Hang tight while we fetch the video data and transcripts. This only takes a moment.
Connecting to YouTube player…
Fetching transcript data…
We’ll display the transcript, summary, and all view options as soon as everything loads.
Next steps
Loading transcript tools…
OSIsoft: AF I Online Course- Modeling Approaches in PI AF: Top-down Versus Bottom-up - AI Summary, Mind Map & Transcript | AVEVA PI System Learning | YouTubeToText
YouTube Transcript: OSIsoft: AF I Online Course- Modeling Approaches in PI AF: Top-down Versus Bottom-up
Skip watching entire videos - get the full transcript, search for keywords, and copy with one click.
This content outlines strategic approaches to building Asset Framework (AF) models, emphasizing that a balanced, iterative approach is more effective than strictly top-down or bottom-up methods, and that value can be realized incrementally without needing a fully complete model from the outset.
Mind Map
Click to expand
Click to explore the full interactive mind map • Zoom, pan, and navigate
So far in the course, you've seen an overview
of the PI system and a discussion of what
AF is and how it can be used, how it interacts
with the client tools and pulls data together
from multiple systems. And we are in the part of the course where we are going to
start learning how to actually build elements to represent
assets and add attributes to those elements that holds the data for that assets.
But before we proceed with that,
I'm going to pause for a moment and discuss some of the different
aspects of building models in AF,
just so it's in the back of your mind as we proceed throughout the
course. And before you start building a model for your own system, you
might want to come back and revisit this video just to
keep these different ideas about how to approach
AF modelling in mind. If we are thinking about a hierarchical
model, there are two obvious choices that come to mind for that.
We can start at the bottom level and work our
way up the hierarchy until we get to the top. Or we
can start at the top of the hierarchy and work our way down until we
get to the bottom. Let's discuss the bottom up approach
first. With that type of approach, we would just right in
and start modelling our various types of equipment assets
and assigning all the data to them. So
we'll jump to another database that has a
complex structure, this New Green one. And we
expand this out until we get to the equipment level.
And you can see we have built out a boiler and it has
a lot of different data associated with it. And we templatize
that and we are able to start generating other boilers.
And we did the same for cooling
fans and heaters
and compressors and
pumps. And so quite rapidly through AF
templatization, we can start
generating a whole lot of asset models with a lot of
different data pulled in. And this is really nice to be able to organize
all this data around our assets. But the problem with this
approach is it lacks the vision. And also we have a lot of
assets with a lot of data and other than the benefit of
having organized our data and making it easier to find and use.
We aren't really sure where we are going to go from there
to really find our added value. One thing you
may notice is that a lot of these have some
attributes in common. And so if we just jumped in
and started building all these different equipment
templates out without really thinking through it first, we may
have wasted some time by specifying the same attributes over and over
for the same equipment. And now we've made it so that we can't really compare
across different types of equipment for like attributes. But if we
had done a little planning, I'm going to jump in the library
here, this one was actually done this way and you can see that if
we built a base unit template with everything that these various
types of equipment have in common and branched off from
there and started making individual
templates from the unit template, we could save ourselves
some time by not having to specify all of these
for every different type of equipment. And then we also have this
set of common attributes that we can compare all
the different equipment on. So the other approach
would be the opposite of bottom up and that's obviously
top down. There we might start at our
enterprise level and then go to a plant level. And you see we are
going to specify some very high level KPIs here like our
energy savings targets, environmental targets, quality
targets, reliability targets. And here we have some
PI tags feeding the actual measurements. But in
reality, those would probably be some kind of complex KPI calculation. And
we know if we want to get these, we are probably going to need some input
from the lower levels. So probably some similar
KPIs at the process level. And then to
feed those, we would need information from all of our
different assets. And so with that in mind, we can plan our
template inheritance for these various assets.
And we can plan out what types of data that we want in order
to feed the calculations like these KPIs at the higher level.
And then jumping up to even
the plant level, we have all of the same KPI
targets. And so this top down approach
has really minor disadvantages compared to bottom up.
It can get, for
complex hierarchies or large structures,
you can spend a lot of time and theoretical discussions about how you
are going to organize these hierarchies. And you may not
even know where you are going to get all of the data for some kind of
complex KPI calculation. And so you don't want to get too bogged
down in those theoretical details and prevent
yourself from getting down to the real work of specifying all of
the different data that you have for your assets. So a good
approach is usually a middle ground where we
spend some time sketching out this
hierarchy, getting an idea of how we might want to do some template
inheritance so that we have a like set of attributes
that crosses as many pieces of equipment
as possible. And thinking about some of these calculations,
and if there is any data that we would need from lower levels to
feed into those, what types of things we need at the equipment
level. But also going ahead and including other available pieces
of data for our equipment and not ignoring those that aren't needed for the calculations
because those can often be found to be useful later for
some other monitoring
purpose or some kind of new initiative such as
efficiency initiative or an environmental initiative.
So if you go ahead and build as much data as you can into these at the
start, then you have it available when you want to try and make
new calculations later. So a good general approach would be
to just figure out what assets you have and
think about how you are going to templatize
those and build an inheritance tree for those.
And we'll discuss more about template inheritance later, so that
should become a little more clear what I'm talking about. Then start to define your attributes for your
assets, thinking about some of the attributes that you
might need for higher level calculations. And go ahead and
configure all of those attributes. And then you can
start really adding in your calculations and your analytics to
flush those out and start finding the value in your system.
And then once you have a good set of attributes and calculations
built up, go ahead and move onto your client
tools and see how you can make some reports or
some sort of displays that can utilize
all of this new information that you've built into your AF model.
And one last thing I want to mention, I want to warn
against a common misconception about AF.
Oftentimes people see something like this
NuGreen model that we distribute as an example, and
they think, I have to have this huge all encompassing
enterprise-wide structure with highly
detailed assets with a lot of attributes before I
can really start using AF and getting value out of it.
And it seems like a very daunting
task. They think about the two hardest parts about the two different
approaches, from the top down level they think they have to have the full hierarchy
before I can really put this to use. And they think from the bottom up
level, they have to have all the different data
associated with my assets before I can really get going.
But I want to go ahead and put your mind at ease hopefully
and let you know that the modelling approach that we
discussed, you can sketch out a plan from a higher
level and you don't necessarily have to start at the very top. You
can start at the process level or start
with the particular equipment class. For example, if we
knew that our boilers are key to the performance of our
process and very much wrapped up in the profitability of our
process, then we might want to start modelling out our
boilers and figuring out how we can add visibility to the
boilers and find some new calculations
to improve boiler efficiency. And then go ahead and find value there
by approving our boiler by using AF
to improve our boiler performance. From there,
we can expand to another asset
class. Maybe we want to look at our fans or our heaters
next. And then we can start building up to our cracking process. We
know that we have some calculations that we want to use to
improve overall process performance. And we start small and build our way out.
All the time, keeping our higher level plan in mind.
The key here is you can start with something small like a
certain asset class or a certain subsystem within your
plant, and go ahead and build out your model for that
particular subsystem or class of asset. And start
finding your key indicators and
building them into the reports and displays that you are
generating with the client tools. And start
seeing some improvement in your process there.
So the idea that you have to have a fully finished all encompassing model
within AF in order to start finding value is
really not true. The truth is, if you are innovating,
then your AF model is probably never 'finished' because
you are always finding new ways to improve it and add onto it.
That's one of the beautiful things about AF, it's very flexible and very
configurable and you can always add new
pieces or change pieces around or add new
calculations and make it more and more useful as time goes on.
So don't be intimidated by something like this NuGreen
database, it's something that you might build up to eventually,
but you don't have to start with from the very beginning.
Click on any text or timestamp to jump to that moment in the video
Share:
Most transcripts ready in under 5 seconds
One-Click Copy125+ LanguagesSearch ContentJump to Timestamps
Paste YouTube URL
Enter any YouTube video link to get the full transcript
Transcript Extraction Form
Most transcripts ready in under 5 seconds
Get Our Chrome Extension
Get transcripts instantly without leaving YouTube. Install our Chrome extension for one-click access to any video's transcript directly on the watch page.
