Hello, my name is Behrouz
and here I'm sharing a part of my research about applying Data science
for the safety of process plants
the truth is, we still need them [Chemical plants]
They are there
And there are so many applications for them nowadays in society
but we don't want them to be like this
...A source of disaster for human being and nature!
So where all these incidents are coming from?
It is coming mostly from human error
What we can do about that?
Let me show some resources of data
in chemical engineering field
This is an engineering drawing called 'Piping and instrument diagram' or P&ID.
It is showing the flow diagram
Please note that this is human-readable
Machine is not able to automatically read data from this kind of engineering drawing.
Another one I'm sharing here that is called 'Mechanical datasheet'
it is full of data
but still it is just human that can read data in it.
And also we have engineering specifications and standards
that we have to deal with
and again, they are human-readable
it is not possible for a machine to understand them, except that the machine
do NLP or 'Natural Language Processing'.
So how do we design process plants?
and how do we decide during operation or shut-down and maintenance?
What we do is that we gather all the data
as you can see on the left side
we gather all these engineering drawings
, specifications and standards
and we send them to a designer, an engineer
or a person in operation field
to design or to make decisions
So that's a dream
That's what we think that happens
and we want them [plants] to protect our planet earth as well, while they are operating.
But that's not the reality
the reality is, there are so many incidents happening in chemical plants
and oil and gas facilities
and that's, as you can see, what's happening to human being and to our nature
This is one example
Happened in Huoston
23 people died and you can see on the table
the amount of money spent for what happened
so what I'm doing here
is to gather all the data
in something called the knowledge base
this knowledge base in AI and Semantic technology field is called Ontology
So what I'm doing is gathering all the 3D models, mechanical data sheets, line lists as you can see
and also the human knwoldege
that is in the format of human natural language
into a knowledge base
here is one example
of extracting and converting data from one engineering drawing called P&ID
to an Excel sheet or any tabular format
from there we are going to have spreadsheets, tabular format
then we are going to have data wrangling
because there are so many terminologies are being used
then we change that into ontology
The other part is natural language, as I have mentioned
as we have specifications in human natural language
what I do, is that I use something called Controlled Natural language
and from there, I can change that to ontology
Now at this stage
we have everything
from engineering drawings and natural language in one database called the knwoledge base
that we called that 'Ontology'
This is one graph database
resulted, or something that you actually can see from the ontology
This is just one bonus here
This is not something that we are looking for actually
and this is something, a side effect of having an ontology
having this, you can do a lot
but this is not the end result
what I'm looking for is the reasoning from the data
or from this knowledge base
As you know, for any reasoning
we need logic
and I need to have something called Description logic for that
I know it is probably becoming boring
but just bear with me for another 2 minutes
Expressing our human natural language
into description logic is not easy
It is not something that we can do manually
I have shown here 2 examples for that
as you can see
manually changing the syntax, as we know in description logic
is not possible
Here I'm showing you how we change the controlled natural language
into description logic.
I have used a platform called Fluent Editor
and I can change the controlled english, or controlled natural language
have the semantic rules in there, and put it into description logic format
using the syntaxes and having the right ontology
this is one example
I'm using another platform called 'Protege'
let me run this
So in this one, you can see that on the left side
I have my engineering drawing
and I have created my knowledge base
and this is the graph database
as I have talked about
as you can see, all the edges, all the relations
are having meanings
it means that for example I know an equipment like 'E-1007' is connected to a line
so in our case, "Connection" is actually being defined and the machine can understand
the meaning of the relation and the "connection" between an equipment and a piping line.
The example that I'm showing here is using 'Fluent Editor'
I have the same P&ID as you can see on the right side
and on the left side, I have my ontology
but this ontology is actually using Controlled natural language
So I have all these information about all parts of my engineering drawing
and I have defined everything in 'Controlled natural language'
and I was able to bring in all my information, all my logic, that I require to make decision
into Fluent Editor
and what we need from here
is to make decisions
so I'm going to have the query, or the reasoner, as you can see here
to make decisions
for example I'm saying : "can you tell me if my plant is safe if I have this kind of diagram"?
and it's telling me that it is 'unsafe'
because of the information it has in the knowledge base.
Thank you for your time! This is just a brief description about what I'm doing, a part of my research
and the way I'm using data science in that.
Thank you very much for your time.
For more infomation >> how to crack movie maker 2016 - Duration: 3:55. 
For more infomation >> Reunited, for Better or Worse - Duration: 42:20. 
For more infomation >> Woman arrested for stealing thousands from memorial fund - Duration: 2:56. 
For more infomation >> That's What Best Friends Are For/In Hot Watermelon - Duration: 22:40. 
Không có nhận xét nào:
Đăng nhận xét