This is Darren Coil.
He loves adventures of all
kinds.
His most prized possession is
his grandfather's rock
collection and his heroes are
his parents.
He's helping transform supply
chain business technologies at
Microsoft.
What I love about my job is the
pace we are able to deploy a
really interesting
manufacturing technologies to
our own factory at a pace I
have never seen outside of
Microsoft in my 20 plus years
in manufacturing technologies
Microsoft as many people are
unaware is a manufacturing
company.
We build Xbox, we build
Surface, we build Surface Hubs,
HoloLens, keyboards and
accessories.
The first big challenge was
understanding the way the
supply chain works at Microsoft.
Everything from the sourcing
the plan, the make, the
delivery, the care, the return,
and the logistics. So just
understanding the scope and the
magnitude of our own supply
chain.
What are we doing in our
factories today what
opportunities are there to
improve the way that we build a
product?
What do we know about the
product? where the product is?
the quality the product, the
speed of deploying products?
There's lots of things in
manufacturing that are
important.
Manufacturing tends to be
fairly conservative in the way
they adopt technology.
In our case we have lots of
pockets of data.
So we had an engineering
database over here.
We had a SAP ERP system over
there.
We had the manufacturing
execution system at the
contract manufacturers place
all these different types of
all these different locations
of data.
The first problem we're trying
to solve is how do we answer
questions about our business?
For example, if a device came
back and we wanted to know the
history of that device: when
was it made, where it was made,
where did it ship to, who
activated it, why did they
return it?
Just the one serial number
would take us days to go to
each one of those different
data sources pull the data put
it together in a report and
answer one business question.
If you want to keep up in this
market, you've got to answer
these questions faster. You
can't spend your time getting
data,
bing it to the forefront.
answer one business question,
and then go figure out what the
next business question is that
you want to address.
That was the business problem
we were trying to solve.
Let's go connect all these data
sources in one location.
That way we can answer any
question that we may have --
today it's a serial number
tomorrow it's a quality issue.
The day after that it's a
sourcing issue and we don't
have to go spend countless
hours curating, manicuring,
stitching together data.
A year ago we took our
manufacturing operations on a
digital transformation, and
that transformation we broke
down into three waves: a
connected wave a predictive
wave and a cognitive wave.
And the connected wave was
really the first step which is
there's lots of data sources
all around the world get
connected to them all.
We're just connecting to the
data we already have that
turned into a trip to China.
We spent about a month there
and we connected to a dozen
different data sources.
They look at productivity they
look at yield.
They look at outputs.
They look at repairs and
inventory levels.
So productivity is data that
comes from our contract
manufacturer.
They use a Oracle based ERP
system.
And so what we did is we gave
them a data contract fairly
simple flat file format that
says here's the different
fields that we need.
We helped them with a script to
extract the data out of their
system, and then we basically
push that data to the event hub
or two blob storage.
So from the event hub or Blob
storage that then moves into
our Azur data like.
Our partner data flows to us
across the Internet using
encrypted packages into the
event hub or into Blob's George
depending on who's sending the
data and what they're sending.
Our product data flows across
police line also encrypted into
our storage containers and then
into your data lake.
For our network design.
It comes down to the data that
we're streaming off of the
location. what does it mean?
How important is it? what's its
time sensitivity?
For example, we just connected.
Process Equipment at the
factory.
And so we're using a IOT
gateway to azure connection to
stream data live into the
factory and then turn that data
back around to make real time
decisions, so the round trip is
seconds.
We just took the basic things
that they use everyday reports
that they trust and we
automated those reports. We got
them into Power BI we've got
the data moved into the cloud.
we got some basic analytics
behind it and brought all that
data back to the factory.
Our first deadline was to get
an operational control room for
the factory.
The four person team went out
there and in six weeks we were
able to automate their standard
reports.
They looked at everyday
productivity, shipments, yield
quality built him a 10 screen
power be-I visualization room
where they could look at all
the data live all the time.
We did all that in a six week
period before the CEO came out
to see an entire digital
transformation of the way that
that factory was looking at
data, and presenting data, and
reviewing data.
We had all that insight in the
slicing available Power BI
immediately because we were
connected to the raw data
source.
We could answer questions about
what happened yesterday or the
week before. We can look at
line to line comparisons. And
all these things were instantly
available and instantly
answered with power BI in the
factory which is what got the
wheel started.
Since we began a year ago,
we've connected to our primary
factory, we've connected to a
dozen of our vendors, we've
connected to our delivery
mechanisms that we connect to
our customer service mechanisms,
We were able to do all this
over the last nine months.
That's in a connected phase.
The predictive stage is kind of
where we are now where the data.
Lets us see trends as they are
occurring real time and respond
to them we can dispatch people
to the floor.
We don't have to wait for an
excursion event to happen.
We can see supplier data coming
in we can make decisions about
how much to build, where to
build, where to ship based on
all this data coming to the
surface real time because we
don't have to collect the data
anymore.
So for all of our automated
test machinery we did all of
the statistical grading and the
back end so that when it came
into Power BI it had already
had a sort of a ranking as to
whether or not this piece of
data was important or not,
Which then goes into the heat
maps which allows us to find
the data quickly on the machine
level data like a lamination
machine or a trimming machine?
All that data gets
statistically granted actually
in Power BI itself.
So we brought it all the way
through and then we developed
the statistical process control
rules in Power BI And so it
executes SPC on the fly.
So we have both.
The cognitive wave.
It's now there for us 24 by 7.
We've had dozens of
conversations with these
manufacturing operations that
see the same thing they laugh
the same way they say, oh yeah,
we run our factory off of Excel
we run our factory off a
PowerPoint, and we have the
same challenges. It takes us a
week to answer one business
question. And we show them in
five minutes and we bring up
Power BI and show our factory
and we're like look, I can tell
you why I miss production today.
Oh, I am short parts from this
vendor and I have a bunch of
stuff stuck in repair, and we
show them the power of
visualization layer and.
the pace that you can answer
questions and then we go back
and tell them again take
advantage of all of the machine
learning and the AI and the
intelligency insights and let
that then drive your business.
The cognitive wave is where we
allow the machine learning to
solve complex problems for us
and we focus on manufacturing
operations and supply chain
operations.
The difference between
predictive waves and cognitive
waves is more around the
fundamental technologies behind
it.
Cognitive is more about using
machine learning artificial
intelligence.
So to get started we presented
the problem to several data
scientists.
We gave them all of the data
streams. that streams came from
fact information, such as
process yield coming off of the
machinery.
We gave them dimensional
information, and this component
came from this physical
location,
The order in which things were
put together, so that the
patterns could then look for
relationships and causality and
create a better optimal
solution for how we built the
Hololens.
Large big data platform
analysis to solve complex
problems not necessarily things
that are occurring at this very
second but things that are
occurring over longer periods
of time this sort of analysis
by a human would take weeks and
weeks and weeks, trial and
error lots of computation,
Matlab and those kind of
computation programs to come up
with the same answer.
But by using machine learning
and doing the pattern
recognition we are able to come
up with the answer in just days.
For example optimizing the
material maximum minimum
material conditions of a kitted
device.
Fact information from what was
happening,
Dimensional information in
terms of where things are, and
what order things go together,
it's like a fax sort of a
construct.
And from there they then were
able to use the machine
learning pattern recognition to
come up with the optimal way to
assemble the Hololens, which
improved our yields.
So these are the kinds of
machine learning algorithm
things that will let us get to
a cognitive in a faster pace in
our manufacturing operations.
The big feature for Cognitive
in AI is to tackle problems
that haven't been addressed yet
in manufacturing simply because
there is not enough data
available not enough
computation not enough pattern
recognition to be able to do
these things.
You can do other examples where
you look at the way that you
build a device and you feed it
design and experiment and it
can generate better ways to
assemble a device or it will be
able to predict what will
happen in the future if you use
a component a component B.
So far Microsoft has reaped a
number of important benefits
from the continued digital
transformation of its
manufacturing operation.
Darren said the team has
learned some important lessons
as well.
I think the things we could
have done differently we should
have kept the dedicated team
dedicated made this their full
time job.
Of course we didn't know that
it was going to have such an
impact we didn't know that the
value was going to be there.
This digital transformation has
been the largest change in
manufacturing technologies in
30 years and it was probably
one of the easiest changes to
bring to fruition the actual
physical part of connecting the
data and bringing the data to
Azure -and making the Power BI
It was actually very simple.
Like I said, a couple of weeks
to automate some of those basic
reports and a few more weeks to
connect to a bunch of machinery
and bring all that data live.
The number one thing keeping
all these companies from doing
this which was the same thing
for us which is business
adoption and change management
that it's the fear of if I go
invest in trying to go on a
digital transformation will my
business accept the answer to
that is that it has to come
from the top down.
When our vice president said
we're going to do this?
He was relentless for the two
or three months that it took
for everybody all the way down
the organization to believe the
data to see the change and to
get on board.
Once you get that change
management started the flywheel
begins and then it perpetuates
it feeds itself.
The beauty of the architecture
is we've moved from systems of
records to systems of records
and a system of intelligence on
top of that. We're not changing
the fact that the data still
exists on the machine or the
data still exists in an ERP
system,
We are simply moving all of
that data up to a system of
intelligence.
We've seen value in
productivity gains,
people not collecting data
anymore.
We've averted product loss in
the tune of millions of
millions of dollars.
We've optimized operations in
and around just the data that
we're getting from the insights
the payback the value to the
implementation, is measured in
days and weeks not years.
In our 10 part video series
expedition cloud, Microsoft
technical experts will share
the inside story of Microsoft's
journey to the cloud including
proven practices,
Things we do differently and
advice for customers on the
same journey towards digital
transformation.
For more infomation >> Contruction Project On 2nd Street To Improve Safety For Cyclists And Pedestrians - Duration: 1:26. 
For more infomation >> Suspect Arrested For Shooting Up San Jose Elementary School With Shotgun - Duration: 0:46. 
Không có nhận xét nào:
Đăng nhận xét