How To Create Intelligent Digital Twins Using Xmpro Ai

Welcome to our comprehensive webinar hosted by Gavin Green, our VP of Strategic Solutions, titled "How to Create Intelligent Digital Twins Using XMPro AI." We invite you to join us on this...

Transcript

Welcome to our comprehensive webinar hosted by Gavin Green, our VP of Strategic Solutions, titled "How to Create Intelligent Digital Twins Using XMPro AI." We invite you to join us on this... hello everybody and welcome to our XM

Pro AI

um for intelligent digital twins webinar

my name is Kevin green I look after

strategic solutions for XM Pro I want to

thank you for your time attending today

if you've got any questions please send

them through I'm not trying to answer

them at the at the end

in some prior webinars we went through

the four pillars of the the excellent

product I'm not going to go into detail

here

but just in a continuation of that which

pieces are we going to be focusing on

this is in line with our i3c framework

and it's in essence where we focus our

effort when we're putting the product to

Market and the different feature sets

that we are are working on

today's Focus however is going to be on

the AI side of things and what is in the

product that can help you for the

intelligent digital twins with the the

focus on the AI site

before we jump into you'll hear me

talking around intelligent digital Twins

and digital Twins and there is a

difference between the two of those

there was a paper that was written on

the evolution of digital twins

um the two fathers of digital twins

Being John Vickers and the second being

Dr Michael Greaves

um this is Dr Michael greve's vision of

the different stages of evolution of

digital twins these are some slides from

the paper the paper is available down

the bottom and if you're interested we

can send you the link where you can

access the paper as well and it outlines

the different stages and it's also used

as inspiration and a guide within our

software and what it is that we do the

the main Evolution steps you'll see them

on the right there going from zero being

traditional 2D

that evolved into transitional so you're

number one which was 3D cat that went

into number two around the conceptual

this is where

um things started becoming a lot more

model based and then that evolved into

the concept of a digital twin platform

the step number three this is where most

people are at the moment

and they are slowly moving towards the

the number four which is around

intelligent digital twins one of the

applications for intelligent digital

twins is front running simulation it's

by no means the the only one it's just

the one that's front of mind which helps

explain the intelligent aspect of that

as well

when we start comparing digital Twins

versus intelligent digital twins there

are certain characteristics of a digital

twin and then there are characteristics

of an intelligent digital twin I've just

built the slides here again I'm not

going to go into into detail here the QR

code in the middle will take you to a

video which will go into a lot more

around the digital twin capabilities and

characteristics and the intelligent on

the other side

the one piece right at the bottom is the

anticipatory or the front running

simulation side of things

keeping with the the XM Pro product

within our intelligent digital twin

suite for those of you who are familiar

you're familiar with the data stream

designer the DS the app designer which

is the ad here and the subscription

manager of the SM here

the new one which you may not have seen

is around the AI side

rounding out the intelligent digital

twin Suite here

we'll go into that quite a bit of what

I'm going to walk you through you can do

without and I'll touch on some of those

areas where you can and then also touch

on the areas where you do need the the

newest addition here which would be on

the on the AI side

we have to stop in at the tip of the

iceberg where does it slot in so when

most people look at our digital Twins

and this is true for normal digital turn

platforms all the way through to

intelligent digital turns as well as

most people look at the top and they see

the uis there's actually a lot of things

that sit underneath

which is where the recommendations come

in and this is also where we slot in the

AI aspect as well the AI integration the

notebooks as well as how you

operationalize this for the agents you

still have everything else as you are

used to you've still got your data

stream and you've got your subscription

manager as well

we're going to leave the slides here and

we're going to jump into the actual

software itself and we're going to go

through a few examples of what it is

that I'm actually talking about here

so what we put together is we created an

app which will just allow me to walk

through the different bits and pieces

that we've got

let me just get rid of my slides

and we'll come back to to this

so again this is all driven from an

overarching framework which is what we

call our i3c

um and it's broken up into a few

different areas specifically with an AI

Focus for today we're talking around

Innovative embedded and augmented but

how do they all fit together where do

they all work together

we like pictures so we like explaining

how all of these fit together

as it is right now without having to

upgrade any accent Pro you can do

everything on the right here we call

that embedded AI

from the data stream designer how do you

bring the models in maybe you've already

got some models how do you

operationalize them and use them

everything else on the left here is some

of the new capability that we've brought

in with regards to Innovative AI as well

as augmented AI it's not required you

don't have to you can choose to use this

or not use this as well

so as we go into these different areas

what do I mean when I start talking

around Innovative AI we start going into

Excel Pro notebooks and there's two main

items in here the one is around the

ability to discover

so you'll see on the left we have a new

tool and that'll open up what we call

accent Pro notebooks

within here for those who are familiar

and comfortable you'll notice a familiar

environment being a Jupiter notebook

that we've got here

so this allows you to discover work

through and come up with the different

models Etc

any of the libraries that you can use

within a jupyter notebook you can use in

here this particular example I've got

here you can see this is around

predicting and the the beverage industry

using linear regression

as I build through this particular

notebook here you can see I'm going to

import certain libraries and it's going

to walk me through the different pieces

here so if there's a library that

I need for my machine learning or I want

to do some data Discovery I can load it

into the notebook not a problem at all

the other thing that as we work through

this that we've added into this is a

generative AI capability as well

what we mean by that is as I work

through this you'll notice at this point

here we've integrated chatgpt into this

environment as well so what that means

is you can actually use the chat GPT

to help augment and enrich what you're

looking for in here this example that

I've got in here what we're actually

asking

um chat GPT to do is to write me some

code on how to visualize the data that

I'm looking at I'm not an expert in how

I put this together so we're going to

ask and it'll actually generate

the items for me and if I go and run

that now you'll notice that it's going

to run and then it will generate and it

once it comes back it'll actually show

me the plot for the data that I'm

looking for

scroll a bit further down you'll

actually see what that looks like

if we keep going you can start

developing different models as well

if I come back here step one is really

what we call notebooks which allows you

to interact in an environment which is

familiar to you

you don't have to use them in here if

you do use them in here you can start

using some of the other features and the

bits and pieces that we've made

available as well if you have your own

models and you just want to use them and

plug them in by all means you can do

that already in the steps here

the second piece is really when we start

wanting to deploy models so there's two

main things that you can actually do

with the model

the one if I go into a different example

here is I can actually run this

it will generate the model I can help

discover that and the end result is I

can actually save this model to a file

that I cannot use inside the data stream

you can do the same thing right now

let's say you have a jupyter notebook

that sits outside X and pro you've

already doing Discovery you're already

generating models how do you just use

those models and bring them in you can

do that and I'll show you in a data

stream as well

the second thing that we're doing in the

prior example though is we're actually

fitting this model and passing it to a

repository and the reason for this is we

want to introduce a concept of

governance

and make sure that there is a framework

that you can actually deploy into and

that your models can live so that you

don't end up with a an Excel of machine

learning model sitting everywhere no one

knows it's got the latest version of

where it's actually sitting

if you have your own repository by all

means you can actually hook into and use

that as well this example here we're

just talking to to Mr flow as the

repository this is just one example it

is not the only example so if you've

already got one by all means you can

just plug that in and actually use that

as I continue to step through all of

this what it's doing is it's working out

the model it's working out the pieces

and then it is actually publishing that

model to my ml flow repository for me

you'll see it has already existed it's

going to created me a version 5 of that

model and now I can go into my actual

repository and decide when I want to

deploy that or elevate it to production

or not

so if we jump into that and again this

is just one example that we've got it's

not the only repository you can see the

one that we've just pushed through now

so number four is the one which is

currently in production I cannot decide

that number five is ready to go and I

can transition it to staging in a

controlled environment versus working

out well which is my live running model

which one isn't and how do you in a

governance structure move between the

different Frameworks that we're actually

looking at here

number four is currently published and

I'll show you now where that is actually

published and running

so if we come back here we call that

Innovative AI so if I come back to the

diagram that we're looking at here this

is the area that we are working in at

the moment is how do we take training

data how do we ask for specific results

that we are looking for and how do we

run that through a notebook you can't

augment that with chat GPT in this

example and then deploy that to an ml

Ops platform of your choice

you don't have to deploy it you can just

take the actual raw artifact which is a

result of that model and then use it in

a data stream as well

the

preferred or the more governance driven

process is to actually push that through

a

repository because then you get

everything that comes with the

repository as well

I mentioned you can do things right now

you don't need to upgrade or get all the

newest toys in the Box to to be able to

do this this is around the embedded AI

side so if I go into and this will open

up my data stream designer for me

you can do this right now you can drag

on this example has got a python runtime

running here which is running a model

this model was created inside the

notebook it was outputted from The

Notebook and then manually put into the

data stream and used and this is

capability that you can do as I

mentioned right now we call that how do

you use it from a manual perspective

what you'll also notice is there are a

lot of other items in the library

so if python is just being one of them

but let's say you want to use a more

standard algorithm like anomaly

detection or binary classification or

clustering

those agents already exist you can just

drag them on and use them right now

there's nothing stopping you you do not

need a repository you do not need to use

the Excel Pro notebook to take um

advantage of that particular capability

however if you do use a repository

and you actually want to integrate and

make use of that inside the data stream

what we've got here is we're performing

the analysis so this is using the Mr

flow agent so on the left you will see

here is a list of library

one of those being the ml flow

the Mr flow is talking to the repository

and here you can see the actual model

that it's talking to and everything else

is governed as well so I don't need to

remember the URLs I don't need to

remember logons passwords Etc I just

want to use that particular model and

actually just run it and execute it

again this is your governance that sits

around it it makes sure that the models

which you're using to do anomaly

detection Predictive Analytics maybe

you're using it to do forecasting or to

do front running simulation or just to

do normal simulation of current state or

past State you want to make sure that

you're running on the correct model

latest model and that you have control

of that you can do that inside the data

streams that we're looking at in here

so again if we come back

Innovative AI this is a really around

the ability to discover

go in through the particular notebooks

and then how do you use some of those

libraries in there the other example in

here is you can actually just create

models that do simulation so you don't

need to create models that do very

sophisticated

um AI number crunching and algorithms at

output you can do something as simple as

just simulation which is what this

particular example here is doing we're

creating a network we are then deploying

the same model to the repository again

you don't have to

There's an actual example of this one

running just the python itself with the

simulation on top of that as well

this one here is doing a remaining

useful life as well as a more advanced

prediction on the data coming in the one

that I just showed you is doing more

simulation so it caters to to both

there's no

you can do one and not the other

if we go into the augmented side

let me expand that a bit here so

embedded again you can do this right now

there's nothing stopping you from using

the current set of libraries and we're

expanding and adding to those all the

time

ml.net are some of those that you're

seeing in there ml flow when you're

talking to repositories maybe you want

to use some GPU enabled

um

algorithms from the the Nvidia Library

you can do all of that right now within

the embedded side of things

the Innovative is around the notebooks

but augmented is not just in the

notebook section over here

the augmented

AI you can actually use inside your apps

when you create and configure them as

well

what do I mean by that oh let's go into

an example

so I've got a particular app here and on

the right you'll see some

recommendations have triggered

these recommendations you can see

remaining useful life has been declining

and I can go a bit further down and I

can see remaining useful life has been

predicted to be below a certain

threshold again these were running

through the data stream and they can

either be Standalone python or any of

the other algorithms or read the matter

or repository and execute them as well

if I drill into that particular asset

it'll actually filter out my

recommendations for the two that I'm

interested in across the top you can

actually view the state of the model and

you can view the state of the data

that's flowing through the model so

actual versus predictive how are we

doing are we getting better we're

getting worse

if you apply this to simulation you

could have a button here that you could

click and it can actually show you the

progression of a simulation as it's

running through the uh the model itself

these recommendations here

you'll see above that there's a thing

called copilot copilot you can actually

bring into the application and use it in

here so if I was to ask it a question

let me just grab my prompt

and paste that in there

I can engage and talk with it and ask it

anything around this particular asset

that I'm interested in

this is using currently chat GPT so the

the data that I'm looking for here is

not sensitive in that regard however we

do have some OEM

Partners who have gone and taken as an

example Azure open Ai and they are using

that on their own data their documents

their manuals and they've deployed that

internally as well

this is just easier to to demonstrate

and show the concept of how you can plug

this in and where we are busy plugin the

different bits and pieces in so even

though this is using a chapter GPT API

here you could switch this out and put

azure's open AI in here as well a lot of

problem to do that

you'll see the the recommendations

coming down

you can go and expand that further and

actually pass these recommendations into

the assets themselves so here you can

see a discharge exception and I can

actually see that in this particular

Unity view as well I've got more space

for my co-pilot

um but I can still see the relevant

information that is applicable for me

if I drill into these recommendations

and I'll drill into this one as an

example and then I'll open a few others

as well

this is your quick start for

recommendations that you've configured

what I mean by quick start

so let me go all the way back here and

open up a different example for you

so when I say quick start you have

certain capability that is available out

the box you get it when you install and

you can just use it there's some other

capability that you can actually

configure within the platform as well so

we have a lot of different widgets we

have a lot of different applications

that you can just import we have a lot

of different templates

one of those templates is around

recommendations so you'll see at the top

I've actually got two sets of

recommendations on this particular page

for the Top If I go into my one example

this is the recommendations that most

people are familiar with XM Pro would

actually be looking at

this is the the Quick Start

you can't really change anything here

from an end user's experience or even

from someone who's configuring it you do

have some some control over the form and

maybe the triage instructions here but

that's about it you can't change any of

the layout you can't add anything to

this particular form maybe you want to

enhance and bring some other data in

some other capability in here as well

so how do you actually do that

if we go down the bottom

you can actually take all of that data

and make it available in different views

that your end users are interested in

consuming

same data presented a bit differently

but what this allows us to do is we can

now bring in let me hit my other prompt

there we can now bring in the copilot

into here as well and this is around dry

gas seals I can see my event data for

the event that has happened I can see

analytics so how often is this actually

happening is there something that

happens at certain times is it happening

when other events are happening as well

so we're dealing with a pressure seal

that is low we may have some high

temperatures maybe all of this is

related and there is a correlation

between these as well

however if we go a bit further down you

can bring the copilot in here as well

so now you've got a co-pilot which you

can deploy and use on an actual

application itself at an asset level or

maybe all the way up at a landing page

for your different views or you can have

a view where you can have it all the way

down with your recommendation data as

well

so now as an end user I can decide and I

can work through and work out based on

what I'm looking at based on what I've

asked and the responses I am getting do

I need to create a work request can I

capture any notes what do I actually

need to do with this recommendation that

I'm looking at here

and again this recommendation was

triggered from within the data streams

you'll see there is a run recommendation

if I go into the other example you'll

see there is a run recommendation as

well so we apply the same patterns in

our data streams irrespective or if if

you're using a model that is coming from

a repository

in this case this one is coming from a

repository or if you're using some of

the standard inbuilt

algorithms are already there or if you

just want to go and run your own model

and actually use that in here as well

they all follow the same pattern the

output of that you can pass to

recommendations and Trigger

recommendations as well because that's a

key thing for us is making sure that we

can close the loop on any of these

events that we find it's one thing to

just have a model it's how do you

operationalize the model and make it

useful

with the outputs that are coming

this view here

you can then look at the data you can

view any of the model information and

now you have a lot of different options

and how you want to react to that

if I close that

and we come back to this particular view

here

so to go through the the items that

we've just gone through

we talk around Innovation AI

embedded AI as well as augmented AI

embedded AI you can do right now

we have the libraries we have the agents

if you need specific agents for specific

algorithms they're pretty quick and easy

to create they need to talk to a certain

repository you don't want to use the one

that I've just shown here maybe you've

already got your own

those agents can be created very quickly

and deployed and used

on the augmented AI site

you can right now bring in

if I open that up again you can bring in

things like co-pilot into your

applications right now there's nothing

stopping you there's there's no

capability that you're missing to do

that the only thing that I would suggest

is make sure that whatever it is that

you want to use here

um you understand the privacy concerns

with the data again we are using chat

GPT here we are busy I can't show you

the the things on the Azure open AI side

because that's typically trained and

running on corporate data or customer

specific data for their customer

so this you would just swap out same

capability exactly the same mechanism

it's just what's sitting behind that'll

actually change for you as well you can

do that right now

The Innovation piece here this is where

you need access to the excellent Pro

notebook here

to be able to run and configure what

you're looking at we do provide if I go

back into the router we do provide a

quick start as well

the quick start for if you're not

familiar with all the different

capabilities the markup that you need to

use and how you can configure this a

quick start will actually walk you

through as well it's available as soon

as you have access to the AI designer

you'll find it in the list here and it

has a index and it'll just run you

through all the bits and pieces you need

if you're interested in some of the

other examples that we've got you by all

means reach out and happy to happy to

discuss how we can share and learn from

this as well

okay let's see if we've got any of the

the questions we do have a few questions

that came in there so first one does XM

Pro have models

um great question we get that quite a

lot when people see all the different

pieces and capabilities we don't have

specific models that we make available

we're not a data science scientist

company

however we do have certain algorithms

that we do make available so for

instance anomaly detection regression

um

classifications Etc what we do do is

give you the vehicles and the tools to

operationalize your own models whether

those are coming out of a repository as

we did with this piece here or whether

those are coming out of a data stream

and you go and configure everything

whether it's in python or even if you're

more comfortable in our script and you

want to use that you can do that as well

so we're more vehicle that allows you to

operationalize versus do we have a model

that does X on this asset

second question

um I did touch on it a few times but

just to to go through it is I have my

own models and repository do you I need

to choose yours being XM Pro

uh or can that you can keep your own you

do not have to replace what you

currently have you do not have to swap

that out and use ours you can just

integrate from ours into yours that'll

mostly get done inside the data stream

from the agent's perspective over here

again if there isn't an agent that will

connect to your repository of choice

it's pretty quick and easy to actually

create these types of Agents we can

create them on your behalf quite a few

of our partners do that as well or we

can even show you how to create these

yourself as well

the last thing I will leave you with

is just before we wrap up again thank

you for your time

um to today and we do have a webinar

coming out next month as well where

we're going to be talking about the 4.3

release

um the QR code here will take you

directly to the registration page if

you're more comfortable you can type the

link in and go from there if you have

any questions

um you can email me directly or just

sales at actionpro.com and again thank

you for the uh the time today and have a

great rest of the day

foreign