the building , if the smoke layer descends , if the visibility is maintained .

Those are the things that we're trying to figure out with CFD simulations . And by explanatory mode I also mean that they are usually done after you've done most of your engineering . They're kind of expensive , kind of slow .

When I started doing them 15 years ago it was not very rare that simulation took you a week of your time and I know today many people would still struggle with that because in some complicated models it can take days of time to process simulation .

So you're not really able to run too much of that and you're kind of not really able to run this as a design tool . You're not able to run this early in your project . You're definitely not able to run this as a part of a parametric design of a building . That's something we've talked about with Benjamin Ralf in the podcast .

Wow , so many podcast things that are going to go to the show notes this episode . Wow , so many podcast things that are going to go to the show notes this episode . Anyway , we have an access to a very powerful tool that can , to some extent , predict physics or represent real physics in real buildings , in large fires .

Yet we're extremely limited in our ability to use that tool . In the end , what we really want from CFD , what most people really want from CFD , is a bunch of images . Therefore , the nickname for CFD is cartoons for directors or colorful , fluid dynamics .

We want some sort of proof that can be given to our authorities as a demonstration that our design works , that our design meets our goals , meets our objectives and hence the explanatory nature of the simulations that we use today . And I disagree . I disagree that this is the best way we can use this powerful tool .

Come on , people are not using CFD for airplanes to explain that their design is actually flying , to explain that their design is actually flying . F1 teams are not using the CFD to show that their car will be the fastest one . No , this is on track . They use CFD to optimize the hell out of the winglets and other things on their car to make it the fastest .

It's not explanatory . They are exploring the designs . They are figuring out how to build their car in the best way so it can win the race . And the race is the final check .

Whether it worked or not , and believe me or not , but the teams who were first to introduce massive CFDs in their R&D process for their vehicles , those were the teams that were winning F1 in the early 2000s . Later on , fia that's the Formula One administration , let's say had to put limits on CFD because of how powerful that was .

So I always looked , you know , at these other engineering disciplines , looking how they are capable of getting so much out of the simulation in the design process and they could not design without it , you know and then going back to my desk and facing the reality that what I'm supposed to do is just run a simulation to prove the design I've pulled out of my

Excel sheet based on some equations , standards and so on , just to prove that it works . That's not enough for me . That is not enough for me and I really , really hope we can change that . And I know the pathway is not easy , but I have some ideas how this could improve . So now let's do an exercise together .

How would it look like if we were to use this powerful CFD tool like a Formula One team in fire safety engineering ? What could we get out of it and what would we need to use it ?

From my perspective , if I had infinite resources , I would just run hundreds or thousands of simulations with variables put into the CFD solver as some probabilistic distributions and get the probabilistic outputs of the simulation , know how often specific parameters exceed my values , get some ranges , figure out the outlier scenarios and try to understand them why in this

particular scenario something went very , very wrong and just take a more risk-based approach for assessing safety in my building . That would be very helpful and very useful and would require hundreds of simulations . In fact , things like this are done today with zone modeling . So if you take bRISC , a zone model from BrandsFire , you can already do that .

You can have probabilistic distributions on your inputs and just run thousands of simulations very quickly . It's just a zone model , so of course it has some limitations , especially related to space . Because again , if you think about it , why do we want to use CFD if we have all the other tools

? The most powerful aspect of CFD is the fact that it can include for the spatial effects not special effects spatial that are happening in the space of your buildings , and this includes the geometries , the architecture of your building , the way how the building is shaped .

Every little detail in your building may influence the flow of smoke and will create very complex entrainment patterns around it . So every time a smoke will flow around a beam , let's say , it's going to change its direction and it's going to start mixing with the surrounding air .

This is something you will never be able to catch with a simple zone model or Alpac's correlation for your ceiling jet . You only can capture that with CFD .

So this aspect , which kind of for smoke control is prevalent in buildings , is something that we really get out of CFD , and that's the reason why I would like to do those probabilistic simulations with CFD and not with B-Risk or CFAS , for example , because those will not give me all the information I want .

Another way I could use CFD would be you know , have a building and I have to design smoke control . What's better ? Two extraction points , four extraction points or 10 of them ? What's better , 100 cubic meters , 150 cubic meters ? 200 cubic meters per second ? How much the location of the fire will influence the design ? Is there any specific location ?

I need to put the smoke curtain inside ? A ton of design questions which we are answering based on our experience and based on some general frameworks , guidances that we have for smoke control design , and I could technically simulate all of those . You know I could create a real design exploration with cfd if I had resources to .

That allows me to compare the outcomes of all of those at once and pick the one that works the best and is most cost effective , and that would be a brilliant design tool for smoke control in large spaces .

However , that brings additional layer of challenges that lie not just in the computational power , and that's kind of what I'm going to talk about in IFE conference and in the New Zealand . Anyway , with the computational power , I may surprise you , but we're almost there .

I truly , I honestly believe that the computational power required to use multi-parametric , large-scale CFD simulations , like running hundreds of simulations . For your case , I think it's very possible today and it would not be excessive amount of money necessary to run that . It's not in millions , it's not going to take a year to run those .

Today we're in the era of GPU computations , so GPU means Graphics Processing Unit . Those are the parts of the computer that are responsible for rendering the computer graphics and , as you know , the video game industry exists and the video game industry thrives for better graphics since its inception .

Some years ago I would say 30 years ago , I think I was like 95 or something there was a game , turok the Dinosaur Hunter , and I think it was the first game that included video acceleration as a main concept , and for that you needed a special GPU unit alongside your computational unit and the GPU was calculating stuff related to graphics .

You need , and the GPU was calculating stuff related to graphics . For graphics you need a lot of interesting things related to how the light shines on your objects , how the shadows are formed , how the textures are implemented on your models .

So that requires quite a lot of processing and because it happens while the player is moving around through the world in which the game is played , you need to be able to render this very quickly , so a new tool was conceived that was the GPU , the graphical processing unit , and this tool was supposed to compute this very quickly in a parallel to the processor .

As the time flew , those units became more and more and more powerful . I remember when PlayStation 2 was coming out . That was insane how much computational power that device had compared to normal computers and other gaming consoles . But that was a long , long long time ago .

We had many revolutions in GPU and today these graphical units are unbelievably powerful supercomputers on their own . The only difference is they have different architecture . You cannot solve equations directly on GPU . You have to use some tricks to use this capacity .

They are normally used for rendering graphics to solve your equations and of course , this revolution is happening . We've learned how to program our CFD computational fluid dynamic codes for the GPUs . For example , ansys that we're using is now fully capable of running on GPUs .

We have a GPU setup in-house and it's ridiculous how much quicker the GPU calculations are compared to the CPU calculations the one that would happen on your processor .

John and Hodges told me that FTS is also developing some capabilities for running on GPUs , so perhaps this will be available to everyone in open source at some time , but for now , okay , perhaps a little limited based on the solver that you're using , but still technically possible to use a graphical acceleration for your simulations , decreasing the computational time by

a factor of 10 , 20 , 30 , depends on how big your model is . You really could cut down from simulations that took a week a few years ago to simulation taking one two hours today . That's the increase of computational power we have .

And in parallel to the GPU revolution we also have machine learning revolution , where people are trying to teach machine learning algorithms to run alongside CFD to solve perhaps the parts of the model that you're not that much interested in . Like , imagine you're simulating a shopping mall .

You would use CFD to solve for the smoke plume region , maybe the inlets , the outlets you know the spaces where a lot of mixing happens but all the hundreds of thousands of cubic meters of the shopping mall in which not much happens .

You could run through a machine learning algorithm and just get an approximate outcome of that empty space , which would be on par with your CFD solution .

So you could technically couple CFD modeling with an AI-based tool to get a significantly reduced computational workload for the CFD where you solve it only for the parts that matter and you approximate all the rest . This is also not something hypothetical . This is something that is happening right now .

There's a company in Poland called ByteLake who's developing solutions like that and we're kind of working with them to explore this in the fire space and it's very promising . So perhaps not just the GPU revolution but also AI revolution will bring the cost of single simulation to something similar like zone model .

So I would risk stating that the most important factor in moving from using CFD as explanation into using CFD as exploration tool it already happened we have the computational power that we would need , like you could really almost almost use the CFD as you would use a zone model a few years ago .

Maybe it's not an instantaneous response like in a zone model it's still going to take a few minutes , maybe an hour to calculate , but if you can schedule them and program them , you can do 100 simulations across three days . That's more than sufficient for design exploration that I am dreaming of .

However , I'm very practical in what I'm doing and how I'm thinking about it . You know , for many people and this is a trap , kind of in the CFD and the increase in computational power . We had those tenfold increases in computational power across the last decade , but yet our calculations times have not decreased by a factor of tenfold .

And there are reasons for that , because when we gain access to more computational power , there are those temptations that you could do more with your CFD . You could include more physical models or more detailed physical models into your calculation .

That means more equations to solve , more complicated equation to solve , more products that you would like to track , and this adds up very quickly in your model and could increase your calculation time very , very quickly . And models are growing . By the way , if you compare FDS version 6 to FDS version 4 , it's a completely different solver .

It's so so much bigger . There's also perhaps a need to go for better meshes , something we've talked with Jason Floyd . Like people would immediately assume that the smaller mesh will give them better solution in their CFD simulation . In some cases , yes , this is true . In many cases , no , this is not true .

The mesh must be accurate for the type of phenomena that you are trying to model . A 10 centimeter mesh on a fire could be sufficient .

A 10-centimeter mesh on a 30-centimeter diameter jet van may not be sufficient to solve it , so a lot matters on the context and what you're trying to solve and there's no ultimate value that will say , yes , this is a good mesh , you're good .

And some people would just like to cut the meshes down and this considerably increases the time that's required for computations . I mean , for good reasons , they want better simulations , but it's not that simple . It doesn't converge that easy that smaller is always better .

Another problem when people get more computational power is they could go a little crazy with the number of variables studied .

I said that when I would be exploring my smoke control in the building I would have those four types of grills , maybe three extraction points , maybe three sizes of fire , add a smoke curtain and not to that , and look , we're already at like 72 simulations with just that simple combination . So yeah , it's a trap as well .

Some people with too much computational power would get into a trap of running too many variables and then they are simply not able to process the data that they just got . So that's a big challenge . I try to be very pragmatic and practical about what I do with my simulations and how I run my engineering business in the ITB .

I believe the true limiting factor of how we are doing our CFD is not the computational power but that ability to process the data we get in the context of the lifecycle of the building design project . I mean , if you look at it , the client doesn't care if your simulation takes a week or an hour to complete .

What they care about is that the report is delivered in a specific time in their design lifecycle , that it fits their objectives , that it confirms their design , that solves some specific engineering problem , that assists them , that can be presented to HHA and so on . It's the client , it's the deadlines that define how much time we can have .

So if I 10 years ago had two weeks for a project in the year 2024 , I still have two weeks for the project . It did not change that much with the growth of my CPU computational power .

The question is how much more I can do within those two weeks , and that's something that I'm very fond of thinking , because now , those years ago I could perhaps run two or three simulations Today within those two weeks . With the new capabilities today , I could do 20 , maybe 30 , if I kept the level of fidelity of the simulation as I did many years ago .

If I want to improve , let's say on a mesh , if I want to include more physics . If I want to be more detailed or more specific about my analysis , I could perhaps still be able to do two or three . I would not be able to do them five years ago but I'm able to do them now at this level . The good question is what's better for my client ?

Is it better to have more of simple simulations and the design decision based on multiple CFDs let's say , 30 very simple CFDs with a little worse , little coarser mesh with the physics that I'm sure is necessary , nothing fancy in it or a few of excellent simulations ?

You know high fidelity physics , very fine meshes , much bigger detail in the resolution of my fire problem , of my smoke control problem in the simulation , but just a few of them . And for me it's almost always that the more of simpler wins over less of excellent . That's my opinion . You may not agree with me and I'm really fine with that .

Also , someone could ask if more is better , why not run CFAS simulations , why not run Zoid model ? And the reason is this ability to include for entrainment ethics around architecture . That's pretty much the main reason I would go for cfd .

But anyway , I truly believe running more , more of more of simple is is better than less of excellent , at least in in the design space . In the science , perhaps it's the other ways . So I'm very into design exploration and now , with this approach , with the new computational capabilities , I can do more simulations for my client .

I can do a proper design exploration . We found another bottleneck , and this one is something that I truly believe that stops us from achieving greatness . This is something that really stops us from being able to use the CFD in a way that I would dream the CFD to be used like . And this is the human . We're not very good machines , after all .

I mean , we're pretty clever . We could write CFD code and figure out how to simulate the fire with it . That's a pretty good achievement for the mankind , but overall we are not that great in processing large amounts of data .

And that data needs to be explained to someone , that data needs to be put into the report , that data needs to be processed , accessed and so on , so on . So the CFD is not just about creating the simulation .

Running the simulation and achieving a solution to the problem that was posed is also about processing , is about being able to extract the important information from the simulation and present it to all of your stakeholders in a meaningful way . Hello , communication , that's what I'm preaching in this podcast .

This is very important to be able to give information to people . If you give people 1,000 colorful images showcasing your smoke distribution every 5 seconds across 20 minutes of your simulation time , this is not very good information . You need to tell people what is the available safe evacuation time .

You need to tell people how quickly the smoke layer descended , what kind of visibility you got . Are the evacuation pathways clear ? Is the firefighter entrance possible in your scenario ? This requires data processing and this requires a fire safety engineer . And I believe this is a limiting factor , a true limiting factor that prevents the scalable use of CFD in fire .

And the good news we're working on that because we first identified that this is the bottleneck . We identified it because of the access to the computational power that we have , because of the research projects that we do For my PhD

. I've done smoke control in shopping malls . I've done like 100 simulations for that PhD and I've brute forced the user data analysis in that , and boy that was . That was fun . I've spent so many days on the colorful plots from that phd project and I've like totally manually went over a hundred of them and compared all those variables inside of them .

Gosh , that that was a painful experience . Since then we've improved a little bit . We've done project for car park smoke control I think I've talked about it in the podcast one in which we found the height of the car park is the most important variable . So in that one we've run like 480 simulations and to process that we've resorted to statistical analysis .

So we've started extracting statistics from the CFD simulations and started comparing the different solutions on a statistical basis . Hence we were able to identify that the statistical output of the simulation , like the average temperatures , the average smoke densities and so on , they changed the most as the height of the car park changes .

And we've also noted that for very low car parks there are almost no changes between different design points of the smoke control systems . So here we've used statistics to analyze the outcome of hundreds of simulations . We did not even need to look at the colorful pictures from those simulations .

But this was quantitative analysis and this is not something you could present to your authority , I believe . Later on we've done smoke corridor simulations . It's more like a digital twin to the experiment that we're running , which you've also heard about in the podcast .

In that project we run literally thousands of simulations and they're like processed along the experiments and they just help us understand what's happening within the experiment .

So another a different approach and I've done my wind and fire CFDs , which you've also heard in the podcast and in wind and fire we wanted to narrow down this to some sort of qualitative analysis , to some sort of qualitative output , and we've succeeded .

We've defined something called operational lap time which told us how often the design is correct and you can interpret this how often the simulation shown that the outcomes are good for us , based on four different criteria that we've defined for the simulations and in this case , the . We've automated the process to some extent .

That used statistical insight but also used qualitative analysis . So this is closer of what engineer would do , do in a building . Now the next step , my next big idea , is to have an automated qualitative analysis . What I mean by that is that an engineer , when processing a project , when processing a CFD simulation , they use very repetitive tasks .

They look at the visibilities on the evacuation routes , they look whether the firefighters can enter the compartment and reach the fire . They look at the temperature distribution what would be the heat fluxes to the floor . What would be perhaps the FED doses that people could take when traveling through smoke ? What's the visibility of evacuation signage ?

What's the visibility of evacuation doors ? You know , those things that are part of fire safety engineering and those are the things that you're really looking for in the project .

So we figured out okay , if we can narrow down the things that the engineer is looking for , can we automate the process of looking , not the process of assessment it's still the engineer that has to interpret the data , whether this is correct or not but the process of assessment itself .

Instead of making engineer look through hundreds of pictures of smoke in the car park to figure out if the firefighters can access , can I just run a script that will create a pathfinding experience in the car park model and just tell me , yes , a pathway exists or no ? There is no pathway . I can do that .

I can also narrow my simulation only to the evacuation escape routes and analyze whether smoke is there or not . So I'm not bothering with the empty , unused space of my building where people will escape from very quickly . I focus on the pathways which they will use for evacuate or maybe the spaces in which the people were queued .

Can smoke , capture them there or not . I'm trying to define parts of the model that are of the most interest to me and run a very specific statistical assessment in those spaces and give the engineering information . Across your evacuation pathways , you don't have any smoke up till 375 seconds .

Across your evacuation pathways , you don't have any smoke up till 375 seconds . In the place where people would queue , you don't have smoke accumulation until 10 minutes into your simulation and , by the way , the firefighters can access the building up till 17th minute of the simulation .

That is a bunch of powerful information that could be automatically extracted without even having your eyes on the drawings , without having your eyes on the colorful images created by CFD . And if you can do it for one simulation , you can do it for 100 .

And then you can have 100 design points and , just based on those information , pick up the design points that are most interesting for you . This is the scalable future of CFD that I see , because we are there with the computational power , we are there with the computer capabilities , we are there with the models that we have .

What we need is to bring up the human part up to the same level of automation and efficiency to allow us to quickly go over those hundreds of variants that we are currently able to analyze . So I'm really excited about this because this is something that I'm going to be talking about in the conferences in the future .

And I'm always asked , wojciech , is there a paper about this ? So we usually don't have those CFD papers published For many reasons . It's actually quite hard to publish CFD papers in this community if you want to do it in a good journal , if you want to do a CFD paper in a fire safety journal , that takes a lot of effort .

But for this one , for the automated qualitative analysis of the simulation outcomes , yes , the paper is in the making , not yet available . It's going to take some time , but this time I'm really focused on writing it because I really want this concept to be established in the community . Oh , and , by the way , you can even go further .

I can actually tell you about a concept that's currently in development . That's the final stage of the idea and that's called the quality of smoke control index . Quality of smoke control index is just an integer value that we assign to the smoke control system in a compartment that represents how good the system is . You know , 10 would be a perfect system .

One is a solution in which people just routinely die Really the worst solution you could have and in between that you have multiple levels at which the system progressively gets worse and based on your analysis on the evacuation pathways of the five-fattest entry of the smoke layer , of the temperature , of the visibility and so on , you can assign progression through

that index . So as my smoke layer thickens , as it goes lower , based on some thresholds I put , the quality of smoke control index would go down . So it would change from 10 to 9 , 8 , 7 and so on . If the temperature underneath the ceiling increases , starts harming people beneath the smoke layer , the smoke quality index would decrease .

If we lose the ability for firefighters to enter the compartment , it would decrease by a lot . So this index is kind of a representation of how many design objectives I'm able to keep and how well , and it just gives you a score . Basically it's one number outcome of your analysis .

So you're running an entire CFD analysis and what you get is one number or perhaps evolution of that number in time . It makes exploring the design space so much easier because you're just comparing one number .

So you can see that , okay , I've increased , I've doubled my capacity of the system , but the quality of smoke control index has not risen by more than one point . That's not much , and perhaps I've increased the height of my building by 30 centimeters and it increased by five points . That's a lot . So it kind of tells you what solutions to pursue .

And the index itself is , you know , a combination of all the things that we've done in our analysis , is a combination of this qualitative analysis of the outcomes , the statistics , the localized effects of the fire that you see in your room , all taken , all merged down into a single number that is very easy to interpret . That's the big idea .

That's something that I hopefully be able to publish , something that I will be able to spread with you . And if people like it , if people take it , I'm sure it could be programmed into post-processing tools like smoke view or maybe pyro sim that could be used to process this automatically along your cfd . And perhaps we could even get fire design explorators .

We have that in in ansys for many types of design , where you just put a lot of your design points and it runs your simulations automatically , just plotting you one variable that you're interested in and you explore your design like that in answers .

You can even create a parameter space in which the design is explored so you can tell the range of , let's say , exhaust capacities or the number of inlets or the size of your openings that you are allowed to to have in your building , and the software would seek for the combination that brings the variable of interest , in this case the quality of smoke control

index , to the optimal point . So it's actually the CFD simulator doing all the work in finding the most optimal design based on the parameter space that you've set . That's really cool . That's my dream .

I would love to run FireCFD like an F1 team and there's no federal administration that will say oh , you have too much computing power , that is unfair to your competitors , cut it down and do less . No one is going to tell us that and I think that's a huge opportunity we have in the world of fire science .

So that was me rambling on what the future of CFD could look like . For me it's kind of the reality of CFD and it looks a little bit like that . It's a lot of effort , but it's work in progress and I see this work is well worth it . Thank you for listening to this podcast episode and staying here with me . I appreciate that .

So , as I said before , I'm entering a crazy travel period right now . So tomorrow I'm flying to Ireland to IFE conference . Perhaps we can meet there . I'm looking forward to drinking Guinness with you and then just after that I'm flying to New Zealand and that's going to be a little longer trip because that's a little longer flight .

I'm going to be at FireNZ conference and some events before that in Wellington next week and then for the next week I'm flying to Christchurch , the University of Canterbury . So if you're in Wellington or Christchurch I'm very happy to meet you up and have a beer together and talk about fire science .

The downside of my travels is that I will most likely not be able to produce episode in time for next Wednesday because I have kind of an intercontinental flight with a jet lag and a massive conference straight after that . So I'm not sure if I will be able to produce episode of quality for you for the next week .

I hope you'll find something to listen in the fire science show or in the uncovered witness fire science revelations , if you have not heard that podcast .

That one has eight episodes to listen , sufficient to keep you entertained until I'm back , I think for the last week of October when I settle down in New Zealand a bit , I will be able to process an episode from there . So I'm pretty sure there should still be one more episode coming in October I think , but we will see .

It's painful to me to drop episodes , but sometimes you really have to do it with the overwhelm of the surrounding world . Anyway , I'm very excited to travel to meet a lot of you . I'm very excited to talk about the concept of scalability of CFD , the concept of automated qualitative result analysis in CFD , and I look forward to what's gonna come out of that .

I wonder if that will make some people change their mind about the use of CFD in fire , because I truly think that , yes , the the way , how we use it today may not be the best , but there are ways in which we could use it really , really well . That's my wish for the future . Thank you for listening to the fire science show .

If there are more developments in this regard , you can be sure that I will share them with you on the Fire Science Show . So see you here in two weeks . Have a nice week , have a nice weekend and , if you are somewhere where I'm traveling , I'm looking forward to meet you , thank you .