Using AI & Predictive Analytics to Enhance Operations

Using AI & Predictive Analytics to Enhance Operations

The current market for AI in predictive maintenance is experiencing renewed interest, largely driven by advancements in generative AI. However, it is essential to approach AI adoption strategically. Organizations should first identify the specific problems they aim to solve and then determine which AI technologies are best suited to address these issues. While generative AI has brought back the hype to adopt Artificial Intelligence, adopting the wrong AI technology, however, can lead to purchasing solutions that generate poor data, rather than more power.

To gain deeper insights into the practical applications of AI in predictive maintenance, watch this comprehensive session. This webinar will showcase proven strategies and real-world examples of how customers are leveraging AI predictive analytics to enhance their operation.



--
TRANSCRIPT

Hello and welcome to today's webcast titled Don't Get Caught Off Guard Using AI to Prevent Surprises. I'm Aaron Larsen, executive editor at Power Magazine. I'll be moderating today's program, during which we'll hear from Janet Webb, director of Advanced reliability, with GE Vernova and Mazen Younes, senior director of product management with GE Vernova. The presentation is expected to last about 45 minutes and will be followed by a Q&A session.  
 
Before we get started, I'll run through a few housekeeping items. In the webinar platform, you should see a chat area and a Q&A area. If you are experiencing any technical difficulty, you can ask for help using the chat function and our production staff will assist you. To submit questions for our speakers, please enter those in the Q&A section rather than in the chat area. You can enter questions at any time during the presentation, and we'll answer as many as possible at the end of the program. Any that we don't get to during the live session will be answered via email after the webinar.  
 
Today's presentation will be archived on our server for up to a year, and future viewing will remain free of charge. You can use the same URL to reach the archive program as you did to reach the live program. PowerPoint slides will also be available upon request. A certificate of completion for professional development hours will be sent via email to every registered participant who attends. Before we begin, I'd like to thank GE Vernova for underwriting today's program. The company's generosity allows everyone to attend the presentation at no cost. So with that, Janet we have the topic of the webinar is focused on using AI to prevent surprises.  
 
So how can AI artificial intelligence and machine learning really prevent surprises and failures? Thanks, Aaron, and we're thrilled to be here with you today. I'm going to start with a use case, that shows an example of how our our machine learning predictive analytics can, can help, prevent surprises and give you advanced warning before there are issues in the plant. And this one is for our, you know, kind of the most modern and sophisticated, gas driven equipment, the H class. In this case, it was the thrust bearing temperature anomaly that was detected. So we have a, a service that we, we offer, a managed service. And in this case, our services team was doing the monitoring, and they found this temperature anomaly.  
 
They looked deeper into the data after they received an alert and saw the trend. You know, steadily increasing over time. Quickly engaged our own GE Vernova product engineering team that that managed the gas turbine hardware and then connected them with the site, in order to take action. So in this case, a number of actions were taken in order to manage, the issue that they were detecting. They did not have to shut down, immediately. They had advanced warning, as I as I mentioned. So, they derated the unit until the next planned outage, and they changed some of those planned actions that they had prepared for that outage, including bearing inspections, having parts on hand and resources available on hand to perform the maintenance, and some supplemental inspections that they may not have done otherwise. And ultimately, they found significant distress on the bearing.  
 
What would have led to a potential six month downtime, which is a, a really, you know, severe event and, significant cost avoidance in terms of, you know, six months could be tens of millions of dollars in cost avoidance. So I wanted to start by sharing that example. This is one of many. I'll go through a few more today as we proceed. But I think it it shows clearly that our machine learning, analytics software is, you know, finding these, these issues and helping prevent, surprises, helping prevent unplanned downtime. We have a long history really with AI in GE Vernova. The product that I'm speaking of is SmartSignal. I'm sure many of you, are aware of, you know, have heard of SmartSignal. In this case, we GE Vernova purchased SmartSignal from Argonne National Labs in 2011.  
 
So so truly, for decades, we have been leveraging this, you know, proven and trusted machine learning technology. And I'll pass it to you, Mazen, to to describe some of the other areas of AI within GE Vernova. Yeah, absolutely. And I thank you so much, Janet. And, Aaron, thank you for hosting us. Great to be here today with the webinar with you all. So, yeah, I think besides what Janet is describing. Right. We have a long history in, AI/ML and definitely much of the products that we have, they can prevent surprises from a cost standpoint, right? From a failure standpoint as, as Janet says, we've had the, you know, AI expert systems like autonomous tuning and boiler optimizations.  
 
These basically look at, they're closed, closed loop, optimization system, that can basically, that basically use neural network models, right? Model predictive controls as well. For combustion optimization. Right. Reducing heat rates. So like we've, we've seen, real world savings with customers, beyond $300,000 a year in savings. Right? These prevent budgetary surprises for example. In addition to that, we have newer examples like autonomous inspection. This is a computer vision, image analytics, solution that is utilizing AI/ML models to detect issues with assets like corrosion, gas leaks, thermal heat detection. Right. It can tap it. It can tap into fixed and non fixed cameras. That uses basically a convolutional neural network. Right. And you basically, translate these images into time series data, give insights to these specific monitored assets. And you can basically reduce the amount of manual errors there, reduce the inspection cycle that you have. Right. And basically, reduce the time it takes for inspection from weeks all the way down to minutes. I think overall. Right.  
 
Like, when, when we talk about AI/ML being a tool of help, I think we've seen a lot of hype with Gen AI these days. Right. So naturally, our world got, heavy interest, focusing on the topic of AI, right. That that probably shifted that focus into that topic. I think the hype is great, but, you know, there are multiple proven traditional AI solutions that are out there, as you see on this slide, in the prior slide that Janet was sharing, that have been out there in the market and that have been bringing real value to our industry throughout the years. Yeah, obviously, GE Vernova has been using AI for a long time. What are some of the top ways that it can be used today to help companies? Yeah, I can start on this one and I'll, maybe make a couple points and pass it to Mazen as well. So we, you know, we've talked about helping prevent failures, getting early notification and in many cases, you know, days or weeks or even months of, of early notification allowing for maintenance, planning, parts planning, etc..  
 
But another one of the challenges that I hear a lot from our users, you know, and from power generators is that they, they struggle with the knowledge gap of people with a lot of experience, you know, 40 years experience retiring and trying to bring newer talent up to speed and fill in that knowledge gap software in general and AI more specifically, can really help close the knowledge gap, by not only having storage of that information and data and failure modes, for example, but also having the AI tools to help you access that information quickly. You know, in a way that that allows you to be more productive. Mazen anything you want to add to that? And I'll build on, the productive piece, right. On the productivity side, I think you're seeing simple tasks like, like text summarization, natural language query when you, when you want to get insights from a specific database can be an expert system or notes. Right. Can be your historical database.  
 
I think a lot of these reduce time it takes for employees just to find the information that they're looking for, right. It could be as simple as summarizing an asset history. Right within the asset performance management world or within SmartSignal. Right. It could be looking at historical failure modes that that asset has seen. Just give me a summary of what are the failure modes seen and what were the recommendations or steps taken to solve for that specific issue. This really puts information at the fingertips of users by just asking a simple question that can generate that information for them, which which goes a long way in improving productivity. I think the other piece, right, is the automation of certain tasks as well. Right. Like, in the, in the, in the field of inspection, for example. Right.  
 
You can, if you're deploying a robot to go and do the inspection and your rounds and taking pictures of certain assets, you're basically reducing or, you know, you don't need a field engineer to basically spend some travel time there and take manual image captures, spend the time to upload that image back, right, into the system, that whole workflow would be automated. Now, the field engineer's, job is shifting from, you know, the mundane task of taking a picture to actually thinking about you know, how to service a specific asset. And I think the other piece is safety, as well. Right. In the same spirit as talking about inspections and going on sites, I think certain, employees may be, prone to dangerous areas, right? Exposed to the dangerous areas. Now that you have robotics, right, Right. That can or drones or, right. So you can even use satellite imagery today to, to, to look at, transformers and other things. I think these are all, tools and technologies that are helping companies today. You know, be more responsible, safe, productive.  
 
And as Janet says, there's an aging workforce out there. You build the, expert system around that. You preserve that knowledge for your new workforce coming in in case they need to refer to things. And that can help with preventing failures. Proactively. Okay. I think that there's going to be a poll question that, we're gonna we're going to roll out. The poll question is my organization currently is not allowing third party software with Gen AI Yes or no? Give people a little bit of time to answer that. And the answers are coming in. Looks like about a half and half. Split a little bit more on the no side than the yes, but, certainly interesting how different organizations, allow or disallow this type of technology to be used. Yeah. And we can certainly touch more on that. You know, particularly in the gen AI space, the level of maturity of the technology, and kind of how we're thinking about it within GE Vernova. Okay. And I know many customers have invested in clean data processes, but a lot of them don't necessarily know what to do with the data. How can AI help in that regard? Yeah. You know, starting with, with good data is really the first step, you know, being able to, to collect the data and have it readily available and have good data, you know, a model is only as good as, as the data that it's built from.  
 
And so if it's if it's not good data, your outcomes will not, be beneficial either. So it's really the next step to me is really about, leveraging that data in these types of AI and ML models, to turn insights into actions. So we're showing another example again in the H class equipment where, and again, this was our our services monitoring team, received alerts. So it's, you know, we've got a model built from this good data. The alerts come in. They've they, you know, can trust the alert as they investigate further. And in this case found, you know, an unequal in the guide vein and very, very variable stator vein electrical actuator. Loading. And we're able to again reach out to the site and, you know, work through some, some recommendations on planned maintenance, you know, ultimately refurbishing and reinstalling and going back again to that model, after the fact to ensure that the actions that were taken did correct the issue that they were seeing.  
 
So that's kind of, to me that the next step is it's going to that, from insights to actions. The other thing that I want to note is we've, you know, I've heard working with some of our users or people who are interested in starting to adopt software that, you know, as they're just getting started, they've got the data in place. They don't they don't feel like they're ready for AI. You know, we don't have a data science team. AI is is, you know, too, daunting of a of a software for us to start with. And really what I would say there is in some ways, you know, AI like what we're describing here, actually makes things easier. And also because we have the flexibility of the services monitoring team, often what we see is, our users, when they're new, we'll start with our services so that they can get that very quick time to value, a model can be built with a minimum of two weeks of data. And when you have an expert who's doing that monitoring for you, you'll see that time, time to value very quickly.  
 
And then most often I would say they they plan for a transition period over, say, a year or two, where their internal teams start to learn more of the software and then they transition to either a lower tier of services or in some cases to to fully self performing and monitoring themselves. So those that flexibility and those options are there. I think we've got another poll question that we wanted to roll out. This one is about the benefit of adopting AI for my company. Is it clear? Yes or no? Give a little time here. Again, is the is the benefit clear for your company when it comes to adopting AI? It looks like most people think yes, they understand where the benefits are.  
 
So that's good to see that they recognize the value, at least in, you know, just getting approval perhaps to to implement is the hardest part. So. And I know many people are also concerned about AI and and how it may take jobs. What are your thoughts on that? Is there risk to jobs when when they implement AI projects? Yeah, I can I can take that one Aaron thank you. I think it's a normal concern. Right. Whenever we hear of like, new capabilities that can synthesize data, automate tasks. Right. It's a normal concern to have. I do think about it differently, though, right? I think these technologies are basically heroes, productivity boosters. I don't see them replacing human, the human aspect of things. Right? They can't they can't think or act on their own. Yeah. That's right. That's the it's it's, it's the far fetched concept at this stage at least I think the, with generative AI and I think more importantly, agentic AI, which is probably what, what triggers people to be concerned. We're hearing about, you know, means of improving day to day tasks right there. The way I view it is that it's taking away non value add work and time we spend on certain tasks, and just repositioning us to think about things or, or act around certain outputs from these AI tools differently. Right.  
 
I think the, the picture says it. All right. And the title of the picture, human in the loop is very important whenever we're designing and like, that's, that's the concept we always use in GE as well, the human in the loop aspect when we're designing these AI tools. It’s basically at the forefront of, of how we design workflows. Right. So if we're automating a workflow, you always need the human inside to ensure that, and acknowledge that the output is correct. Right. And ensure that tasks are safely and correctly processed, that we're in a highly regulated industry and environments. So like you always need that human aspect and oversight and experts and SMEs to be there as part of the process that you're building. At the end of the day, I'll give you an example. Right. Like in our monitoring and diagnostic centers, we'd have an AI agent, for example, that's disposing an alert, right.  
 
As true or false, the monitoring engineer right can basically acknowledge the disposition of that specific alert as correct, or they can override that agent to disposition it appropriately. Once done, the agent can go and trigger the creation of the case and auto generate the case with specific recommendations to take actions on and fix the particular issue that came up. Right. Now the monitoring engineer can approve these prescriptive guidelines that are generated and basically ensure that, there's, you know, there's also a feedback loop, right? If the agent needs to improve that output, there's a feedback loop where the the human in the loop, the person who's taking action can give feedback back so that system improves the output the next time.  
 
So I think from my perspective read the human in the loop concept is crucial, because it's, you know, AI basically, at the end of the day, as a means of improving the user experience, and people are becoming more focused on tasks that are important for them on actually applying their subject matter expertise within that automated process. Right. Where we keep a human to say, yep, take the next action, carry on with your automated process. I know many power companies are facing extremely thin budgets, and it makes it difficult, when they're trying to come up with decisions on software investments. So what is the return on investment or ROI for predictive analytics? And and what can customers expect from from implementing these types of programs?  
 
Yeah, that's a it's a really common question. And I'll I'll start by saying that, you know, we recognize that it can be overwhelming the number of products that are out there in, you know, various areas, and actually spending time with our, our sales teams early in the year, as they kicked off the year, a lot of conversation was around, you know, a power generator may be overwhelmed by the APM suite of products, for example, and as the experts internally who have seen many customers find success on, you know, different journeys of adopting the software, it's important for, for us to be advisors in that space to understand and, you know, listen to the the existing situation of someone who's considering software, and be able to make recommendations and point to, you know, successful use cases where other people have done this and found a lot of success. So in some cases, we do see, you know, somebody wants to adopt a full suite, across many sites. But I think that tends to be the exception, where it's more common that we would say you know, SmartSignal as a use case.  
 
Let's implement SmartSignal in one site and see the benefits and learn through that adoption and change management process, and then be able to apply that to multiple sites or consider, you know, additional products in that space that that would add more value. So, so it's a it's a journey and it's one that, that we can help with. Certainly. I'll walk through a few more examples here. That kind of address that question of what can we expect in terms of dollar savings? You know, as we consider the adoption. So this one I wanted to show, you know, I showed a couple of gas turbines, but, you know, smart Signal will cover across the balance of plant. Any OEM in terms of manufacturer and actually across multiple industries, including oil and gas, mining, petrochemicals, etc..  
 
So it's a very versatile tool with a catalog of what we call blueprints that have these failure modes built in, for all these different types of equipment. So in this case, looking at, a generator example, I think there was maybe it was the previous slide. There was also an HRSG example. So you can just see some different pieces of equipment where, again, the software finds an anomaly. It sends an alert to, to the team whether it's our monitoring team or, the end user. And those can then be actioned. And as you can see through some of these examples, the other thing I'll point out is it's it's not a single tag. Typically it's not a single sensor that's showing an alert. It's a, it's a multivariate, solution. It's you know, analytics built around indications that are coming from a number of different sensors. So you might see a change in temperature. And that could be an early warning that might come in at a, at a lower severity. So again these alerts are are prioritized in terms of their severity.  
 
And then when they see a different sensor change maybe that's an indication of something that that could be more severe. So these alerts that are that are coming in and the, you know, multivariate nature is allowing you to put those pieces together, within this prebuilt blueprint with, with these failure modes, you can just jump to, sorry. The next slide. And this is kind of a more holistic, you know, so I've, I've shown specific what we call catches, but this would be an example. And this one's published. So you can you can read more about it, Verdantix published this. This was a user who had 12 plants. So, you know, a large fleet and saw, you know, an increase in complexity in their assets.  
 
They were having challenges with availability. So they implemented in this case, not only SmartSignal, but also our thermal performance solution. And and with those two solutions combined, they were able to see over a two year period an increase in availability of 2%, which is very significant. As well as $5.5 million US dollars in savings per year. And that's the combination of fuel savings from, from the thermal performance improvement as well as what we've been talking about with, avoided costs, leveraging our predictive analytics. And then the last thing that I'll touch on here in terms of, you know what, what can you expect to get out of this investment is, you know, Mazen talked a little bit about safety earlier. I also want to draw this connection of reliability to safety. It's not one that we necessarily think about. We think more about you know the improved availability and not having surprises. But I think intuitive intuitively when we think about doing reactive work, it makes sense that there would be more safety issues when we're under pressure and under the stress of knowing that, you know, the company is losing money the longer that we're that we're in this state of being unavailable. And so that that actually shows up in the data as well.  
 
So, it's interesting, you can see here the trend of, you know, doing more predictive and preventative maintenance reduces safety issues, whereas doing more reactive work increases those safety issues. And, you know, not surprisingly, it's it's the most likely person to get injured is the maintenance technician who has less experience and is doing this in a reactive state. So to me, that's that's a really huge value as well to think about predictive analytics also having that positive impact on safety. Yeah that's a good point Janet. You know it's always hard to quantify events that don't happen. But you know and safety is a good one where somebody doesn't get injured. There's a real benefit to that. So what about companies that don't adopt AI? What sort of consequences could they face?  
 
Yeah, I think, you know, it's it's it's there's a few things that you've heard through the examples from, that, that Janet showed, you know, that there's a risk of higher operational costs. You see, slower decision making, potential availability disruptions. Right? These these cost money for companies. Right. And at the end of the day, it's it's it's not just the adoption of AI. You know, it's it's the it's one the adoption of the tools by let's say power generation companies. Right. Oil and gas companies, companies and other industries that may find value in that. Also, the vendors need to make sure that they're baking in these solutions that are rendering the products to be more, reliable, to render the product to be more useful for the, you know, for the task at hand. So I think, you know, in general, we're even seeing higher energy costs. You know, if it's a company that doesn't have the ability to do you know, generation optimization, switching between specific energy sources, between renewables and gas at certain, you know, peak hours or off peak hours, right. How you manage that? There are tools out there, generation optimization planning, even.  
 
We have some tools out there that that do help with, you know, with, with, enabling these or unlocking some of these benefits for companies. So I think, I mean, it's, it's the adoption of AI as a, as an enabler. And then the tools themselves that are built by, you know, expert vendors or like SMEs that are going to help with that. And I think at the end of the day as well, we must acknowledge that probably in our industry, it's a highly regulated industry. Right? So we need to, that we need to operate in, like safety and compliance right around the adoption of AI is at the forefront of, of where we need to stand. Right? Things like the EU AI act that we must ensure that we're complying with. Right. Ensuring safeguarding, our data, safeguarding customer data and making sure no IP leakages happen. These are all very important, measures to put in place. And I think, it I would say that, you know, it's safe for companies to adopt the technology slow and do it, do it doing it right, then going faster, not adopting it at all. And I think, yeah, I think we wanted to also, have a poll question as well here for our audience. We'd like some insights as well from Right now on this question. It's about the expectations for 2025 this year and 2026 next year.  
 
And, attendees can select all that apply and how you plan or what your plan is for adopting traditional AI technologies or newer technologies. If you're still gathering information about AI and machine learning, or have you already adopted traditional AI technologies and, already adopted newer AI technologies such as gen AI? Or not looking into it at this very moment, but, hopefully planning to start looking into it. It would be a good option. And we'll give some time to answer. You can select as many from the list is, applicable. It looks like a lot of people are still gathering information, which is probably why they're on the webinar, learning about some of the options that are available. And some have already started adopting. So it's, good to see. All right. You can keep entering. There's pretty broad split through all five of the categories. A pretty equal split, although still still in the gathering phase. Seems to be one of the the higher options right now. I guess my my last question for you is why GE Vernova?  
 
What sets your team and your products apart from others that are working in this sector? Thank you, thank you Aaron. I think you can definitely read the bullet points there, but what I would say is, you know, we're continuously evolving our solutions, right? I think with the introduction of new tools like the computer vision tool that I was talking about earlier, the autonomous inspection tool, the prescriptive recommendation aspect that we brought into our, you know, SmartSignal tool has been around for a while. But we didn't have that prescriptive recommendation aspect. And it's right then I think, that coupled with our ability to have flexible deployment options. Whether it's on prem, whether it's in the cloud, I think is, is, gives flexibility to our customers in terms of picking and choosing the, the right solution that would cater to their use cases and needs. And I know, Janet, you have a few, few more, things on that one. Yeah. Sure.  
 
I, I agree, you know, certainly on the, the ways that we're investing in AI, when I think about Smart Signal in particular, but really even more broadly, when we talk about generative AI and, you know, kind of the hype around it that that Mazen mentioned, it is a newer technology. I know if I go out, you know, just on, a web search and use Gen I, I have seen wrong answers come back. So there's, there's, there's reason to be hesitant with that technology, you know, today. But we're but we're still continuing to do proof of concept and pilots, and partnering with industry experts in that space so that when as that technology matures, that, that we're ready for it and, have the ability to, to adopt it in our software. But knowing that in general, we're always going to make sure that any AI/ML that we're adding will add value and not add that level of risk that takes away the value.  
 
You know, and also when we talk about generative AI, the, the cost, the data cost of, you know, data storage, data query, etc. is so much higher in generative AI than it is in, you know, a typical, search agent, as an example. So those factors have to be considered in as we develop the solution, because otherwise, again, we wouldn't be passing on value if we were passing on so much additional cost, that it that it made the solution not worth it. So I would say with GE Vernova, we are very keen on adopting the right level of technology at the right time that you can have confidence in. And, you know, with something like SmartSignal where we're using this internally as well, having our own monitoring team. And so we have, sort of a testbed of, you know, a group that can use this and prove out the technologies before they would even go to a beta opportunity for our, our end customers. We also have within GE Vernova on the power generation side, their biggest monitoring and diagnostics center is, is leveraging APM software as well. So, you know, we have we have multiple ways that we can kind of prove out these solutions. And one that all that I'll talk about briefly that I'm very excited about, that is in, you know, internal trials currently internal beta testing, if you will. I talked about, you know, the alerts.  
 
And they come with with an alert priority. We're also looking at machine learning to then give you, percent likelihood of, of an alert being a false positive so that you can even further prioritize. And we're testing this with our internal teams today. So really exciting new technology that's coming. And then again, you know, Mazen mentioned the flexibility, I think having our services team helps enable that flexibility. Also having just the deep expertise that’s built into, you know, our library of, of blueprints that include thousands of failure modes. This has been built over time. And it's something that that we have always seen as something that differentiates really on equipment where we have a deep hardware expertise in addition to software. But I think from my perspective, I mean, what gives me confidence is the fact that you guys have been doing this for so long.  
 
I mean, you said have been really incorporating AI for over a decade. AI has come into the general mainstream when ChatGPT came out and and if all of a sudden you decided, oh, we got to get AI into our programs because it's a new buzzword or it's a new thing that that everybody's trying to do. That would give me a little more pause. And the fact that this is just a natural evolution that you guys have already been working on for a long time. So I think that, is very encouraging and I think exciting for, for the industry. So, We have had a lot of questions come in. I'd like to thank again, Janet and Mazen for providing such informative and and interesting information. I'd also like to point out that there are four items loaded in the Webinar Handouts tab, so you can access them and download the handouts at any time. And be sure to do that before you log out today. Or you can come back in the on demand version and get them there as well.  
 
In the meantime, we'll get into the Q&A. We will consolidate similar questions and answer as many as possible in the time that remains. Any questions again, that we don't get to due to time constraints will be answered individually via email following the program. So at the top of our list here, we've got question that came in and it says is there a progression from the old gen aid monitoring service that was offered on generators in the late 80s? Is that part of what's kind of evolved into this new AI technology? Yeah. And I can I can start on that one. So I'm not actually familiar. I’ve been with GE Vernova and GE prior to that for 19 years, but but not in the 80s. So apologies. But but but no the truly the the monitoring has evolved. And there are I think still existing. You know particular to GE Vernova generators. A program around that specifically where you could additionally have Smart Signal and typically so I can I can certainly speak to this in the area of the gas turbine where, the, the same engineers who are involved in the development of the hardware, are behind the monitoring in the in the analytics of the gas turbine.  
 
There's a service that is, that is in, our contractual service agreements. And in that case, and I believe this would be true with the generator too, but can can confirm, leveraging SmartSignal on top of that gives you an additional layer of, early warning. And there's there's a difference in the kind of inherent nature of that when it comes to the hardware development and that expertise. It's a knowledge of failure modes that are, that are particular and understood well by having the data for the entire fleet. So, you know, GE Vernova has data for all of the GE Vernova equipment that we monitor as a fleet. And that's a particular sort of type of monitoring. Additionally, what we do with SmartSignal, because it's advanced pattern recognition, we can see other anomaly detection as well. And that's why Smart Signal is effective across GE equipment, non GE Vernova equipment. Because it, it doesn't necessarily require a full fleet of data. You can you can leverage the patterns in the data in these empirical models or data driven models, that that have different outcomes. And so, you know, to, to our end users, it's less relevant who's doing the monitoring and how how we're, you know, looking at the equipment and helping manage it internally.  
 
There's because of that, there's a reason that today it's two separate teams using, you know, kind of different methodologies, if you will. All right. Thanks, Janet. And can you talk a little bit about the cyber security of the system and any necessary security needed to keep the data available? but also keeping it secure. Yeah, I think, from a from a security perspective and data availability in general, right where we're talking, you know, in a, in a cloud setting, for example. Right. We, encrypt any data, address, any data in-flight. All of these are encrypted. We have access controls, security controls, user based Rbac, role based access controls that are, implemented. I think that the question touches upon data availability. So within the cloud, we ensure high availability of our storage systems as well. Right. With data applications across, what we call availability zones. Obviously backing up the data and making sure that the data is safeguarded from any, disaster for disaster recovery and ensuring that's where, you know, we have point in time recovery enabled. For example, to safeguard that data. So, yeah, we employ, these tools and techniques, to make sure that that data is there available at the fingertips of, of our customers and, you know, available for the software that's being used, that's, accessing this data. And I think you probably just answered this question, in, in the answer you gave. But does the analysis you offer operate through a cloud service? Yes. So and I think this is probably also the question is geared towards SmartSignal.  
 
It's actually in the cloud. Yes. We host it as a SaaS based offering software as a service, that we provide we also are able to provide that specific, software in an on prem environment, in a customer's specific, data center. Right. So, so, both are, fair game. Fair asks from customers. Okay. What instrumentation do I need to use AI predictive analytics software? Yeah. This is a I think a pretty common question that I've heard too. It's we would we start with, a services team member looking at the existing instrumentation that you have to understand what's there, and then they can help explain, you know, how much of the software in the modeling would be applicable. Often we can work with the instrumentation that's there, and potentially make recommendations to add instrumentation, but fully recognizing that that's not always a viable option.  
 
And so there is flexibility within the products to, you know, I mentioned the, the pre-built model, and blueprints, those tend to, to have some minimum requirements of instrumentation, but often most often that's critical on your very, you know, your high value but high criticality assets as well which which more typically you're going to have instrumentation for your turbines, you know, your generators. There is flexibility that if there's limited instrumentation, models can still be built with the implementation that's available, in order to monitor with, with the tags that we have. So there's flexibility around that. And we're actually, starting to look at further future enhancement. That would include, more ease of, if you have a team of data scientists, you know, their ability to, to build and add additional models. So we're, we're just, starting to get into some discovery sessions in that area, connecting more into the space of data ops and MLOps for those for those familiar.  
 
And if you're not familiar, you know, it's interesting topic. Okay. Thanks, Janet. I guess here is how accurate are these predictive models? I don't know if you can answer that or get into it. Yeah. It's, you know, I talked a little bit earlier about, you know, the model is only as good as the data. And so, if there's not good data available, it's hard to build a, you know, a model that's going to have accuracy, where you do have a high data availability and, you know, processes for for having clean data, you can build a more highly accurate model. I think part of the evidence, too, is in those examples that we have and your account owner or your customer service manager, or you can certainly reach out to us. We can point you to that library of catches that we've seen, where you can see the results of the models providing, you know, highly valuable output. Okay. I know, security is obviously a big issue.  
 
We've already touched on a lot of the, security issues. I don't know if there's anything to add, but another question about AI and security. Do you have any anything else that you can reassure people with? I think was probably is a good aspect to tackle. This is, the emergence of large language models and, you know, new attack, a new attack surface. Right. When you think of the OWASp top ten, for example, of large language models, and for those not familiar, that's basically the top ten security risks. You're looking at things like prompt injection, right. Training data poisoning, model denial of service. Right. And these are not any different than a DDoS attack type. Right. Or, like trying to trick the system, basically. So I think, when we think about these, it's building guardrails, right? Like prompt guards.  
 
Right. How do you guard and safeguard against, particular, prompts that a specific user is trying to have you basically set these guardrails, basic authentication authorization. Right. Goes a long way in all of this. At the end of the day, some of these newer technologies as well, they're accessing your APIs, safeguarding your API as well as part of the the work that that's that's put in place. So yeah, I would say, you know, yes, security is a risk, but in partnership with the cloud vendors and you know, with and even for on prem, right, the on prem use cases with our customers, we ensure that the proper guardrails are, are set and put in place. So we, we definitely, take this very seriously and, the emergence especially of AI, new AI technology, I think, brings us to make sure that we're adding these new security layers as well, in place. Very good. Does your system automatically connect to existing monitoring systems, or is it necessary to upload the data like trends to the cloud?  
 
I would say, depends on the use case, right. If you're looking at real time data. Right. So more of a vendor event. Yes. You can plug into an existing system, and the like, if it's in the cloud or if the customer gives access to it from the on prem, you can stream and read that data through. And then the other use cases, if you're requiring more of historical data, we do, actually connect to these, monitoring systems. You know, it could be a historian. It could be, CMS system, an EAM system that may have your information as well for the asset. We do ingest that into our cloud to make sure that our software has access to it. Okay. Thanks, Mazen.  
 
Can the software work with other utilities besides the electric generation, transmission and distribution, such as thinking of water supply and distribution, sewer collection and treatment, roadway monitoring and maintenance, life, life remaining, things like that. Yeah. In this case, we don't we don't particularly have, advanced predictive diagnostics in, in these spaces. But what I would say is that within the suite of APM products. So APM is asset performance management. There are a number of ways that we handle, moving from reactive maintenance, for example, to condition based maintenance, time based maintenance. So there's, Rounds, for example, that you can, that you can do with, handheld device and, have that incorporated into the same software suite. There's also, you know, asset, what we call our Asset Health Indicator that will give you, you know, a broader picture of asset health and some other reliability tools as well. So when you talk about life remaining, we do have, the ability for our users to develop their own reliability analytics in terms of life prediction.  
 
So thinking of Weibull analysis, if that's familiar, but basically using statistical methods to say, you know, we've seen even with a limited amount of failure data, maybe you've only got one failure point, but you have a lot of points of equipment still running, what we would call suspensions. In that model, you can actually build these statistical models with with very limited failure data to predict, a lifespan of a part. And there are some other, some other tools within that functionality as well around root cause analysis, managing the the analysis and documenting throughout the process and that workflow of doing a root cause analysis, so that the APM suite gets fairly broad. And I would say there are tools within that that would that would span across many industries.  
 
And I think I know, Aaron, where it looks like we're out of Q&A in the chat, but there was one that I accidentally deleted. Instead of clicking the the answer button, that I wanted to bring back up. And it was, asking for the root cause in the bearing issue that I, that I noted in the first H8 gas turbine example. And I just wanted to share that, unfortunately, that's not information that that I have the ability to share. Knowing that that's, you know, there there are things that are more, proprietary in terms of, those investigations. Okay. I guess one question I have, from past experience working at plants, managers very often like to be able to access data from off site, you know, and dig into some of the trends and things like that. How easy is that to do with with these types of, solutions? I think, you know, if, if, if again, these are either it's an on-prem or a cloud solution.  
 
I think the access to the to the data is going to be available. As I answered already, one of the questions there, if it's, depending on the use case that you need, but, overall, you know, it's, it's a simple login into the system, based on, the path you choose, if it's in the cloud. Right. It's a SaaS based delivery. You log in, your access, there are analysis tools that are available within the software that we provide that you can do trending chart thing. You can look at the historical, trends of, you know, whatever the use cases, you can do some BI as well, visualization as well within these tools that we have. So I think the access is, is, depends on the path that you're taking. Right. If it's an on prem, you go within your, within the route that is established within your cloud company.  
 
Everything is federated at the end of the day, whether it's cloud or on prem, over to the, to the customers, IDP environments. So you would just log in, like, you would log in to any other enterprise application that you log into today. All right. Well, I guess, we've got a little bit of time if you'd like to have any last words that we haven't talked about or things that we haven't shared. Just to thank you.  
 
For me, it's been a real pleasure to join you today. And feel free, to follow up with us anytime. Mazen. Likewise. There, plus one to what you said. We're happy to address anything, you know, outside of the the webinar as well. Feel free to reach out. And, Aaron, thank you so much for being an awesome host. All right. It was good. Good session. I think provide a lot of great useful information. So I hope people gained some insight from that. With that, again, I'd like to thank GE Vernova for underwriting today's event and thank everyone for attending. We hope you found the presentation beneficial and hope you have a great day. Thanks everyone.