2024 - AVEVA World - Paris - HMI/SCADA
AVEVA: AI use cases for HMI/SCADA
Session description coming soon.
Company
AVEVA
Speaker
Brian Leonard
Brian has held various positions within AVEVA over the past 20 + years in technical support, application engineering, and product management. He currently is Product Manager for AVEVA Teamwork, Reports for Operations, Integration Studio and Development Studio. After earning a Bachelor's in Electronic Engineering and Technology, Brian began working in this industry as a Systems Integrator in fields ranging from oil and gas, liquid chemicals, water/wastewater, power, and building management for data centers. In his spare time, Brian spends his time making new, interesting things with his hands, making craft beer, and has been known to dabble in world Dungeons & Dragons.
Session Code
AW24-HMI-D2-SESS-363
Transcript
Hello, everyone.
My name is Brian Leonard. Pleasure to meet you. I'm, one of the product managers at Veeva. Been there for about, twenty one years. Actually, yesterday, I think, was my my anniversary. So my my my tenure at Aviva connect is old enough to drink, which I guess means something. Right?
So going, AI use cases, it's sort of like a, AI and machine learning, I guess, from from a, HMI skater perspective. First thing we'll talk about is Vision AI, and this is a product that is out is has been released. Nathan Snyder is actually our product manager for that. Any questions, feel free to direct him to him because he's the he's the guy who knows.
But but if you're not familiar with this product, what it's designed to do is to be able to to interface with real world applications, real world products, and and and things out in the field to be able to detect whether or not there are anomalies of, an ideal situation. So you'd be able you start off with connect it to, hardware, cameras, things of that nature, lighting, all set up properly for particular, scenario. And as as a process runs, it can and it will interface with your applications, whether it's on my system platform, etcetera, and then be able to to alarm and get warnings based off of, situations that are out of the boundaries of normal operations.
So, just kind of an example, pass or fail, not ultra high speed, stuff, but sort of a generic, run rate applications. So, there are some applications out there that are, like, ultra, ultra high speed. This is sort of not that that level. It's designed for, relatively, modestly paced, detection.
So, here's an example of what that looks like, how it can be implemented within an application. So we have various implementation capabilities of it. You'll see here, so we have a process going on. There's an alarm situation.
You'll see that the vision assistant has generated an alarm. So we drill down, figure out where the process is. Select it there. You'll see that this particular tank has got an alarm situation.
Back over here. And you'll be able to go take a look at those alarms, see what what triggered those particular, alarm sets or alarms there, and go ahead, proceed there. You take a look and take action, based off of those particular settings.
Another, product we have here is, that's actually out in the field, been out for a while, is Aviva Predictor Analytics.
What this is designed to do is be able to take, near or real time data coming in, compare that with your historical data, and then generate your, anomalies and things of that nature, to be able to determine if there's any failures out in the field, if there's sensor data that's that's failing, to be able to detect these, discrepancies and these anomalies, before before problems occur. So you'll be able to schedule your downtime, fix the problems before it actually has costly results.
I'm sure you've already seen this before, the AI Vision Assistant.
This is in this is coming out, I think just already did come out. This is designed to take that language model, sort of your typical AI, applications, to be able to interface, interact with your applications and then give, help based off of language, models. One of the things we wanna be able to do with this is be able to interact with things like Teamwork. So if you got a problem with a particular piece of equipment or maybe some data is out of out of whack, you take a look at that. The interaction to Teamwork to be able to take action. So you could be able to work with the process itself, figure out any work instructions that are associated with a particular piece of data or whatever, and we'll go from there.
On the topic of Teamwork, one of the products that, I'm near and dear to me.
If anybody here not familiar, I guess, for a show of hands, people who are familiar with Teamwork or who already have it. Okay. That's about right. Looking pretty good.
We're kinda bringing the numbers up there. But what that's designed to do is to be able to interact, with people. A lot of the stuff that we're talking about is data. How the historizing data, analyzing it, collecting it, and all that that nature.
What teamwork is designed to be able to do is be able to work with your users, work with your people that are actually working out in the field and and interacting with with one another. A lot of the times what what winds up happening in these facilities is that you'll have these heroes in a in a in a place. If something goes wrong, you contact them. But when they're not there, when they retire and a whole new new workforce comes in, you wanna be able to exchange that knowledge.
Teamwork aids in that that that, transfer of knowledge that way.
And in doing so, you do things like create issues, identify issues that happen, people interact with one another.
Some people say, well, I could just use teams for that communication, but this takes that to a whole new level because it it it's the interaction of the people with particular pieces of equipment.
And so I think, Nathan, earlier if you were on a UNS, discussion there, we have some other enhancement. I'll kinda show that that a little bit later, but some of the AI stuff within teamwork that we're working on are things like the automatic translation. So in the tip case, let's say something goes wrong, this, you know, mixer is not working properly or a conveyor is not running properly, you would just basically have a call for help or an issue that's been created within the product. But some of the stuff that we're adding are things like automatic language translation.
So you type the problem in, something's wrong with this, if there's somebody else who speaks another language, it'll all be automatically translated for you, on the spot. Some other things that will be enabled, are intuitive search as well. So if something's wrong, but maybe a pump is not working or something like that, there'll be associated things that may may come up after, along with it that maybe can help you solve the problems that way.
So, on on the topic of teamwork, we have the ability to to take an application and transport that into Teamwork. So you have your system platform Galaxy, and and extract that those that hierarchy and then automatically populate or programmatically populate a Teamwork instance. So that'll really aid in the deployment of it and and and and make it a lot easier. So to take it a step further on an AI level, well, we have we're working on and this is in the labs.
So actually the the transformation to go from, system platform galaxy over to Teamwork, that will be released at the end of the year. So if you are an ops control customer and you want to try it out, let me know. We can get you fired up. But some of the stuff that we're also working on is to be able to extract, a hierarchical model off of a flat tag structure.
So we have, the we we imported the structure in from, plant SCADA and it's a flat model and what what, dev studios are started to do is it take that structure and then be able to try to interpolate, the actual, model that you it might may think that you're trying to use. And so, we do here there's several formats that it's that it's found on on the left side here And then what we'll see here on the right side there we are. Okay.
So we'll see that, we'll look at this first the various formats that it that it's been, calculated. We'll take out the first one and say, okay. I think I think it's cluster. I think there's a a particular, asset name over here, the AIT.
And but on the second sec over there let's see. That should let's take a look at the structure. On the seven one two one three two, it didn't really know what that was. So it said, okay. I don't know what this is.
We'll figure that out. Let's see. Take a look at that structure.
And we take a look here.
The cluster that's cluster in cluster one. So we'll say, you know what? Actually, this maybe add a little bit more smartness to this and say, okay. Well, instead of cluster and cluster one, I'm gonna join the two together and give it a name.
There.
That's that. And that name is gonna be called a cluster. It did figure out what that section was there, so it automatically applied a name to that particular asset.
But the seven the the the big number there didn't really know what to do, so we'll go in and try and define that, for it.
Here we are.
We'll break it up.
And this number actually is a particular sequence. This is just for our knowledge, we know that this is the the the actual structure of it, so we're gonna take that model and put it populate it to to have the AI learn how to do it.
So we are there. So find that up really quickly. We'll break it up.
So we know this section is this at, and then we'll, apply names to them. So when we make our asset structure, it'll automatically be applied that way.
There.
Zoom it in.
Alright. So they go to the next one. And then that one now they give it the first one. It's gonna re re, calculate it all and then take the things that it already learned and apply that to the the other formats as well.
So that so you see that it automatically decided to do that.
You join those two together, give that a different name, correct that, and then go to the next step. It recalculates, does all that. And but now that the rest of them see, on this one here, what we wanted to be able to do is, take that I think on this one here, what we do is we actually there may be some tags that you don't wanna have an create an asset. Maybe it's just a tag that will be translated into an asset. So you can just skip those as well.
Let's see. This one didn't calculate that properly, so we fix that.
Adjust the names.
Alright. And this is the one, I think, where we, just disregard it, ignore that format, and don't don't convert to an asset.
Alright. So now that that's done, we go to the next and configure the standards.
And so now that we've defined all of these standards and all these set points, we go through and actually create the asset structure.
So we take the ones that have been created, define the asset structure, and we and it'll also tell you if there's certain tags that aren't aren't a part of that. If you're missing some out on some, it'll, it'll let you know.
Create the structure here.
Sequence ID, and there we are.
K.
Alright. And this is, now that we've created now that we've created the structure, you'll notice that sorry. It it takes all of it, sets up the mapping, and process it that way.
Now that we've done that, we can start taking that and publish that to Teamwork.
Okay.
And within Teamwork, it's a little bit of a different structure. So, you can have a very wide structure set up in either a system platform or or plant SCADA, but then in Teamwork, it's relatively rigid. So you wanna be able to map those things accordingly. You maybe take an end to one ratio, but, with this, it makes it much easier because you just drag and drop and create your structure that way.
One of the things too that that this aids with not only the creation of the your your structure within Teamwork, but also links the the applications and links the assets to a system platform. So then if you have a your your OMI, you put a widget on there for Teamwork, the the the applications are already linked. So if you're looking at a pump within a your OMI application, when you bring the Teamwork widget in, it'll automatically, be contextually aware of it and ring up the pump that way. So things like work instructions or issues can be automatically, applied there.
Alright. So this is the creation of the structure, drag and drop and stuff in. If you've ever set up Teamwork before, this process typically takes a day, two days, three days, and now it's done within a matter of, of minutes.
But, also, the important thing is that linking. Alright. So we'll publish that, and then we're good to go. So we go into Teamwork, and you could be able to check it out and see how that structure's been created and then take it from there.
Alright. And this this path of the part the the the video, the part from the transferring from system platform to, Teamwork, that will be available later this year.
Alright. Let's take a look at the report. If there's any data that went wrong, you could take a look at it there.
Alright. And here we are defining teamwork.