Episode Description:
In episode 464 of the Smart Buildings Academy Podcast, you'll explore the game-changing ways ChatGPT can enhance your work as a building automation professional. From troubleshooting and system design to programming and sales strategies, this episode is packed with practical examples of how AI can streamline your daily tasks and drive innovation in your projects.
Key topics you’ll dive into:
►Leveraging AI to simplify sales pitches and proposals
►Enhancing troubleshooting techniques with AI-generated insights
►Using ChatGPT to assist in creating sequences of operations
►Automating programming tasks for BAS systems
►Unlocking AI’s potential in facility design and optimization
If you're ready to see how AI can transform your approach to building automation, this episode is a must-listen!
Key topics you’ll dive into:
►Leveraging AI to simplify sales pitches and proposals
►Enhancing troubleshooting techniques with AI-generated insights
►Using ChatGPT to assist in creating sequences of operations
►Automating programming tasks for BAS systems
►Unlocking AI’s potential in facility design and optimization
If you're ready to see how AI can transform your approach to building automation, this episode is a must-listen!
Click here to download or listen to this episode now.
Resources mentioned in this episode
- Free Technical Skill Assessments
- Smart Buildings Academy Workforce Development Program
- Smart Buildings Academy Courses
- Building Automation Training Guides
Podcast Video
Transcript
Phil Zito 0:00
This is the smart buildings Academy podcast with Phil Zito. Episode 464. Hey folks, Phil Zito here and welcome to episode 464. Of the smart buildings Academy podcast. So I'm re recording this intro because actually, as I got about more than halfway through the podcast, I realized that there was some really powerful stuff that we showed at the end of the podcast. Now let me be clear. What I am not telling you is that the beginning of the podcast is not powerful or is not interesting. That's not what I'm saying. However, I encourage you, because I do cover some more salesy use of artificial intelligence in the beginning of the episode, I encourage you to stay for the entire episode, because as we get deeper into this episode, we really push the limits of chatgpt, and some of the things we get it to do are really, really awesome, like goosebump inducing, awesome. And so my encouragement is for you to stick through the entire episode. This is definitely an episode you want to watch. So if you're listening to this, you want to go on podcast that smart buildings, academy.com forward slash 464, once again, that's podcast that smart buildings, academy.com forward slash 464, so you can watch the video. With that being said, we're going to kick over to the podcast. All right. So the first thing I'm going to teach you all is how to actually use chatgpt. So when I see a lot of people approach chatgpt, they ask it to do things, but they don't set the correct context. You really want to set the context for what the large language model and the prompt is doing. See in each prompt, especially if you have the enterprise version like I do, it will start to have memory. It will start to learn. There will start to be context. And based on all of this information that it learns, the memory that it creates, it will be able to answer questions better and better. So I'm going to tell it. You are a mechanical engineer. All can see, I can't spell engineer with an expertise in building automation, design and operations, I am uploading a document, and I need you to answer questions from this document. Okay, so I'm gonna upload the document and grab it, and it is a MEP document that I found just going through Google and searching for.edu sites with MEP documents. This one, I believe, is for a chiller replacement. So I upload the document, and I hit Enter, and it's going to say, what questions do I want? I'm going to say, please tell me what mechanical systems in bullet format are part of this project. So what I want you to start thinking about here, and you should always trust but verify. You should always verify any of the information that chatgpt gives you. It can still hallucinate a little bit, and it will do quite often lazy responses, where it only gives you a little bit of detail. So for example, I'm going to say, Hey, give me the sequence of operations for this chilled water plant.
Phil Zito 3:59
For this chilled water plant. For those of you listening, you can also watch this recording after the fact. I will try to articulate what it says, but it's going to be a little bit difficult, because this is very visual. So for those you listening, you'll want to go to podcast that smart buildings, academy.com, forward, slash, 464, and actually, watch this video. You notice it gives me kind of a little bit of response. And this one is better than what it gave me previous when I was testing this out, but it says all these things. Hey, it's Johnson Controls, Bas, enable, start, stop status and alarm monitoring gives me all this data, and then if I said something to the effect of, can. You tell from the document where this chiller will be physically installed. So I want to figure out words here by parking lot. Okay, I gotta be more specific what state and city, so Greenville, South Carolina, all right, based on the state and city, how should I operate this chiller in order to maximize the energy efficiency of the central utility plant? I that's gonna give me a bunch of And okay, so you're maybe wondering, like, what would you do here? Well, if I'm in sales and I want to go provide more value to the customer, and I see this PDF, come on the street for bidding, and I'm going owner direct. What I'm going to want to do is go to the owner and be like, Hey, I saw your sequence. Love to bid it for you. Here is a couple strategies that I think we can do that will optimize your central utility plan. Now we see that it's Greenfield Technical College. Does Greenfield Technical College have a sustainability plan and or target? So now it's going to go search and it's going to see if there actually is one so finds their website strong focus. Lead, okay, cool. All right, I'm pitching the facility director and Energy Manager of Greenville Technical College, to let my company bid this project, I want To create a narrative that shows understanding of their sustainability, of their sustainability goals, and also positions myself as a subject matter expert. Give me a talk track for my initial sales visit. Boom. So we'll focus on some sales stuff in the start of this episode, and then once we get that knocked out, we'll go into like, some technical stuff. So, you know, it gives us some talk tracks gives us some things we could say, etc, etc, do, okay, but it's a metastasis system, and we don't do Metasys B, are not a medicine contractor. This could be any system. We're not a mess. This contractor. Help me compel the customer to accept an alternative building automation system. Put this in bullet points to make it easier for me to create or to present. All right. So, right, we're doing this, blah, blah, blah, right stuff about open, etc, etc, right? So it gives me all these potential things, and what I want to do is convert this into a slide deck format that I can then present to the client. Okay, boom, so. So what I hope you see what I'm doing here is I started off with this PDF. You're in sales, maybe you're outside sales, maybe you're inside sales, at a distributor. You're trying to help people out. So I put the context in place. I tell it what it is. I tell it to answer questions for me. It details out the system. It gives me the sequence. Then I'm like, Hey, how can I make this better? First, I have to know where it's installed. Okay. Then, how can I make it better? It gives me some ways. Then I say, hey, research the customer. Find out if they have a sustainability plan or a target gives me some information. And I'm like, Okay, well, help me tailor a sales pitch. And then I realize, oh, it's got a system that maybe I can't install because I'm not a provider. And I say, All right, then help me put together some bullet points to deal with that objection, and then put this into a slide deck that I could then present to the customer. All right, let's open a new chat. Let's get something going. For those of you who are technicians, how can you use this to troubleshoot? I'm going to go on the other screen real quickly, and I'm just going to grab some random images off of Google search, building automation, graphics, Hu, I'll go to images all right. I've got just this random image here. I don't even know what this is, so I'm going to type in create an optimal sequence of operation. Well, just say, create a sequence of operations. Create an SOP for this document. Now you can put this actually into a context, if you wanted, you could say you are a building automation expert who has a who has A skill set in site surveys for bas retrofits, all responses will be in that context and will be providing guidance on how to best identify existing systems provide optimal sequencing and retro fit recommendations. Please give me a sop for this image. So you go to a job site, you are able to look at their graphics. You find an image on the graphics, you copy it, you paste it in here, and you ask it, give me a sequence. All right, so it's going to go and try to develop what's going on here. Supply fan. It looks like it's on, so indicates that it's running control strategy. How do we want to run it? It's saying, use DCV. Gives a bunch of things, okay, gives me a bunch of information I'm going to and see it also gives me right my retrofit potential. And then finally,
Phil Zito 13:58
convert this information into a points list and a sequence of operations. So right here we can see, right we got Sensor Calibration. VFD gives me a bunch of potential optimizations. I'm going to say, convert this information into a points list and a detailed sequence of operations. Where'd my coffee go? Here we go.
Phil Zito 14:38
So it gives me detailed points list. It actually tells me the current state of things. And then I'm able to take this information. It's going to give me a sequence of operations, and I can go then present this to the customer. Let me find a. Image of a VAV box that potentially has an issue.
Phil Zito 15:09
So let's look at this. Let me see what I find.
Phil Zito 15:20
I think this one's going to provide a whole slew of issues.
Phil Zito 15:29
So I'm going to just open a new prompt. I'm not going to put us into a context here, just for the sake of time. But what potential control issues can you identify from this image?
Phil Zito 15:51
This is a great way to do kind of a quick spot check and see if there maybe are some potential issues. So airflow versus set point. The actual airflow is 11. Set Point, 75 significant discrepancy. Right reheat valve position. It's at 34 which means the reheat. However, the room temp is already at 87 which is much higher. I mean, obviously this is just a graphic with garbage data in it, most likely supply air temp is showing 17.8 which is low. Maybe an incorrect sensor reading is air temps, the slow are unrealistic. So this is an interesting thing that you can do if you are someone who's in a facility, or if you're a newer technician and you're trying to figure things out, all right, sorry about that. I thought we had a issue with the recording, and I didn't want to keep talking and make sure or and lose everything I talked about up to that point. So when would you use this? Well, what you could do if you're a building operator, if you're a technician and someone sent you a screenshot, like a customer sent you a screenshot, you could go and actually pop this in the chat, GPT, ask it to identify potential issues and potential recommendations. Where would you recommend beginning in troubleshooting this unit.
Phil Zito 17:35
So it says, I would recommend the following steps. Verify this, because this is the most likely issue, check this, then check that. So this can actually help you create a priority to how you go and troubleshoot. I hope you all are starting to see kind of the power of the couple use cases I've shown you here. So far, I've shown you the use case of as a salesperson. I've shown you a use case as someone doing a site survey, and now I'm showing you a troubleshooting use case. I'm going to try to do something a little squirrely here. I don't know if this is going to work. I have not tested this yet, and I have no idea what result it's going to give me. So let's try this out. I'm gonna go upload from my computer. Let's grab I believe it was, excuse me. Did was this document? Let me double check. Actually, I know what I'll do. I'm gonna pause the video, and I'm actually gonna grab a different document. Just give me one second. Alrighty, here we go. I have no clue if this is going to work, but we're going to try it out. I'm going to grab a sequence of operations here. Let me zoom in on this. And I am going to try to tell chat GPT to program this using the kit control library from Niagara. You are an expert in Niagara with a with knowledge of the kit Control palette, which I realized I just spelled wrong. Take this SOP snippet and tell me how to program it using blocks from kick control. I have no clue if this is going to pull this off. I have no idea. Let's see this could totally just go. I feel like it's going. Go wrong. I do not think this is gonna go right.
Phil Zito 20:09
Okay, so it's identifying some modes or needs a state machine block from the num switch from the kick Control palette, okay, I'm use binary blocks for each of the and then handled using a relay.
Phil Zito 20:36
This is interesting. I did not Wow. I did not expect this to actually, even potentially. I'm not saying this would work, but Okay, can you represent this in such a way that a junior programmer could follow the instructions and build This program specifically which blocks go, where and what connections to use, or let me use the Niagara terminology, what links to use between blocks. I think we're going to break it. Let's see. All right, this is so wicked sick. Go to the kick Control palette. Drag the enum switch block into the program. Add the following states, do, to do, to do, drag a schedule block into the program. I should have done this at the beginning of the video. I feel like this would have been more exciting to people than what I did. I actually might do a disclaimer. I might record the intro again and be like, Hey, make sure you stick around, because at the end, we're gonna show you something that's really cool. All right, so does this perfectly get us to where we need to be for programming? No, I don't think so. Does it get kind of close? I mean, this is pretty cool. It gives me the blocks. It tells me what alarms to do, tells me how, Oh, I was not expecting it to do this. This is so cool. I want a full block layout, like the example layout you did
Phil Zito 23:12
the future, I'm just going to be wicked. This is so cool. This is going to change how we program.
Phil Zito 23:32
So a vision of mine, something that I thought would be so cool, if I had unlimited venture money, I would upload each programming manual and programs, like screenshots of the programs, and then build an API into the programming software, or build like a mouse control interface. And then I would tell it which building automation programming software I'm running, tell the AI, and then the AI from its library, I would give it a sequence, and it would know how to open the executable, and then it would control the mouse, and it would actually lay out all the blocks. So basically, automated programming. That'd be pretty cool. This is interesting. I hope you all are starting to see kind of the things you can do with AI, how you can make it help you out. Let's, let's try something squirrely. Change the directions to the JCI, CCT software, change the programming directions. Change the programming directions for the. For this to utilize the JCI CCT software, let's take a look the tree structure. Do go into the CCT software? I want to use the system selection tool to build the initial program. So this is pretty cool, folks. I hope you all are finding this enjoyable. I want to use the system selection tool to build the initial program also make any recommendations on the state table changes. Surely, there is a lot more to be done to build this, to make it actually usable, I'd really have to spend time like I'm creating these prompts right now, while I'm talking to you, I'd have to actually spend time and think through the prompts and make sure that prompt engineering is done, right? And it gets me the outcome, and the outcome is predictable, because the worst thing is to hand this off and then not get the same output. Let's do this. Want to use the system selection tool to build the initial program. All right? So using so open the System selection tool from CCT, holy crap. I did not expect it to, oh, that's wicked. So it must have actually went and found the document. And, oh, I'm sure this is not precise. I'm sure, having used the system selection tool in the past life, that there are more options. Okay, I want you to be very precise in the SST. Oh, you little. It's building the state table. Look at that. Oh, this is so cool. Oh, man, this is awesome. I want you to be very precise in the options. I options and sub options I should select in the SST. Well, let's see what it does now. And next I'm going to do Siemens. I just want let's do line code. Ah,
Phil Zito 28:09
look at this. Look at this. This is so cool, like, my goodness, this is, I hope you all are really seeing because I know I'm gonna get people that are like, Yeah, but you can't trust it. It's not there. Blah, blah, blah. I hope you all are seeing that there is a potential, like, if I asked GPT to do this two years ago, no way. But now I think it's, it is wicked. Close. This is amazing. I'm like getting goosebumps. Okay, write the line code for this program using the Siemens. Oh, is it? Ppcl? Ppcl, programming language, forgive me, Siemens, people, I remember. Ppcl, if there's a new program, then I apologize. Oh, this is so cool.
Phil Zito 29:39
What does Delta use, do any of you know I'm talking to you like you can actually respond like you could be like, yeah, Phil, of course, I want to do Delta after this. Use the Delta line code, programming language for their. It's not web control. What is it? Or no web controls? ALC, obviously it's it's not orca anymore, though, is it? I don't remember. Let's try this. Use the Delta line, code, programming language. I don't know what delta. I don't remember, hmm, GCL plus, okay, see if it's any different. Is it any different? Feel like it's going to be the same. Yeah. Basically, oh my gosh, this is wicked sick. Okay, so what I hope you've seen through this episode is that you can do some pretty amazing stuff with AI. Also, I Oh, okay, let's see. I want to try one more thing, and I don't think we're gonna I do not think this is gonna work. Cargs back, net, wire, Shark captures. I There is no way this works. Okay, let's find here. I want to find a failure. Find a failure.
Phil Zito 31:39
Ppmt, error, that's one I want to find, like a,
Phil Zito 31:51
is there, like a bad device issue, like a baud rate mismatch or something bad hub, critical, all right, whatever. Let's grab. Let's grab this read property. Let's, let's try this out real quick. Analyze This BACnet. No, we're not even going to tell it that it's BACnet. Analyze this capture and tell me what I need to know. We're going to give it absolutely no context. I mean, it knows it's building automation, probably because it's in this prompt. I don't even know if it's going to be able to open this I'm
Phil Zito 32:45
let's see, I highly doubt. Let's see what it does.
Phil Zito 33:01
Possibly from Wireshark. I would need to parse it as a package capture or, Oh, crap, I figured out what it was. I didn't even so cool. It's, it's thinking I
Phil Zito 33:25
oh yeah, I couldn't do that. It doesn't have anything to analyze in here.
Phil Zito 33:38
Oh well, okay, so we hit a limit, and with that, I think we will end the episode, folks. I hope you had fun with this. I hope it gave you an idea of some things you can do with GPT. Gave you some really cool ideas about how you can use it for programming, design, for sales, and a bunch of other things. As always, if you have any questions, you have any thoughts, do not hesitate to hit us up. Everything can be found at podcast smart buildings, academy.com forward slash 464, once again, that is podcast smart buildings, academy.com forward slash 464, thanks everybody, and I hope you all have an awesome day. Take care. You.
