158 - Fire Fundamentals pt. 9 - Know your boundaries (in CFD)

I was asked over and over again to do some basic physics episodes in a kind of approachable way and it kind of took off and I really enjoyed doing it . So let's continue doing fire fundamentals .

The last time I did it solo was on compartment fires and previously , before that , we talked about CFD , and the CFD episode was one that many people really enjoyed and yeah , unfortunately for me , you have asked for more . I was not really sure what's going to happen after that episode .

I was pretty sure that I'm going to get a lot of rage emails that I am not covering fluid dynamics in a way that I should , not in a scientific way . But actually people have enjoyed that . There were no complaints . There were questions for more . So here we are .

I'm going to do another episode on the fundamentals of CFD modeling and today we will be discussing boundary conditions and , as boring as it sounds , I'll try to make it exciting and interesting for you that even such a simple thing can be taught in a way that is fun and that really highlights the importance of the subject .

For the goodness of your models , it's critical to know your boundary conditions , so let's spin the intro and learn our boundaries . Welcome to the Firesize Show . My name is Wojciech Wigrzyński 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 . I wonder if you have this feeling that you wake up and you realize that what you have planned is perhaps a really bad idea and you should not be pursuing that . I go through that every second podcast episode , but I'm very bad at listening to my voice of consciousness , or perhaps my imposter .

And here I am talking on podcasts about boundary conditions . You could claim perhaps one of most boring topics one could pick , but I really see some funny and interesting things related to boundary conditions , on a very fundamental level of their use actually , which can make or break your simulation .

And actually every now and then I see some people posting on LinkedIn their simulations and you immediately realize that their boundaries are not really the ones they should use or they're perhaps misused in a way . So let's try and talk about boundary conditions in fluid dynamics , in simulations .

And note , before we start , you have to understand that I am ANSYS fluent person . I am an ANSYS fluent user . I'm trained and , let's say , skilled at FDS , but it's not my daily software that I would be using for everything . So if I confuse the name of a boundary condition in FDS , please feel free to correct me and I'll try to fix it .

I'm learning every day . Hopefully I'll get my boundaries done in Ansys and correct If I mess that one . That's going to be hilarious . Anyway , if we talk about boundaries , what are boundary conditions ? Why do we need them in our simulations ? If you would like to attempt a fluid , dynamic simulation of an endless space , let's think planet Earth .

You swim to the east , you eventually reach the point from where you've started An endless continuity . If you look down , well , there's a boundary condition . There's a solid surface or an ocean . If you look up , it ends in with a vacuum of space , also something you could call a boundary condition .

So even in an endless space of Earth , you are still constrained by something . And obviously , when we simulate our fire cases , it's not that we're going to simulate an endless atmosphere around our buildings . I mean time is money and resources are everything . They dictate what we can and what we cannot do .

In our simulations we have to narrow the field of what we're investigating into a part of the space and , as you remember from the previous CFD episode of the Far Fundamentals , cfd works basically on solving multiple equations momentum transfer , heat transfer , species conservation and so on , so on in a confined space .

And for it to work , it must understand what's happening when it reaches the end of that space . Does it magically disappear ? Does it bounce back from a solid obstacle ? These are your boundaries . These are the elements that constrain your fluid , that constrain your model and through which you can actually act and define your model .

These are critical parts of your model where you have the ability to interact with the fluid that you are modeling . Actually , if you think about fluid itself , you can define fluid as a substance that continuously deforms under the application of sheer stress and adapts its shape to the constraints . So here you are .

That's your boundary conditions , things that act on your fluid . Now , if we want to narrow the discussion into particular types of boundary conditions I'm not going to list them , that's pointless . I would say I would define two major categories of the boundaries .

One would be the boundaries that actually interact with your fluid , so I would call them the flow boundaries or the fluid boundaries . It's a very unprofessional way to call them , but it kind of delivers the message .

It's the boundaries that interact with your fluid and act on it , and boundaries that are pretty much solid walls , like solid boundaries of your model . And there are some special boundaries which we'll talk in the end . So stay tuned for a surprise hopefully interesting For fluid boundaries .

Into this big bucket I put all the boundaries that you use as a CFD user to interact with the fluid you have in your model . That most likely is air or some mixture of air and smoke .

I often call the smoke spoiled air because in CFD modeling the smoke properties are actually quite same as the air has , and the only difference being that there are species in it , like soot and other products of combustion , and due to that it has different emissivity .

However , for most cases the smoke and air are pretty much the same thing from the fluid dynamics perspective . Anyway , from my perspective , the first , most basic boundary condition I would always have in my model . It's hard for me to figure out a single model in which I would never use that boundary condition .

These are conditions that simulate an open end to your numerical domain . In ANSYS we call them pressure boundaries , pressure inlets , pressure outlets . We call them pressure boundaries , pressure inlets , pressure outlets . In FTS I think we commonly refer to them open , and those are types of vents that are used in FTS .

Regardless of the name , the point of this open or pressure boundary is to define space through which the fluid can flow freely and there's a defined pressure dynamic pressure on that boundary condition . Now , what does it mean ?

If the dynamic pressure on that boundary is positive , then the flow is coming from that boundary into your model , unless the pressure in your model is higher than that one . If the pressure is negative , there's going to be a suction effect . If it's zero , then the flow will be an outcome of the pressure value inside your domain . Pretty simple , isn't it ?

But there are some issues that you can easily incorporate into your model by simply applying these boundary conditions without thinking . If you think about pressure in your domain , well , it's not that the pressure in every point of your domain is exactly same . There's hydrostatic pressure . That is simply an outcome of your fluid having some weight .

It just weights and you'll have a hydrostatic pressure buildup . In normal conditions that would be roughly 12 pascals per meter of height . So that's quite a significant pressure . Now imagine , on your other side of your boundary condition you always have zero and on the domain in which you are simulating your fluid flow , the pressure is growing .

With every meter down , it simply increases , which means that you could end up with artificial pressure difference between your domain and exterior just because you've defined the pressure on the other side as zero . For the CFD software , along the height of your pressure boundary , the pressure is zero . It's not hydrostatic .

Perhaps in other softwares than ANSYS it's predefined as hydrostatic , being included automatically . For us , there's actually a necessity to consider that and include that in our pressure boundaries . So here , even the most simple condition has a challenge for you .

Now another thing that's interesting in pressure boundaries or open boundaries , sometimes we would be only interested in modeling the interior of our building . Let's say you're modeling a warehouse and you want to put some sort of inlets , outlets , to that warehouse .

You're working with a natural smoke control , so you're obviously using things like natural smoke extraction ventilators . You're using doors as your makeup air sources . We've talked about that in previous Fire Fundamentals episodes . Now you would like to model them , and there is an urge that the simplest way to model them was a natural opening .

That's like literally something that connects your interior to the exterior , right ? So perhaps a pressure boundary is a perfect way to simulate

that . Now the thing is that when you have natural ventilators , especially smaller or ones that do not open fully , there are different types of natural ventilators that are less efficient In those devices . It's not that the air flows freely through the entire cross-section of the device . It may be blocking the pathway for the air .

It will definitely include some resistance for air to go .

So if you have , let's say , a square meter of a natural ventilator on your roof and in that space you put a square meter of a natural ventilator on your roof and in that space you put a square meter of a pressure boundary , the effects of pressure boundary will be more pronounced , more significant than the effects of your natural ventilator , because the natural

ventilator is less efficient than an idealized hole . Another thing outside you would have different temperatures . So you also have to define this temperature on the other side of your open boundary to really capture the chimney effects and all the buoyancy effects that will come in play .

So just dropping your open boundaries , because they are not openings , they're pressure boundaries . You have to understand how you use them and be conscious in where you place them . If you want to simulate natural ventilators , they're not the way to go .

You need to simulate some sort of a roof structure and some sort of openings in that roof that at least try to mimic the effectiveness of natural ventilators . Perhaps you can go away in the most simplistic case by using smaller , but I'm not going to give you a recommendation of how small you should make them .

If you want to play like that , you need to do your own validation and figure out values that would work for yourself . Now we've briefly talked about the pressure boundaries . Another boundary that I would very , very commonly use in my models would be velocity boundaries . So here we are talking about mechanical vents . In ANSYS we would call them velocity inlets .

In this case , instead of a pressure on my boundary , which again is something that fluid can flow through , I define velocity of the flow and I define a vector which way the flow goes . So I am capable of defining a fixed suction or inlet point in my model through which air will come or be extracted through with a specified velocity .

I also define stuff like temperatures . I also define in Ansys stuff like turbulent intensity and other important parameters for my simulation . But the most important thing , what's the velocity at the boundary condition ? Now , again , we're talking about boundary condition . You have a patch of your model that this boundary condition is applied to .

When I say that the boundary condition of velocity is three meters per second , every single cell , that's a part of this patch that I called my velocity inlet now will have three meters of velocity introduced into my model , exactly this value , and the model will make sure that this value is the value introduced to my model . Now there's an issue with that .

It's pretty robust and simple and in most cases it works without playing too much with it . But if you would like to be really , really precise in modeling your vents , it's not that the entire ventilator has the same velocity at its surface , whatever the surface is . It's actually a virtual concept the surface of a ventilator .

Anyway , through the cross-section from which the air enters your domain , in a real case , in a real building , you will have some sort of velocity profile , whereas in your CFD model you will have idealized averaged flow over the entirety of the velocity patch that you put into your model . How big is the issue ?

So if you have a very narrow duct work , you have we call them pancake ducts very flat ducts in which one dimension is pretty long and the other is pretty narrow and you have a very big vent on such a pancake duct . You usually end up with extremely uneven flow through that duct .

So you could have 2 m per second average on such an opening , which could sound acceptable in some sort of cases for smoke control , whereas in reality you would have no flow on most of the opening and 5 m at the bottom .

Now those 5 m will cause you trouble with your smoke control system because that's a lot of kinetic energy introduced into your building and a good chance that this flow will mix something in a place that you don't want it to be mixed .

And especially if you're modeling stuff like corridors or some narrow , tight spaces , this error can occur , lead to a system which works perfectly in simulations by applying this averaged velocity over your inlet , whereas in reality the velocity is skewed and you get much higher velocities and the system actually does not work .

We've seen that happen and that's a pretty big challenge . The ways to overcome that we usually try to model the ductwork and so the the profile on the inlet is actually some sort of an outcome of the calculations of the model , and my velocity inlet is actually where my extraction or inlet fans are physically .

So instead of simulating just grills that are in my compartment , I would would simulate the ductwork For the more challenging projects . I wouldn't say that we did it every single time , but when we need to pay attention . We do it like that . Another way you can define profiles . This is a very interesting way to work with those boundary conditions .

So you can actually force CFD software to apply a specific profile to the velocity introduced so it's no longer even but actually conforms to however you define it . So yeah that , that that's the way .

Another thing interesting with velocity boundaries is that when you think you introduce flow from a mechanical ventilator into your compartment , you would most likely think about your horizontal velocity component from that fan . Right , because it's just blowing air straight in front right .

But in fact , because the fan is a rotating machine , it actually introduces quite a significant tangential component to velocity . It swirls and in some cases , in some aspects of modeling , this swirl may be actually quite important to the outcomes of your simulation . Here . I wouldn't say it's something we would use commonly . I would say it's very rarely used .

But especially in tunneling projects when you're modeling big jet fans , you're not usually doing that with velocity inlets , but other flow conditions also can apply as well . This could actually make a big difference in the outcomes of your modeling of the flow profile of your jet fan .

So yeah , sometimes you have to go in your understanding beyond just the basic characteristic that you're pushing air forward . I know some cases . There's a gentleman , fahd Itrada , who is doing amazing modeling of chat fans .

They've developed a fan called Mojet and truly the modeling they do to model their fans is remarkable and the level of details they go into to actually capture the behavior of that chat fan is stunning . So , as with most of the things you do in fire safety science and fire safety engineering , when you go deep inside it becomes pretty complicated .

Another aspect of velocity boundary conditions is when we try to apply wind . That's a big trap . A lot of people fall into that . So wind is not just a boundary condition that blows air at a constant speed . Wind is a complicated phenomenon , physical phenomenon , and you really have to introduce wind first with a profile . So it has to follow some profile .

The wind at the ground level will be weaker than the wind at some height . There are logarithmic profiles that define the velocity of wind . With the height , wind will introduce turbulence , so you have to capture the turbulence .

If you're modeling it with large eddy simulation , you may be even in need to get some sort of periodic boundary condition that increases , decreases velocity continuously so you capture the formation of large eddies or large vortices that come with winds . So you get the spikes in velocity correct .

It's actually quite challenging , but just dropping a massive velocity inlet boundary condition in front of your model , it it's not really cutting . That that's not wind , that's a big chunk of velocity , not a physical phenomenon of wind , which is beautiful and complex , and I have many other podcast episodes that go deeper into that .

There's another thing about velocity . I mean velocity inlets . This is very simple but it could be a very problematic boundary condition or it can have a very far-reaching consequences for your models .

So if you define the velocity inlet boundary condition , you define the speed at which the air is put into your model or extracted from your model , which basically defines the exact value of volumetric flow that's flowing through that space , and this will be constrained by the model . Now imagine you have your model . It's a cube , it's a compartment .

You have one velocity inlet that brings air into the room . You have velocity outlet that extract the air from the room .

You've defined both velocities to match each other and now you introduce the fire into the room and what will happen is that you will observe a pressure increase into the room , perhaps even up to some unphysical levels of pressure and your CFD model will crash . It will give you very odd results . Why ?

Because the system is not able to blow more air through the velocity inlet , which has a defined velocity . It has to extract or input exact the number you've put into it . If there's a pressure increase around that boundary condition , it's not changing the outcomes , so the pressure increases , decreases .

You can get into big trouble in your models if you only define very fixed values on those types of boundary conditions , and that's why I said I use pressure boundaries in almost every single model . You almost always need a pressure boundary somewhere to improve the convergence of your simulation , to have a way to manage the decrease , increase of pressures .

Of course this has to represent some physical leakage in your building . If you're designing a , the most sealed building in the planet , you perhaps should do velocity boundaries , but then again the fans don't work like that and they would adjust . Oh yeah , let's go to fans .

Actually , that's accidentally a very good segue to the next boundary condition I wanted to talk about which , which is fan boundary condition . And this is a boundary condition in which we have a slice of space through which the air can flow through . So it's a boundary condition that does not mark the outer regions of our simulation .

It's placed inside the fluid and it accelerates the fluid . It acts as a fan , as a ventilator would . Now there are so many ways you can design this boundary condition .

Actually , my friends from Silesian University of Technology , alexander and Gosia Kurl , have wrote a really nice paper on all the different ways you can use to model fan boundary condition to capture the behavior of a fan , especially jet fans , in your simulation better . And yeah , I really like to use this , this boundary condition in ansys .

It gives me very good , uh , first , it gives me very good control over the flow of my fan and secondly , because it's based on decades of answers being used to model fans for fan manufacturers . They actually got very convenient and good and accurate models of fans built within the software .

So for me it's just convenient and if you think about it , it's kind of funny because it's a very simple answer to a problem that have been quite complicated in the past

. So if I remember correctly , the first times we tried to do jet fans we were mostly using velocity inlets and you know , having air sucked into the velocity inlet on one end of the jet fan , having it released on the other end . The issue was it was a different air because it kind of disappeared when it reached the other end of the boundary condition .

It was not modeled anymore . Then we went into recirculation inlets , outlets , which were velocity inlets connected with , let's say , a magical duct that connects one end to another and that allowed us to transfer air from one side to another .

But this did not allow us to model the behavior of the ventilator where it receives different pressures on the inlet or the losses change on the outlet side . The fan , of course , is a mechanical device . It has its power curve , it has its fan curve . Those are the things you can actually model with a good fan model and by introducing that .

It takes a lot of work actually to introduce that correctly , but when you introduce that you are actually modeling a real ventilation device in your system and I think in some projects it's absolutely necessary , just like capturing the profiles on your velocity inlets .

This is sometimes also critical to get your smoke control system modeled in a way that you can be fairly sure that what you've done is accurate and in the building , when they finally build the building and they finally do the smoke control system in it , it's going to work like you've simulated .

It's a very stressful situation when reality validates or verifies your numerical modeling and you really want to be on the safe side . And you get on the safe side by knowing your boundaries , knowing your boundary conditions and applying them correctly .

An interesting thing for modeling fans and , in general , smoke control systems or ventilation systems is the hvc model in fds . Massive shout out to jason floyd who was spearheading the efforts to write this model .

I think it's a brilliant addition to FDS and it's simply a nodal model where you can basically build a network of nodes that interconnect to each other and simulate the ductworks , but instead of physically simulating the ducts , you know the shape and the flows inside .

There are physical relations that describe the flow of air within the ducts , through ventilation devices into your rooms and so on , and this network is modeled along your CFD model and the boundary conditions are defined by that network .

So it takes a lot of hard work from you onto the network and it actually calculates what are the flows , what are the pressure points on your fans and so on . A brilliant way to actually make the lives of fire engineers simpler . So highly appreciate that this was included in FDS .

I actually wish we had something like that in Ansys , like if we wanted something like this , we would have to program it ourselves . And uh , yeah , sometimes we we do some attempts not as fancy as the solver in fds . I'm highly envious in this regard . Perhaps we need to recruit jason to write one for us .

The hvc nodes are also something you can use to model your jet fan . So , as I said previously , we have the fan boundary condition in answers , which allows us to define the performance of the fan based on what comes in , what comes out , what's the losses and so on .

You could actually do the same thing with HVAC nodes in FDS to efficiently model your jet fans . I'm not sure if you can add the tangential swill to your flow in FDS . Probably you can , but please don't quote me on that . You have to submit an issue on issue tracker and figure out if you can or not . If someone knows , let me know .

I'm actually curious now because I have not fact-checked it before . I've never done it in this detail in FDS . Anyway , this would be a quick summary of the important flow boundaries . No wait , there's one more . There's one more very important flow boundary , that's mass flow inlets . So again , very similar to velocity inlets , very similar to your mechanical vents .

But in this case , instead of prescribing a specific velocity or volumetric flow rate , you would specify the mass flow rate . Now you could think it's the same thing and yes , that's true if you have your flow in your ambient air .

So if you're talking about delivering air to your model at the same temperature over the course of your simulation , yes , that actually is true that it's the same boundary condition . But if you define extraction in a form of mass flow , then you would get completely different outcomes as the fire grows .

The reason for that is that as fire grows , the air heats around , and as the air is heated it changes its density . So if you say you're always removing five kilograms of air from your model , in a cold situation , that would be like four something ish cubic meters . But if the air is at 600 degrees , that would be twice the number .

So you have to be careful with those boundary conditions . Actually , we in the past we found a very clever way to use mass flow boundary conditions to define a completely new way of extracting smoke from buildings , extremely efficient . We coined it as a smart smoke control . There are papers on that .

If you wish to dig in deeper , perhaps one day I'll do an episode on that . That was my pet peeve in 2017 . My biggest discovery you really deserve to learn that . Perhaps that's a topic for episode 200 . We'll see .

Anyway , there are clever uses for mass flow inlets and also one common use of mass flow inlet boundary conditions is related to your burners , to the fire source . So we would very often like to define the inlet of fuel to our model with a mass flow inlet . Why ? Because it translates very easily to the heat release rate on paper .

You just multiply by calorific value and efficiency of combustion and you're there , you got your heat release rate . So it's very easy to control it in that way . I think when you do burners in FTS and you define the amount of heat release , it actually calculates it down to the mass and then it's a mass flow inlet boundary't . Quote me on that .

I'm not 100 sure , but the outcome is like that you you bringa specific value amount of mass into your model in every second of your simulation . Are there any challenges with that ? Yeah , of course there are . With everything there's challenges .

In fire safety engineering we've actually found one very interesting challenge If you're having very small burners and very small heat release rates , and that's the case when you have modeling of scaled down experiments in fire science . So we have this technique called the Froude number scaling .

Previous episode of fire science show was very rough and here's another pebble to the pile saying that it's not the best technique to study fire science .

So if you have a very small mass flow rates and the fuel is released from the surface with very low velocity , the efficiency of combustion actually changes , and my student , jakub , is working very hard on a paper that summarizes our findings in that regard and once this sees the daylight , I will be first to to tell you about it , because I'm very proud of this

work and , yeah it , it highly relates to the mass flow inlet boundary condition . So , yeah , that was uh , my good summary of uh , boundary related to flow . But there's a second group boundary conditions that I would call the solid boundary conditions , or the physical constraints of your model , basically , the walls in your buildings , the walls in your model .

So , even though you could consider it as one boundary condition , a wall , there are a lot of things that go into it and I think we can simplify them into things that are related to heat transfer and things that are related to sheer stress and forces and resistance . So let's try heat transfer first . How does wall interact with your fluid ?

First of all , the fluid is constrained to your walls . It takes the shape of whatever container is put in . So , yeah , walls define the shape of your fluid and , of course , the fluid cannot simply penetrate your walls unless they're porous . But we'll reach that as it touches the fluid . And we're talking fire science .

So obviously the fluid is not ambient air , it's some outcome of the fire physics happening inside your model . There are heat transfer phenomena that are happening and actually there are multiple ways how a wall boundary condition can handle the heat transfer and can interact with the fluid . So there are these different types of boundary conditions .

Formerly they're called the kinds the first , second , third kind , drichlet , neumann , robin , I think . But that's a very challenging way to explain it to people who may not be fluid mechanics For me . Think about the modes of heat transfer that we have .

We have radiation , conduction , convection , and that's basically what we try to represent with the model of our boundaries . So in the simplest way you can just say that the boundary does not participate in heat transfer , it's adiabatic , it simply does not participate in the heat exchange and whatever heat is released into the fluid , it's not going anywhere .

So that's the simplest way . When you omit the process and in some simulations I know this is practiced I don't see huge benefit in that , but I know in some models , some cases in which adiabatic walls were chosen to make the numerical case more simple , another way is to simply define the temperature of your wall . So yeah , that's a way .

That's a way structural engineers would be defining boundary conditions for their structural models . They would tell what temperature the wall has and that's it . You can also define the heat flux on the wall , convective heat flux , so how the wall takes the heat from the air .

Or you can actually make a combination of all of those boundary conditions and make a mixed boundary condition where the wall takes the heat flux from the fluid but its temperature is defined by a heat transfer phenomena inside the wall , and that's the correct way . That's the proper way to model them in fire science . In my opinion .

There are some challenges into that . Of course there are . First is with the convective heat transfer . So to assess the convective heat transfer , you need to know the convective heat transfer coefficient , and that's not a simple number to know . Eurocode tells you 35 or 8 , depending on where the wall is .

That's watts per kelvin per square meter , if I'm not wrong . That's watts per Kelvin per square meter , if I'm not wrong . But there are also mathematical models that let you figure out the value of that coefficient , based on Nusselt number or some other simplifications of flow phenomena in boundary layer . Anyway , for simple fire modeling we have a value .

That's a pretty good guess If you want to be a real scientist . I know people who are making PhDs of calculating that value and it's pretty fun to actually go deep into what goes into convective heat transfer coefficient . We had to go that way when we were modeling linear heat detectors .

Another paper that's currently being written in my laboratory by my Jakub , who seems to write too many papers at the same time and doesn't have time to finish any of them . I will chase him , then you'll have a nice source to learn about convective heat transfer for thin elements . Anyway , there's also heat transfer inside into the solid .

So the heat goes somewhere . It doesn't magically disappear . So you could imagine that in your fluid dynamics model you also have a solid model . So you're actually modeling the solid with discrete cells and you're solving the heat transfer phenomena that go into that solid wall . But that would be very computationally heavy and perhaps not necessary for fire science .

In this case we have simplified models for heat transfer , very simple models based on Fourier's law . It's very well known . The heat transfer through solids is actually quite well known . We have one-dimensional models in which you just simulate the heat penetrating the depth of your material .

We have a simplified three-dimensional models where the heat actually transfers through a virtual assembly , let's say , and you can track it . It's especially useful for structural engineering to have this type of an assessment and with this we are capable to say where the heat went . And it's critical .

Solving heat transfer from your fluid to your structure is critical for two reasons . First , for your smoke to cool down . Like you have to capture how your smoke is cooling down through heat transfer because this influences the buoyancy of the smoke and buoyancy influences the smoke behavior , the flow velocities , the ceiling jet and everything .

It's critical to capture that and it's critical . The second reason it's critical if you're studying structure . So if you want to solve a structural fire engineering case , you need to understand how much heat has went to your structure and then assess what that heat does to your structure . Sometimes you would get with extremely complicated models .

Another proud paper that I participated in with Michał Malendowski . We wrote a paper that was Michał's model . I cannot take much credit for that . I delivered the experimental validation of his crazy ideas .

The concept was that if you have a very complicated shape , let's say an I-beam , you cannot really model that within your fluid because you cannot afford putting one , two , three centimeter cells to solve the fluid around your very complicated structure of an I-beam .

However , miho came up with the concept that you could build a virtual box around an I-beam , simply solve the boundaries of the virtual box .

So now , instead of a very complicated shape of a beam , you have basically a rectangular shape of the virtual box and then with another set of equations you can calculate all the modes of heat transfer within that box to that exact shape of an I-beam .

And that's like seven pages of very hardcore mathematics that give you a very quick and robust solution to heat transfer problem from fluid into a complicated thin wall structure . So yeah , I'll link the paper into show notes if you want to see it for yourself . It's quite a sight . I'm still waiting for someone to implement that model in a CFD tool .

But when one does , I think it's going to be implement that model in a CFD tool . But when one does , I think it's going to be very , very useful .

I said thin walls . So another aspect of wall boundaries related to heat transfer , how thin or thick they are . So if you have a very thick wall , it will behave differently than if you have a very thin wall . We sometimes even call them thermally thick or thermally thin , which doesn't really define their width or depth or whatever dimension is called .

It defines how quickly the heat transferred through them and the point is whether you can assess that the entirety of the wall is at one temperature .

So if you have a thin sheet of metal think trapezoidal steel sheet that would be a very thin boundary because you can say that the exterior surfaces would have a very similar temperature or the same temperature as your internal surface . There's no gradient inside that wall .

Or you can have a thermally thick wall , let's say a thick concrete wall , where you would have a spectrum , a gradient of temperature going through that wall . And also because how long the fire phenomena are that we model , it's very unlikely that in your short CFD simulation , the heat would actually penetrate to the other side of the boundary .

Now it's important in some cases . There were cases , for example , in the Iris project . I know they came up with some new observations regarding flashover in compartments with thermally thin walls . If I'm not wrong that was a paper authored by Mohamed Bashir and if I'm not wrong he's part of AFAR now , so good career choices there , not wrong .

He's a part of our farm now , so good career choices there . Anyway , they've observed that if you had thermally thin walls , which kind of represents an informal settlement that's what the iris project was studying .

In that setting , the the flashover cue is in a slightly different manner that you would predict with , let's say , mqh correlations or or some classical compartment fire science . Very , very interesting science . Perhaps another thing to bring to your attention in the podcast I'm noting down the ideas they come up so quickly .

So I hope I've summarized the thermal aspects of the solid boundaries . But there's one more aspect to the solid boundary and that's their roughness . So the solid boundaries not only interact with the flow by exchanging heat with the fluid , but they also interact with the fluid by sheer stress .

So the fluid flows against the wall and there's some momentum transfer between the wall and the fluid . It's because the walls are not perfectly slip . They cause resistance to the flow of air and this actually can be quite impactful . It's impactful in , I think , two situations for me .

First situation is where I'm modeling ductwork , so it's critical to capture the pressure losses in my ductwork . I need to get this shear stress correct Because if I don't , my flows in the duct work will be completely off , my pressure point of my fan will be completely off , my CFD will be bullshit . So I don't want my CFD to be bullshit .

I want my CFD to capture reality as close as possible . So that's one case where we really need to control that , that slip condition . And the other one is when we're doing wind engineering . So in wind engineering we would introduce a very , very large numerical domains , so the wind profile is not affected by the building that is inside of that domain .

It's something called the blockade effect that we want to avoid . So our domain must be significantly larger than the building that's subject to the wind or neighborhood .

If we're modeling a neighborhood , then it's kilometers in size , and when I have kilometers of space between my inlet point on which I meticulously put my wind profile with a profile function logarithmic wind profile .

Usually I want that wind profile to reach my building , which I'm modeling , but when the flow is flowing through my domain it interacts with the ground because the ground has roughness in it and this roughness actually changes the wind profile .

So if the wind flows for one kilometer until it reaches the building of my interest , I can have a completely different wind profile than the one that I've defined . There are ways to handle that . We have in-house modifications of ANSYS code which my postdoc , paulina , has published .

We've took some clever models from the literature and applied them into our ANSYS fluent simulations and this allows us to precisely control what's happening between the air and the ground . How does the roughness of the ground , the surface , the affects the flow ?

And actually it's quite important because in wind engineering that's a well-known technique to uh to introduce the turbulence intensity and some shape of wind profile that you would expect in a very specific terrain case .

Actually , if you go deeper in that , what's happening between the house , between the flow and the wall , is something we often refer to as a boundary layer problem . So a boundary layer problem is exactly that interface between the fluid and the solid and it's something that we're very ignorant in fire science about .

So you don't really see good boundary layer control in cfd simulations for fires , perhaps because the velocities are very small , perhaps because it usually is not a part of interest , as I said , it's for me it's wind engineering or ducts , and this would not be the main parts of cfd analysis in fire now if I think about it .

But yeah , boundary layer problems are are a hell of a problem . They are very , very challenging to solve . They require fine meshes , they require near-wall models , they require modifications to your turbulence modeling . They are quite a fun of a problem to solve with and if you ever have problems with them you can reach me out .

Perhaps the way we've solved it with paulina for our wind studies will will help you out . Um , that would be it quickly for the wall boundaries . Okay , there's one more wall boundary I've kind of teased .

That's a porous boundary condition , so boundary condition in which some of the flow can flow through your porous , medium porous wall , and we sometimes use this boundary condition to simulate inlet grills . So you can actually control the flows on your complex duct network by those boundary conditions , and they would be very efficient in doing that .

However , it's quite difficult to get the data necessary to define those , and in my career I've also used the porous boundary condition to actually model some baffles on the roofs of buildings to break wind acting on the roof of the building . So that was an interesting case study .

I know my friends also use porous boundary conditions to simulate resistance of a tunnel network , like if you're modeling a road tunnel or something . It's very similar to the duct . In that case the pressure loss along the length of the tunnel is important and you want to capture that .

You can do it by simulating the entirety of the tunnel and the roughness , and you can also build some surrogate model in form of porous boundary condition and it actually kind of works . It solves the problem , allows you to input an artificial resistance to your tunnel that mimics the pressure losses that you would have in your tunneling system .

So we've reached the end of the solid boundaries . I've promised you special boundary conditions , so we have a magical boundary condition called the symmetry . It basically is what it says it is . It's a symmetry plane . So if you're modeling a duct , you can model half of the duct and assume that the other half is symmetrical . It's interesting boundary conditions .

A lot of people are using it in a funny way . So some people are using it as a top of their wind tunnel in their modeling , which is kind of ridiculous because that means that you have a mirror image of your wind tunnel in your simulation .

That's not really great , capturing physics of your model , but it's a boundary condition that came to life because of savings . Necessary Time is money and CFD simulations cost a lot of money .

When you have a symmetrical case , especially you're simulating something simple like a duct , then symmetric case is your best friend there in answers we have a very interesting boundary conditions called interiors , though that's basically the connections between every single cell in your model .

But I can also artificially put that condition on any other uh , let's say physical obstacle in my model which makes them disappear . In fds you have the magical boundary condition called the hole , where you put a hole in the wall and that allows the flow to go through .

In ANSYS we call that interior , so it changes the boundary condition for something that's completely permeable by the air . And there's also a variant of that boundary condition called the interface . So I can actually have two different meshes in my model and connect them with interface , so they don't match each other perfectly .

But the interface boundary condition would work out how the flow from one domain flows into the other domain , very , very convenient , especially in wind engineering when we have to rotate our models . So I have my domain changing every time I change the wind angle because I change it through the rotation of the internal domain .

Anyway , I was rushing at the end because I see the timer and I'm already talking about boundary conditions for like 15 minutes . That's crazy . But yeah , that's it , we've reached the end of my list . So we've went through the flow conditions . We've talked about the simple pressure inlets or open conditions . We've talked about velocity conditions .

We've talked about the simple pressure inlets or open conditions . We've talked about velocity conditions , the velocity inlets , and we've touched a bit the mass flow inlets . We've talked about fans , fan boundary conditions . We talked about HVC capabilities of FDS , which are magnificent .

Then we moved into the walls , the solid boundaries of your model , and we've discussed modeling the heat transfer phenomena model . And we've discussed modeling the heat transfer phenomena and we've discussed modeling shear stress on those boundaries . So I hope I took you on an interesting adventure through boundary conditions in numerical modeling .

I've tried to give you as much real life examples as I could that are relevant to use the good , correct use of those boundary conditions .

And what I hope to achieve with this episode is that the next time you're building your model , you perhaps will reflect on what boundary conditions are you using , what physics they represent , what role do they play in your model and if everything is perfectly fine with them . I think such a reflection is necessary for every CFD model .

If there's an obvious error and you see it and you fix it , that's not a big problem for CFD . If there's an error that you don't see and don't recognize , it is there and you get the faulty results which you are not aware that they are faulty , that's an issue and to avoid that we need to know our boundaries .

Thank you very much for being here with me . If you stayed till the end of this very difficult episode . I thank you twice Great job , thank you for staying here with me and I know you're a big fan of the Fire Science Show and I am sure I will see you here next Wednesday with the next episode of the Fire Science Show .

And once again , congratulations , rory , for being the first episode ever in the fire science show to break . 2 000 listens , insane 2 000 people . That's , that's crazy . Thank you very much . See you bye . Thank you .