2025 - AVEVA World - San Francisco - General interest
APM | Data Intelligence and Generative AI: Unlocking productivity, performance and profitability
Imagine transforming your industrial data into timely, informed, and optimized decisions to improve the productivity and profitability of your assets. In this presentation, we'll show you how our customers embraced the concept of a digital reliability center transforming their industrial PI System data into a strategic optimization solution for maintenance. The conventional approach to operations is siloed, slow, and disconnected. Data intelligence, supported by generative AI, is the future foundation of operational excellence and reliability. With a clear vision of profitability across all your sites, operations, performance and reliability stakeholders can troubleshoot in real time and quickly identify the best course of action to remediate issues. In terms of reliability and performance, you gain the ability to accurately predict time to failure and take action before major issues occur.
Company
AVEVA
Speaker
Sean Gregerson
Sean Gregerson is Vice President of Asset Performance Management at AVEVA, with over 20 years of experience applying advanced software technologies - including artificial intelligence, machine learning, and predictive analytics - to help industrial operators improve the reliability and performance of their production assets. He focuses on enabling companies to unlock the full value of industrial Big Data and AI. Sean holds a Bachelor of Science in Electrical Engineering from Bradley University.
Session Code
SESS-286
Transcript
So myself, Sean Gregor said, I head up AVEVA's asset performance management software business. I've spent the last twenty years in the industrial data intelligence space, along with the application of artificial intelligence and machine learning for improving the health and performance of industrial assets.
And the first thing that I thought we'd start with is just the challenges that we see our industrial manufacturing and critical infrastructure customers are really challenged with from a reliability perspective.
And so we see this maintenance of performance optimization challenge where seventy percent of operators are unaware when to replace or repair or upgrade their assets. And this problem can be a bit tricky because you can have very high levels of reliability, but often that can come at an unrealistic cost structure for the business.
And then asset failures, we know that eighty two percent of asset failures are random. And so what that really requires is us to leverage these new technologies that we have available to us today, leveraging things like AI and fusing that into our reliability process to prevent these random failures from occurring.
And we know that unscheduled downtime, this cost can vary from industry to industry, but on average, it's two hundred and sixty thousand dollars per hour.
But this doesn't take into account the damage to the equipment assets, which could easily run into the millions to men, tens of millions of dollars per incident.
And safety and risk management is at the forefront and center of any comprehensive maintenance and reliability strategy. And we could easily add to this workforce productivity, which is top of mind for most of the executives that we work with today. And software solutions like we're going to look at here can have this really massive scaling effect on the workforce.
And certainly sustainability, the more the higher levels of efficiency, performance and reliability that we operate our assets at, it's going to have a net positive effect on our business from a sustainability position.
And along with these challenges, this is what we often see is that many are operating in this disconnected environment today. They have these data connectivity silos, these stakeholder silos, these divisional and functional silos across the business.
And what this really results in is that people are not able to access the information that they need to make timely and informed and optimized decisions for the business.
And along with this, many today are struggling to progress in their digital transformation and in their digital maturity.
And a key contributor to this often being the case is that many are getting trapped by point level software solutions that are solving a very specific point level problem for the business.
And the way that we're addressing these challenges today from an AVEVA perspective is through the power of the AVEVA industrial software portfolio.
And on this bottom layer, we can see this industrial information management, digital backbone or digital foundation.
And half of this is based on AVEVA's PIE technology.
The other half of this is based on this technology that we've built that fuses together all the domains and dimensions of the industrial data across the entire asset life cycle from design to operations, to asset management and to optimization from a performance perspective of the assets.
And then once we have this Industrial Information Management digital backbone in place, we just plug in these pre integrated applications like asset performance to improve the reliability and performance of the industrial assets and to decrease and lower the cost of maintenance.
And then again, all of these applications are pre integrated to this single pane of glass, this unified visualization layer that sort of encapsulates all of these applications and all of the information within them into this single pane of glass to visualize this information. Now, of course, anything and everything we do from an AVEVA perspective is open agnostic.
It integrates with the existing software systems and IT infrastructure that you already have put in place.
And so when we look at this from more of a functional perspective, this is what we see and we have hundreds of interfaces to these OT data sources. And so this is where we're plugging into and connecting to all of these OT devices and all the sensor data and operational data. We're collecting that, we're archiving it, we're organizing it, we're contextualizing it. And then the other half here, we have all the other data and it has even greater complexity and more dimensions to it than the operational data alone.
And this is where we're connecting. We have hundreds of interfaces to connect all of the engineering data, the 1D data, the two d data, the P and ID drawings and diagrams, three d models of your plants, point cloud related information, the enterprise asset management, work order history information.
And what we do is we fuse all this together. So we're doing the data capture, the validation of the data, we're archiving all of the streaming data.
We have this data harmonization capability that fuses together all those domains and dimensions of the industrial data across all these different systems, across that entire asset life cycle.
And then we take a layered approach from a data analytics perspective and provide this data search capability.
And all of this is underpinned by this intelligent industrial data model and industrial workflow capability. And what it does is it unifies all the data and information across all these different systems, across that entire asset life cycle into this unified taxonomy that defines and creates the relationships on how all of these different items related to that asset are related.
And then once we have this industrial information management capability in place, we just plug in these different capabilities, Predictive analytics for using AI and machine learning in a no code environment for early warning detection and diagnosis of equipment health and performance problems.
Predictive quality, predictive production throughput, predictive energy consumption.
We have purpose built AIML solutions that solve these domain specific problems like monitoring your renewable assets, solar and wind assets, purpose built AI machine learning capabilities.
And then performance analytics, where we're leveraging first principle modeling of the assets, using physics based modeling of the assets to drive the assets to the highest levels of performance.
And then simulation and what if analysis. So we have these asset strategy libraries that we've built up over the last twenty years, twenty two thousand man hours of experience built into them.
And we use these asset strategy libraries to help our customers define a comprehensive asset strategy for their business across all of their assets.
And it has the ability to define the strategy and then simulate the strategy through Monte Carlo simulation, balance the strategy according to your business objectives, deploy the strategy, measure its effectiveness and continually improve it.
And then mobile inspections for driving consistent execution of work in a digital way so that work is being done the same from site to site, plant to plant for operator rounds, maintenance rounds, safety, environmental inspections, things like that.
And then now we have this new industrial generative AI capability where we're leveraging the latest in large language models and GPT engines to be able to interact with all of this industrial data in this very humanistic type of way.
And then the enterprise visualization for this unified single pane of glass that we can visualize all the information about the asset.
And as part of this, we're doing what we call hybrid AI modeling. And so this is where we're leveraging this first principle physics based modeling of the asset, taking the output of that physics based model as soft sensor inputs to the AI machine learning model. And what this is going to do is improve the fidelity of that AI machine learning model by taking the physical sensors and combining these pseudo generated soft sensors, if you will, into that model.
And what that's going to enable is the ability to provide these fault diagnostics, not just from a reliability perspective, but also from a performance perspective.
And here's an example of one of our customers that has implemented this solution. And they did this as part of their digital transformation strategy and program.
And the journey is what's the important part. We'll learn a bit more about this in a little bit here. But the journey is you don't have to do all this at once. It's not that I just dropped this entire solution in place, but it's about a journey that you go on with these components that just are pre integrated and plug into each other. They can go on at your own pace and in your own way.
And so here they had a multimillion tag PI System in place for many years, but they realized as many have that there's a lot of value in that data and how can we start to drive more value back into the business from all this industrial data and information?
So they put in that other component of that industrial data infrastructure to fuse together all the domains and dimensions of that asset life cycle, the design and operations and asset management and optimization data, and then put in the unified visualization layer and then plugged in these different capabilities from using AI machine learning for predictive monitoring of the health of their industrial assets. And then putting in the process simulation to use physics first principle based modeling of these assets to provide this complete solution. And in the end, the result of this is this unified data model that's linked back to the physical asset itself and its connectivity within the plant in three d.
And a big part of this is really taking the data that's inside many of these systems, often trapped inside many of these systems or oriented in a very textual, non intuitive way.
And we're transforming that into something that's actionable for the business, something that you can take action on to improve the reliability, lower the cost of maintenance.
This can then be presented from a three d perspective. So we link that unified taxonomy and that unified data model to that physical asset itself in the plant. We can do that in three d and this can be incredibly useful for troubleshooting all types of design issues, operations issues, reliability related issues, as well as from a remote work planning perspective so that you can really start to see, do I have any other early warning alerts? That's what this predictive technology is giving us is alerts weeks and months in advance of failure, not operational alerts that need to be acted upon immediately.
I can start to see, do I have any other early warning alerts in this part of the plan where maybe I could join or group work jobs together?
And as part of this, we now have this great generative AI technology.
And so what we're able to do is we're able to integrate contextualize this information and now leverage the power of large language models and GPT engines. And we're doing this through this rag orchestration layer that we've built and developed that understands the context of the data sources and the elements within those data sources that can help to answer the questions that are being asked. And we'll see a quick video here of how this works.
This is in the context of a wind farm operator that's having a problem with one of their wind turbines.
So what we're going to do is we're going to open the AVEVA Industrial AI assistant and we're going to start to ask it some questions.
So the Hornsea wind farm is operating lower than expected over the last And we could ask some more questions.
Do you have a asset utilization dashboard for this asset?
So I'll pull up the asset utilization dashboard. We can clearly see where this asset was running and when it came to this stop condition.
It looks like there was a high temperature event for this wind turbine. Can you find me all the bearing and related temperatures?
So give us a list of all the related temperatures here and also a link to a more contextualized trend view of this data.
And we could see clearly here where the temperature was continuing to increase and then the unit reached a stop position and starts to the temperature starts to decrease.
We could ask some more questions. Something caused all those bearing temperatures to go high at the same time. Can you find me the maintenance manual for this asset?
It'll give us a link to the maintenance manual. We can open up this maintenance manual. We could start to sort and filter through the maintenance manual, but this is one hundred and sixty four page document. And that would take quite a bit of time. So we can simply just ask some questions now of the maintenance manual. Does the maintenance manual have any information on high bearing temperatures and what actions I should take?
And so it will give us a summary insufficient lubrication, a malfunctioning temperature control valve, and it will give us some steps to take towards remediation of this issue.
And we can ask some more questions here. Do you have a three d model of the turbine showing the oil lubricating system?
Maybe we haven't performed maintenance on this part of the asset before, or maybe it's been some time since we performed maintenance on this part of the asset before.
And so what we can do now is see this oil lubricating system in three d, and we can start to remotely plan how we're going to perform maintenance on the asset when we get out in the field.
And then we can simply ask to write a summary of the conversation that we can include as part of the maintenance work order.
And then in this case, part of the maintenance team is located in Denmark. We could ask that this is translated to a Danish for the maintenance team in Denmark.
So it gives us a perspective of how this new technology can really be leveraged from a reliability perspective within the business.
And just to conclude, this is the problem that we're solving. Fifty percent of all the industrial data that we have available to us today was created in the last two years alone. And it's expected as we look forward two years from now, that this will likely still hold true.
And along with that, the challenge becomes that the study has been done and only seventy three or only twenty seven percent of all the industrial data is actually being actioned and put to use within the business.
And seventy three percent of all the industrial data remains unleveraged.
And so unless we start to take advantage of these new technologies that we have available to us today, these new capabilities from an AI perspective to translate this data into actionable outcomes for the business, then it's expected as the amount of industrial data continues to grow that the amount of unused industrial data is going to continue to grow in a very similar way.
And we're doing all of this to enable that connected worker, that worker that is at the heart of all of this, the reliability engineer, the performance engineer, the production engineer, the field workers.
We're leveraging the latest in technologies, the latest technologies that we have available to us today from the cloud and AI and augmented and virtual reality technology to really enable this connected worker with new, more effective, more efficient ways of operating and doing business and working.
