Matt Price:
Hello and welcome to this episode of ASHRAE Journal Podcast. I'm your host today, Matt Price. Thank you for joining us. I'm joined by Nathan Kegel. Nathan and I have known each other for several years. As CTTC chair in East Tennessee, Nathan was one of the first speakers I brought on for chapter meetings and after several years of doing it, he's still in the very, very top tier of who we've brought in to our chapter and one of the best speakers I've seen. So today's topic will be load calculations in commercial HVAC.
Nathan, you're sort of one of those rare people who've worked in most of the visible roles in our industry. Can you tell our listeners a little about yourself?
Nathan Kegel:
Yeah, sure. Thanks, Matt. Good to be here. And I spent over 25 years in this industry in a variety of roles. My first jobs were working doing load calculations for projects all over the country, including Canada, so not just in the country, I suppose, but all over the continent. So I saw a wide variety of design conditions, a wide variety of projects, and had a lot of time starting to review those loads. I was trained like many of us are at that age, new in the industry by more senior engineers. And through the course of learning more about load calculations, the methods, et cetera, really started to question some of the things that I learned. And that kind of led me to my next role, which has been to support simulation, building performance, and a variety of roles across architecture, engineering, contractor spaces across this continent as well. So happy to be here.
Matt Price:
Yeah, absolutely. And I think all of us have a path to HVAC that's different, and I think that's pretty cool. So also my path is not completely direct. So I started as a process engineer at an automotive manufacturer after college and did not love that. And then I wanted to basically move back to where I grew up and I started going to ASHRAE meetings. And then in the second meeting, I had a job and was buying a house in the town that I wanted to come to. So basically, if you want to live somewhere, ASHRAE is an excellent way of getting to it. So Nathan and I are both very active in ASHRAE.
So my HVAC background is as a manufacturer's rep in HVAC equipment sales. It was a fantastic job. And now I run and own an education company for new professionals in the construction industry. So interns, estimators, salespeople, designers, EITs, et cetera. So that's what I do now.
So in other words, what I do not do now is deal with load calculations all day every day, and you do. So that is why I'm the host and you are the specialist here, Nathan. So for an introduction, we're going to be talking about load calculations today. So I think the best way to get us there is to sort of explain the why first. As the ASHRAE Journal Podcast has such a broad audience, I think we need to consider new people to experienced people. So let's start with why. Why are load calculations needed?
Nathan Kegel:
Yeah, it's a great question. Load calculations are probably the foundation of nearly every HVAC system design. So load calculations by extension are also a very significant source of risk in the design process. In short, you really can't figure out how much cooling you need or the type of HVAC system potentially until you understand what the load is and how the load breaks down. So that's different for cooling versus heating. It's different for different climates. It's different for different types of buildings. There's all sorts of factors that go into it. And load calculations, while it's easy to sort of describe their importance, are actually really complicated and how heat moves and is transferred into building is very complex. The foundation for all of this, and I'm sure we'll get into this more, is really chapter 18 of the ASHRAE Handbook of Fundamentals. And there's a lot of other things that we'll talk about in that space as well.
Matt Price:
So can you paint a picture—we have reps listening who don't do load calculations, but are very involved with sort of the outputs of load calculations. So in the design process, where does this occur for the engineer, for the consultant? When does this occur in that process?
Nathan Kegel:
For the consultant, ideally we'd start doing a preliminary load calculation as soon as we know three things about the building. We would want to know where it is so we can get an assessment of what the design conditions are going to be for heating and cooling. We do want to know what type of building it is. Is it going to be an office? Is it going to be a hospital? So we can get an idea of things like how much ventilation do we need. And the location will also tell us things about reasonable assumptions we should be able to make for what the envelope is going to look like. So what's the R value of the wall, for example, what's the solar heat gain coefficient? And all of those things are going to be assumptions very early on, or at least educated guesses in a way.
As the design progresses, hopefully, and ideally some of those assumptions that you're making based on code or based on just experience are going to come more into focus. And I say ideally because realistically it doesn't always happen that way. And also, even in the case where you do get a lot of those blanks sort of filled in, there's always going to be unknowns. You're never going to know what the actual infiltration rate is until after the building is constructed. You can make some educated assumptions. You can refer to standards like ASHRAE 90.1 or IECC, the International Energy Conservation Code, to get an idea of what you should expect or realistically expect. But the engineer doesn't build the building. The architect doesn't build the building. So we will always have those sorts of unknowns. But there are things that the engineer can control. For example, the engineer can and does have control over which method they choose to select for doing load calculations.
They have the ability to change some of the inputs. The handbook in chapter 18 rightly points out that the inputs are the bigger delta in terms of the result than the method itself. And that's, I think, a very valid point. And I'd also point out that the handbook also states that the heat balance method is the most accurate method. So there's a lot of things to weigh in the process of setting up the load calculation, but also determining which method to use and what's the most appropriate one for the project.
Matt Price:
So you spoke about this a little bit, but can you tell us what is a load calculation? So probably starting with what is a load? So we in HVAC don't talk about conduction, convection, radiation. There are certain situations where you do, but it's actually much more common to hear envelope load or internal load or ventilation load. So what are those? And could you give a couple of examples of each one?
Nathan Kegel:
So technically speaking, there's a difference between gain and load, and this is also described in the fundamentals. So the ventilation can be a load, that's typically a gain. The envelope is typically a gain, not a load. So gain is what you get from the sun shines through your window, falls on your floor, warms the floor, convection picks up the heat from the floor or an internal wall wherever it's striking, whatever surface it's striking. And then that eventually becomes load in the space that has to be removed by a cooling device. So load is what, and very simple, probably overly simplistic—
Matt Price:
I need overly simplistic, Nathan.
Nathan Kegel:
Gain is what you have to deal with and load is how you deal with it, if you want to think about it in those terms.
Matt Price:
Overall, could you name a couple of envelope loads or internal loads and then explain how the ventilation loads integrate into the overall load calculation?
Nathan Kegel:
Yeah. I'll reference ASHRAE Standard 183 here, which is sort of a condensed reading of chapter 18 in the Handbook of Fundamentals. And just for reference, Standard 183 is eight pages. So it's very condensed. And it does a good job though of laying out what the minimum requirements are for room and zone load and not, and it explicitly states, it is not for system sizing. So I'll make those distinctions here as well. The room and zone gains or loads that you would deal with would typically be conduction. You have a wall or a window and there is a temperature differential across that surface and heat is moving because of the temperature differential.
You have the radiant component of the sun, which is warming up the surfaces themselves. And then as mentioned, you have convection, which is picking up air moving across that warm surface, picking up the heat and bringing it in, transferring it from the surface and into the air. So those would be some of the envelope based loads. Then you also have the internal gains, which you have your lights, which give off heat. You have people, which give off both sensible and latent heat. So we haven't talked about sensible and latent heat here, but those are important distinctions as well. And then you may have miscellaneous gains, like you may have televisions or computers or some other refrigerated cases or any number of devices that are either giving off heat or absorbing heat themselves that you have to account for in the process of doing a load calculation.
And typically your load calculation is done at sort of a moment in time. For heating, that's true. For cooling, we move to looking at the totality of a single design day. And in either case, you're looking at only one or two hours of the year, and the rest of the year the building is probably nowhere near that peak condition. So there's a lot of other things that we know are needing to be studied, but don't necessarily consider when we're doing load calculations.
Matt Price:
Awesome. So I think that takes us pretty directly to outputs. So no one's doing this for the fun of it, or most people aren't doing it for the fun of it. So if an engineer or an EIT or designer, they sit down to do this calculation and we'll get to methods, and this is one of the times that we do talk quite a bit about method and design. When they're sitting down to do it, what are their outputs? What are they actually sitting down to do to get out of it?
Nathan Kegel:
Yeah. So it depends again on your—if you're looking at room and zone load, you're typically looking to figure out what this room or zone level device needs in terms of capacity or airflow. So if we think about, I'll just pick a VAV system, a multi-zone VAV system, for example. And if we're jumping ahead to what a multi-zone VAV system is, well, my apologies.
Matt Price:
We're not doing that today, so we got to scope it. We have a time requirement, Nathan. So note, we're not going to get into that. But yes, take it from there.
Nathan Kegel:
Yeah. In the room and zone condition, the VAV system, you're trying to figure out, given a certain supply air temperature, typically 55 degrees in a VAV system, what quantity are CFM in American units, liters per second in other units, how much air do I need to maintain a given set point? Typically, let's say 75 degrees, and then I have a 20 degree delta T or temperature differential between the space temperature and the supply air temperature, so I can determine how much air I need. And that's typically what you're doing at room and zone load levels.
At system, you're trying to figure out things like how much coil do I need? Do I need a six row coil, an eight row coil, a four row coil? What are the conditions of that? What's my sensible heat ratio? How much latent heat do I expect or do I need the coil to extract from the airstream? And then all those coils would then be connected to some other central device potentially, or they may be connected to their individual contained box in the case of air cooled DX. But you're also effectively trying to size that plant, if I can use that term a little bit liberally there, but to figure out all of those capacities need to be, and then that by extension has ramifications for the electrical engineer, how much power draw do they have to have? How big do the duct work does the duct work? How much insulation on the duct work do I need? All of those have cascading effects from figuring out or determining what the peak size of the equipment or the maximum capacity of the equipment needs to be to satisfy space temperature based on load.
Matt Price:
Yeah. And that takes us directly to inputs because you are showing that the output would be directly applicable to the building design and construction, which means budget, risk, complexity, things like that. So in the inputs, what would a good application look like? So when the designer's putting in inputs, they're going through and setting up what would eventually be heat balance method or whatever the calculation ran is, what are they inputting into the software? And the second part of that, what is the software grabbing from databases when it's running that calculation?
Nathan Kegel:
Yeah. There's a wide variety of software platforms available and there's a wide variety of load methods that are also still available for use. But fundamentally, the load methodology that you, I'll start with the load methodology and then I'll work back into the inputs because the inputs are probably largely similar across all of these. There's some fine-tuned details between some of these different methods, but I'll start with the methods first.
So as I mentioned earlier, the heat balance method is described in the Handbook of Fundamentals, chapter 18, and it is stated to be the most accurate method as anyone who's seen my chapter presentation knows there's ASHRAE research behind how ASHRAE has made that determination. And then the Radiant Time Series is the other one that is currently described, which is a simplification of the heat balance method. And I think a really important—there's a bunch of distinguishing factors between those two, but the biggest one for me is you can't use Radiant Time Series to do an hourly load calculation. You can only use it to find the peak, just quoting the handbook here.
And so heat balance method is being the most accurate and being more flexible in that I can use it to say, do a heat pump system, like a ground source heat pump system, figure out the hourly load throughout the year. How many hours am I in a given range? It's going to have the confidence then to do more of a balance of the loop load, for example, in a heat pump system. Or if I want to use it to size a thermal energy storage tank, I would use the heat balance method. I couldn't use radiant time series for that. So it's more accurate. It's more flexible. The trade-off is it does require a few more inputs. So typical inputs are going to be your internal gains. We talked about those earlier, people—
Matt Price:
Golf simulators.
Nathan Kegel:
Golf simulators. You have baseball batting cages, all of those things.
Matt Price:
Yep, everything. I just want to golf sim so bad.
Nathan Kegel:
Yeah, it’s that time of the year. The Masters week, it’s perfect.
Matt Price:
Restart with the inputs. It takes more inputs.
Nathan Kegel:
Well, there's some fine-tunes in the input. So the internal gains are well understood or they're generally well understood. And I mean, yes, there's unique cases. So this is the case where the exception proves the rule and the handbook gives you a lot of numbers that you can refer to for some of those unique cases of internal gains. I think where the inputs vary is how detailed are you with your internal surfaces? So let's just set aside the envelope for the moment. I'll leave that for last. If I look at Standard 183, it tells me I need to include the interior surfaces, all of them, because I need to understand where the sun is hitting an interior surface and potentially transferring heat further into the spaces or into the building. But it's also because all of the internal surfaces have thermal mass to them. In other words, they absorb and release heat at different rates.
So the peak, if I'm going back to the handbook again, the sum of the peaks does not equal to cooling load frequently or maybe even ever. And so the time delay effects in those aspects is one of the trickier bits to get right in the load calculation methodology. Now, where the heat balance method differs on the inputs specifically is you need more than just your R value for a wall, for example, to really leverage the capability and the accuracy of the heat balance method. Do you have insulation in your wall? If so, what kind of insulation? How much of it do you have? Or are you using concrete or brick or some other heavier thermal mass material that has a different time delay effect than say lightweight metal panel?
So the Radiant Time Series simplifies those calculations, you don't necessarily have to get into all of the details of defining those surfaces. The trade-off with it has already been described. So the inputs are generally the same, but the finer details can matter significantly depending on the method you pick and the product—
Matt Price:
Yeah. And I'll say outside looking in, again, formerly as a rep and getting more into design interests, it seems like method is so constantly discussed for load calculations. It's like a common discussion where in duct work design, piping design, equipment selections, it's not super common to discuss method. So over time, or historically, I guess there's been simplifications of what was ideal or interesting, sort of, "We know we could do this, but we don't have the capacity to do this." Is it mostly in your mind because it was pencil and paper 100 years ago, and then now we have calculation capacity and computers that's like, we can do what we want? Roughly how many calculations is occurring using heat balance method versus the rule of thumb that would've been done 100 years ago?
Nathan Kegel:
Oh, there's certainly a huge difference. And technology has definitely played a role in the availability of using more sophisticated methods or more accurate methods. So yes, to answer your question directly, if we think about how load calculations were done before everybody had a PC on their desk, we don't have to even go back 100 years, we can go back probably 50 years and everything was done. There were worksheets and ASHRAE did distribute load calculation worksheets. People would work through the problem by hand. There were time weighing factors. You can still see the time weighting factors for the rating time series. They're still published in the handbook if you're curious about where those come from. You can do all of those calculations by hand and then somebody makes a change and now you got to go through that process potentially all over again. So we learned as an industry by hand and to trust our hand calculations.
We could see the work and we could find errors in that work. So with the move to software, and I think a parallel change here that I think might be helpful for people to understand is moving from T-squares and drafting tables to CAD is sort of analogous to moving from hand calculations to load calculation softwares. So there's a certain element of the more sophisticated mathematics that's occurring in the software routine or the software program that lends itself to be a little bit more of a quote unquote "black box" or a little bit less transparent than manually doing it by hand. The trade-off, this is of course the calculator, the computer, can do those calculations much more quickly and doesn't make mistakes because it's coded in. So as we get into like the heat balance method, now we're doing sophisticated like differential equations and unless—me, myself and I would not want to do diff EQ by hand.
Matt Price:
Speak for yourself here, Nathan. I don't mind doing it.
Nathan Kegel:
Multiple iterations multiple times per hour, because what actually happens is the calculations happen in a six minute or 10 minute time step or even smaller in some cases to resolve exactly the time delay or calculate the time delay rather than to apply a weighting factor, the time delay in the heat balance method, for example. So yeah, more sophisticated math, more frequent calculations, more rigor to the calculations to get a more accurate result. So this also has changed how we think about reviewing a load calculation report. So at the end of the day, the old hand method, you'd get a number at the bottom of the page, that was, once you went through the QA process, made sure that was right, that was what you had. You'd apply a safety factor, all of those sorts of things. You still apply safety factors in the calculation softwares that are available, but the way that those are reviewed is different.
So Radiant Time Series is probably more akin to internal gain plus envelope load, plus ventilation, add those up, you get a number, kind of easy to understand, it's arithmetic more or less. With heat balance method, it's different since there's so many more calculations happening and the calculations are inherently more interconnected with differential equations. Solving that and pulling that apart and how you do that QA to make sure it's quote unquote "right" is a different process to go through. And this is where I think transitioning from relying only on reports and numbers and a spreadsheet or on a page and using more visual tools like graphs and charts really comes into play to understand the physics, especially when you're looking at a totality of a design day. And I love loads reports, don't get me wrong, loads reports are great. They give you a lot of information in a format that is usually pretty familiar to someone like myself.
They fall short of being able to show you what happens over the entirety of the day. No matter how many reports I generate, it's only going to give me one number. That number is static on the page and what's happening in the calculations is so much more than that one static number. So I think one of the transitions that we've had to adapt to is that moving now from CAD to BIM. So moving from calculations that are algebraic in basis and moving into more sophisticated math encourages us to do QA/QC differently as well.
Matt Price:
And at that point, it becomes more of a simulation and less of a calculation. It's like more dense than just a single calculation. You're getting multiple calculations iterative and then you're arriving at an optionality heavy set of choices that you can make because you have more outputs.
Nathan Kegel:
Yeah. In a way. The simulation is still simulation. So simulation, when I teach this, the way that I kind of think about this is if I want to look at one day, that's a calculation. Yes, I might simulate a lot of things within that one day to arrive at the calculation. And I think that's what you're kind of driving at and I'd agree with that, but I want to be careful to draw the distinction here of I'm going to do a simulation over an entire year, and that's going to contain exponentially more of those calculations than just doing the calculation for one day or a series of calculations for one day.
Matt Price:
And a specific calculation that I want to ask about is ventilation calculations because to me it seems that ventilation quantities, which we will not get super into detail for because that could be multiple other podcasts, of course, but there is within the practical application-based workflow, a need to calculate ventilation. And that is often wrapped in when we're talking about load calcs, people are also thinking that's the point at which you sort of think through the ventilation calcs. Is that correct? Is that when that's typically happening in the workflow of the design engineer?
Nathan Kegel:
Yeah. Ventilation calculations, figuring out how much outside air you need to pull through your HVAC equipment is an essential part of this. I think we neglected to dig into that. We were talking about inputs. I think I still got stuck on room and zone and never transitioned the system. But yes, certainly. The challenge with getting ventilation right is that if you follow the 62.1 ventilation rate procedure at least, which is probably the most commonly used one, at least in North America, the challenge with that is you really do need to understand your system type in many cases to figure out what the system level ventilation is going to be. Now that said, you can still estimate it reasonably well by looking at your zones and adding them up before you've picked a system so that you have some reasonable expectation what that's going to look like, and then fine tune that as the design evolves, but you cannot pick a coil size without understanding your ventilation.
Matt Price:
Yeah. That's what I was kind of getting at is you need it and then the load calc also informs the system top and then you might have to iterate back and forth through that. So unless, like you said, you love doing—
Nathan Kegel:
Definitely enjoy the process.
Matt Price:
Unless you love doing differential equations, it's much better to have a robust calculation method such as heat balance method to do that over and over.
Nathan Kegel:
Well, it's also, to your point, I think it's really important. It's a big time saver and one of the things I know is happening across consulting engineering is there's always pressure on fees. So any place that you can save some time helps your fee and helps your profit margin. So it's really helpful if your load calculation tool can also do your ventilation calculations reliably and accurately in the same tool, and you don't have to have multiple things to manage across those bits. Of course, you have to make sure that you understand how the ventilation calculations are working because there is legal risk about under ventilating space, for example. So there's a lot of points where you can go wrong and therefore there's a lot of points of checkpoints that happen that I think engineers and people who are training on how to do these processes need to be aware of.
Matt Price:
Right. And as we sort of turn toward closing this out, I think to me, understanding the ventilation calcs is critical and it's the world that I lived in a little more and then it would be, here's the tonnage you use. But ventilation, I thought about quite a bit because it swings the cost of equipment so, so heavily. And you want to go heavy enough that it's not a risk, but you want to go light enough that you're not buying a ton of equipment that you don't need. So I would say in my load calculation view, that would be a common mistake or misconception is we're overdoing or underdoing ventilation, or we just need to really stop and think, "Are we doing this calculation correct?" Because it has outsized weight on the outputs. What you're getting out of it is going to be very expensive or undersized, which are two different problems. So are there any that you regularly see or when you give this a similar talk at chapters, are you getting the same questions over and over? What is a common friction point or issue with load calculations in 2026?
Nathan Kegel:
Oh man, there's a lot of them that are all over the map.
Matt Price:
Well, and this can be an opportunity to also say any more methods, things that you would like to say. So any more things that we didn't cover that you commonly are interested in hitting because it's an important thing.
Nathan Kegel:
Yeah. A couple of examples that I use in the talk that you've seen, and maybe others who are listening to this have seen as well. With the heat balance method, the ability for materials like a concrete or a concrete masonry unit to hold heat for an extended period of time and slowly release it can lead to unexpected peaks. I use a performing arts center with a large southwest facing curtain wall as an example of that. And then the interior space is unexpectedly peaking in the middle of the night because that the sun essentially cooked that CMU all day long and it took hours for the heat to work its way back out of that, which is why in Standard 183, you've got to have your internal surfaces. You can't just assume that they don't have any bearing on your load.
Another one is, a common assumption that I've seen used is the design day is a 24/7 period, even though my building is an office and it's not going to be occupied 24/7. If I follow Standard 183, I need to use an occupancy profile and not assume that it's 24/7 for that space type. And then if you have setbacks in there, you can have unexpected peaks occurring due to morning cool down where you have a setback, say of 85 and a space temperature of 75. And then what triggers in our sort of heuristics when we review this is we say, what I see a lot of times, "The software is wrong. I couldn't possibly be wrong." But I see a lot of, "The software is wrong."
And then with a good software platform, you can actually dig in and say, "Oh, actually that does make sense. I just need to tweak my inputs because heat balance method is more sophisticated than the other methods, and I wouldn't have seen that with a hand calculation." It doesn't mean that the hand calculation’s right or wrong or that the software's right or wrong, it means that we need to take a step back and understand our first principles before we proceed.
Matt Price:
First thing that I was taught in college for a computer class was the computers do what we tell them to. They're not thinking and we can get into the future here in a second, but they're probably still not thinking and that we're responsible for what goes into them and really therefore what comes out of them too.
Nathan Kegel:
I think what you're getting at what I would get at with that is professional liability. It doesn't matter what software platform you use or if you want to talk about AI or whatever.
Matt Price:
Oh, I do.
Nathan Kegel:
You're still liable. As the person that had put your stamp and signed that set of documents, you're still liable. So if you want to go down the path of trusting some autonomous agent to do something for you and you don't understand how it works or how it arrived at, that's risk.
Matt Price:
Yeah. But same goes for a simple spreadsheet calculation. If you're using a heuristic spreadsheet calculation, you're still responsible for that. And I give a talk to chapters on, it's called From ABET to HVAC. So it's about engineering programs and how ABET accredits programs at the program level. And we always get to the discussion of judgment and we are responsible for judging it. And ABET literally says in their text of what a program is required to do, that judgment is something that engineers should be learning. So that takeaway is we feel something. So I live in a red brick house that has south facing stairs and at 9:00 PM in the summer when I take the trash out, that wall's hot. So you know, you can sense it. So loads are something that we do sense, but at the end of the day, we're getting into a territory that is unknowable by your mind and you need some software assistance, essentially. You need some calculation assistance to do it properly.
So to me, it's something, yes, you can double check. There's a QC/QA process to this. Does this make sense? Is this even reasonably close? I call it an idiot check for myself, but not for others, of course. But we feel these things, we know they're occurring, but we need some help. So we better get our inputs right so our outputs come out correctly as well. And also method selection is important for this, even more important than some other topics that we deal with day-to-day.
Nathan Kegel:
Yeah. I do think that method is often overlooked and I don't think it's overlooked intentionally. I think there's just... You use the term judgment, and I think I've said, and maybe not at your chapter, but I've said this before and I read it somewhere, it's where it came from. What engineers call judgment, a scientist calls bias. And it's the same thing, but it's described differently from a different perspective. We have to apply judgment largely because there's so many unknowns. And the scientist is trying to get rid of bias because they want to eliminate the unknowns to get to the foundational truth of it all, if you will. But to your point about methods, yeah, we're all still human. It doesn't matter if we're trained as engineers, we're trained as architects, we're the garbage man on the street, we're all still human to use your garbage analogy.
Matt Price:
Right. We feel it.
Nathan Kegel:
We all have our inherent, we learn something and we become a creature of habit. We know what we know and we're comfortable with what we know and to learn something new requires a little bit of getting out of our comfort zone. And when we mix in the professional liability piece of that, that can feel really, really scary. So I think that, at least in my view, we cannot rest on T-squares and drafting tables. In other words, we cannot rely on technology that we know is not put into practice and used regularly. And the heat balance method's been described for 25 years now. So at some point we do as professionals, I think, need to take the onus upon ourselves to keep pushing and learning and exploring so that we continue to deliver not only better buildings for our clients, but also understand how we can be more profitable in the consulting engineering practice.
Matt Price:
Well, and that brings us to a very good question, which is where do you see it going now? These are sometimes junior roles doing load calculations. Obviously, that's not an everyday, every firm situation, but this is often a task that newer folks use. And we're hearing a lot about where is AI going and what is the agentic AI going to do in our workflows? What do you think and see? Thinking about this all day, what do you think happens and what do you think remains? And I assume that's going to be the judgment thing that we're discussing. What do you think is going to happen with AI? What do you think is going to happen with the next five to 10 years of load calculations?
Nathan Kegel:
So AI in the context specifically of load calculations and not the utter collapse of—I mean growth of society.
Matt Price:
Yes. We have a tight scope here, Nathan. Very tight scope. Yes, sir. Yes, sir.
Nathan Kegel:
In the context of load calculations in AI, I think there's a number of things that are going to happen. And it really comes down to tasks and it comes down to availability of what you can use to train AI. So ASHRAE has an AI policy and you cannot upload ASHRAE standards to train AI. That's the ASHRAE policy at the moment. Whether that stays the same or changes, I think will have a big impact on where the AI trajectory ends up going. But at a minimum, you can kind of break down load calculation tasks into, "Well, I can reference what's in the code book, IECC, whatever, in terms of minimum U values, solar heat gain coefficients, lighting power and density.
Those are the things that are known and I can train an AI to do those inputs for me, automate a bunch of stuff essentially. Not do the thinking, but do the population of the data in a reliable way. The part that I think AI is going to struggle with, at least in the short term, and this can change because I know AI is evolving rapidly, but the judgment piece. I've heard people talk about this like in the '70s, there was a New York Times article about professors and calculators and how students shouldn't get calculators because then they wouldn't learn math. And we all know how that turned out. Now calculators are deeply ingrained in the curriculum. You're expected to buy a certain calculator at this point.
Matt Price:
That costs $160.
Nathan Kegel:
I don't know what they cost but they cost money.
Matt Price:
It costs a lot of money. Yeah. Specific ones. Yeah.
Nathan Kegel:
But the difference is the calculator, to your point, is you're putting the numbers in and you're doing the function to get the output. AI is not a calculator. AI just takes some things that you give it data and you tell it what to do and it does its own thing in a way, and it gives you an output that you cannot necessarily calculate yourself. That's a really important distinction when it comes to risk and understanding how much risk you actually carry. So I think we could probably put AI's impact into probably at least four different categories about in the process of load calculation, sort of the known knowns, the unknown knowns, the known unknowns, and the unknown unknowns, if you will. And AI probably has a bigger impact in two or three of those quadrants and probably a pretty minimal impact in the fourth one.
And because AI is not a calculator, it's dramatically different from a calculator, I still do have concerns about how we're going to teach and make sure that the younger generations that are coming up in this industry are still getting that really important knowledge that doesn't live in a book. The judgment and the application of that judgment in the design decision making process as it comes to down to load calculations is not something that you can write down in code or script because every building is different. Every process is different.
And yeah, you can rely on certain things that you learned from this project and certain things from that one, but that's accumulated over time in a pretty uniquely human way, at least at this point. So unless there's some major breakthrough in AI can learn on its own without the need to be trained, then I think there's still going to be that need. And that bridge, that gap concerns me for how we're going to address that as an industry.
Matt Price:
And I'm sure you're similar. Obviously, I started pre-AI. There was heavy machine learning and stuff going on when I started, but I wrote an entire hand calc out for a load calculation and did all the wall calculations and the floor calculation and added internal gains. And it's rough, but it's akin to having the arithmetic until seventh or eighth grade when that calculator starts to come into play. So I do think that's part of it, but like you said, it's also years and years of seeing it, which is currently not being trained into models. And how does it get to that point? I don't know. Chips in our brains maybe, and where does it go from here?
Nathan Kegel:
I think I put it this way, if we think of AI having the capability to function as an EIT, to quote unquote, "replace an EIT," well, how do the actual EITs learn if they're being replaced?
Matt Price:
I have big theories on this. I have big theories on this. This is what I think about all day. But I think honestly that firms will grab people and pay them with an expectation that they're not making the firm money for potentially a little bit longer. Maybe they go do other things, but we see it every day now. That's really a lot of times how a rep firm runs is they have underutilized salespeople for their first year or two years. Sometimes I heard two to three years when I started where it's like, you're supporting, but you're not really making money for longer.
And MEP firms have in mind, "This should take this long," and it actually, it just makes those first couple years look different. And like you said, replace an EIT, which I don't believe is actually—it might reduce the number of them on a macro level from a need perspective, but my bet is jobs-wise, it's probably pretty similar to what we have now because the firms are still going to want to grab the good people or who they view as powerful and potential good PEs or whatever, eventually good designers, they're going to grab them early and just figure it out.
Nathan Kegel:
I think it's a big unknown and that's certainly a possibility, but I think there's a lot of things that are... The future is unknowable and unknown and it seems more—
Matt Price:
And this is going to age very, very badly.
Nathan Kegel:
Yeah. I guess we could get into the ramifications of that, but yeah, there is the other side of that coin, if I can play devil's advocate on that side, is if I look at this, we were talking before we got onto recording this about how many hours we worked. There is a shortage of people in this industry, of young people in this industry that are really keen and are capable of learning this rapidly. I can talk to any engineering firm probably, most of the engineering firm that are going to listen to this would say, "Yeah, we're trying to hire people. It's really hard to find good people." That's been a problem for decades probably. And ASHRAE has had recent presidential themes around we've got to have a more powerful and educated and engaged workforce. So that's all part of that as well.
And I think if AI is deployed in a strategic and powerful and correct way, if you will, then it has the ability to enhance or augment or improve a lot of those bottlenecks. But I also am aware of what people who are creating AI, what their intent ultimately is, and their intent is to be more profitable for themselves. So there's these competing forces about what is going to happen in the future. And short answer is we don't know. I think we have a hopeful of what we want... What we hope happens, but then there's also some valid concerns about what could or what might happen.
Matt Price:
I want to close out with more of what we do know and not about what we think might happen because I think you and I have historically went on very long about what could potentially happen at CRC, but what can you say about load calculations that we know? What's a big broad overview summary, good inputs get good outputs, methods have been proven, ASHRAE has positions, et cetera. What would you say our closing thought is for this?
Nathan Kegel:
I think load calculations are the single biggest, I've talked about risk, but they have the biggest potential about the actual performance of the building that is often not considered. I don't have a hard set of data on this, just anecdotal conversations with people, but probably 5% of the buildings that get designed get an actual energy model, like a full-blown, like understand the hourly load, understand the EUI, et cetera. But every project gets a peak load calculation. Amory Lovins wrote in Natural Capitalism in the '90s, I think it was, that you can be sued for insufficiency, but you can't be sued for inefficiency. So that's really been the guiding principle. But load calculations impact the size of the equipment, which is the first cost of the equipment, and typically your oversized equipment, and there's studies on this both modeled and measured.
Oversized equipment generally uses more energy. It's got bigger motors, bigger fans, bigger compressors. It generally requires more refrigerant, which refrigerant transitions are a real thing right now, they have environmental impacts and all sorts of things. So how much refrigerant you need is also a source of risk, especially going forward. Because of short cycling, they could lead to control issues and thermal comfort problems, humidity issues inside of spaces. They have a big impact on the actual performance versus the estimated or the modeled performance and design. And there's potential code compliance ramifications going forward because Standard 183 now is referenced as how you must do your load calculations for ASHRAE 90.1 and IECC. So the simplified methods that we've used are really rapidly disappearing because our legal minimum floor is changing very, very rapidly. So it becomes another motivator for engineering professionals to understand more up-to-date or more current methods to minimize those risks, not only for themselves, but frankly for their clients.
Matt Price:
So in short, "Slow down. Get the load calculation correct. Make sure it's correct before moving on to the other design pieces."
Nathan Kegel:
Yeah. You've got to have confidence in your loads before you're going to have confidence in your design, for sure.
Matt Price:
Awesome. Well, thank you, Nathan. I really appreciate it.
Nathan Kegel:
Thank you, Matt. Good stuff.
Matt Price:
Yes, sir. Have a good one.
ASHRAE Journal:
The ASHRAE Journal podcast team is editor, Drew Champlin; managing editor, Linda Rathke; producer and associate editor, Allison Hambrick; assistant editor, Mary Sims; associate editor, Tani Palefski; technical editor, Rebecca Norris; and creative designer, Teresa Carboni.
Copyright ASHRAE. The views expressed in this podcast are those of individuals only and not of ASHRAE, its sponsors or advertisers. Please refer to ashrae.org/podcast for the full disclaimer.