Some of the work with energy
airfoils actually their advisor for

further up the through
the final exam right now.

So they're probably having less fun but

so Dr Ahmed asked me to
do the really quickly and

basically he highlighted a couple things
I talked to him about about this team.

First he said the project itself
just sort of kept getting bigger and

bigger and bigger and bigger and bigger.

Or maybe they cared but
they never played and so

they just kept doing whatever they needed
to do and it turned out whatever they

needed to do involves downloading
research papers and reading and

understanding and coming back to him and
saying OK here's what we learned.

Here's a really good woman and

there's just the kind of things that you
don't necessarily do in a regular class.

You have no problem just going
out there and reading it.

Understanding and so
he was really impressed.

So I think that probably proximately
what he would say if you use your.

You know

well

some people mentioned we did
a project we're going to

work with me on this project where

Michael Chan pricing myself late son and
the team mates who are here are parked.

And it's all mental to give
you a little background about

your purpose which are multi-million
dollar she used for

generating power to produce of over
four thousand Air Force but we.

And there are a number of.

Different types of Air Force that sixteen
different types of turban that they rate

and the region size from
about three inches tall.

Although it was three feet tall.

These airfoils require
extremely tight speeds and

quality is extremely important for
their for their production.

You can see a picture of a turban
on the side of the production

process is laid out about forces
enter the production process

orders are pictured on the on the right
there on my right or left there and

they go through numerous steps before
becoming finished their files the most

notable of these steps are the inspection
process noted in blue and

more where they undergo coordinate
measuring machine inspection.

These are measuring machines trace
horizontal sections around each plate

inspecting over twenty
features in each section and

each of these each of these
scans in each of these features.

If there is one section
of the specification or

out of tolerance the entire fleet
is considered the fact that

there are three main problems that our
group pursued in our seniors on projects.

The first is a lack of manager
visibility managers when looking

on the leads being produced on their part
in their plant were unable to determine

when consecutive defects were occurring
until of the until multiple defects and

multiple blades had the screens.

This is a major problem for
them as in two thousand and

eight they believe that over two
hundred thousand dollars was wasted

due to consecutive
consecutively defective blades.

To start stop this problem.

We created a dashboard system
which allowed the US and

allowed them the managers
to detect early trends.

And start and consecutively.

The second problem we dealt with
was excessive inspection time

with an increase in demand and
increasing throughput through the.

That is required.

We we believe that the best
place to increase the throughput

is in the inspection times.

Currently the cm's inspect multiple
sections as I mentioned earlier and

we perform the correlation study using
the methodology found in the paper

mentioned by Dr Sokol to
eliminate redundant sections and

increase the throughput of these
operations without significantly

decreasing the quality of
the overall finished goods.

The third problem we dealt with was
an inaccurate in process tolerance levels

the tolerance for specification values for
the police produced sent to be

to be used in turbans was provided for
us by the target manufacturers' but

in process that values such as
those made after the plane for

Milt were not provided these remind of
estimated by the plant managers currently

and there is no good system for
calculating these values.

Currently there has been a lot of waste in
the plant where the plant would produce

and keep on working on blades that would
eventually be deemed out of tolerance

due to the lack of appropriate
tolerance values within the process.

We believe this is cause and

excess waste over a hundred eighty
thousand dollars doesn't make

the first problem is that there is poor
visibility within the production process.

There's two kinds of problems that
stem from the lack of this ability.

The first is that managers are on a level
twenty five machine shipped machine

these machines that produce the blades
constantly need to be recalibrated and

the managers are unable to know when
the machines need to be recalibrated

until multiple blades are produced
defective and they are then able to

identify when these blades are what
the root cause of these problems were and

go back and read help raise the plates.

The second problem creating a lack that
stems from a lack of quality visibility is

the inability to identify which of these
processes which of the many preachers and

sections for each of the many blades.

Scanned by the seat number sheens are the
most important and or the most variable.

These are all basically the process is
the need to be looked at more closely and

with hundreds of different kinds of
processes going through their plan and

if and at any given week.

It's virtually impossible for

them to look at all the different kinds
of plates and so there is going to be

consecutive effects because of
a lack of information on their part.

Again we believe that this is caused an
excess of two hundred thousand dollars and

the fact continued

the major brunt of our problem
here is primarily data aggregation

first Sheens originally
the machines output data into.

Discrete Excel files after each
scan after each blade scan.

They had no way with four thousand being
produced each week they had no way of

actually pulling together all this
data and trying to monitor trends with

I mean with four thousand different
excel files being produced every week.

We created a access aggregation software
that would allow them to easily identify

to easily pull together the data with just
the click of a button in under a minute

and a little easily identified which
allowed them to easily solve the right

next to the first ability that our access
suffered able to do was run the stache

board software the software allow them
to identify high purity high variance

processes by identifying which of these
processes these scores they could easily

pull down which part number of whichever
portal they wanted within a programmer and

characteristic and choose from a date
range and these built to populate

allowing them to easily see which And as
you can see in this example none of these

scores of lives in the process of
would be worth looking up any further

they could easily identify which processes
are the most volatile and which.

Further instruction in addition to
this dashboard system we also have

made them able to use control charts which
are things like using control charts.

They can process required for
that need to be looked at further and

as you can see by this control
chart this is just an example.

There is a mark out of control and
they would

after looking at the controls where
they would look back at the process and

be able to quickly determine what
part to be replaced in the machine.

If that was appropriate and
no one to last to describe the problems

are so far a project deals with the issue
that there is too much time being spent on

inspecting careful and there's a lot of
time because these therefore have such

strict requirements and so the problem is
that this is a bottleneck in the system.

And right now they can't
explain to people who demand so

they actually actually has to purchase
additional intervals from outside vendors

and that cost of one hundred fifty
thousand dollars extra a year and

demand is also expected to increase so
specially in the future.

There's no way that the current number of
machines is going to be able to keep up

and inspection costs one hundred
fifty thousand dollars so

it's a little expensive to
just keep buying cheap.

So what do we do well

to go about this problem too if we

reduce the time it takes to respect
each other and so how do we do this.

Well if you if you look at this picture
of this kind of gives you no idea of how

these reforms are expected.

They're split into four zones sections and

the machine basically traces around
each section and and so on like this.

And so.

So the best way to reduce the spending
time is to reduce the number

of sections that they inspect and
now but we can't just go around

removing any random sections so
how do we determine which sections we can.

Well this is where we use
a correlation methodology and

determine if this section
don't need to be checked.

So for example on here there's a high
correlation between sections A and B.

and B. only gives us back when it was out
of spec that maybe we don't need to check.

Maybe we'll check and that way we
can save a lot especially when so

but if we remove any number of
inspection points there's always

going to be some cost some loss of
the time and the idea is to minimize it.

While still gain the benefits
of reducing special ops.

So a little bit more detail

explanation of our methodology
is that on from a high level.

Basically we take all the inspection
points and we break it down into two

sections that we want to keep
inspecting the retain sections and

the sections that we want to remove
from the inspection process and

we do this for every single possible
combination that we can break the data

in to break those sections into different
retainer groups and there's hundreds of

thousands of combinations per different
type of Air Force so we do this and

we plug this into a linear estimation
model that uses the retained sections to

estimate the points from the roof sections
because if we are moving the sections.

Then we should be able to develop and
the song.

So how do we compare the different
combinations of the refinery sections

is that the dollars you produces an index
value which is basically a representation

of how much tension power losing by not
inspecting the group of new sections so

for example if an index
values five percent.

That means we're losing five
percent the touch of power and

that means that we should be able to
buy only inspecting retains auctions we

should be able to catch ninety five
percent of all the facts and so

we run this calculation for
every single possible combination and

it spits out in the next value
very possible combination and

we find the best group
of sections to reach

that minimizes that and
that's how we see what's which sections.

We can and so here is a table of routes

that show results from this now on for for

this particular example there's there's
a total of thirteen sections on

this are full and for each number
of sections that you want to retain

this is the combination of
sections that has a lot of size So

for example if you want to retain twelve
sections out of the thirteen then these

are the ones they should retain to
have the least loss of detection and

then this is for eleven sections and
so on and so when you graph this out

there is so on the X.
axis is the number sections retained and

on the axis is the index side
which is the last section

then you see that there is not
a linear relationship and so

if you don't inspect half the points that
doesn't mean you lose half the information

and so this is where the same times when
you can actually remove a lot of these so.

Actions and
still detect the good of the facts.

And so here here's an example if we

choose to remove if we choose to keep
only four seconds I was fourteen.

Then coming from this methodology the best
actions would be to change an arc and

so the inspection would go like this.

The only trees here and
here and you and there.

And so all this that the machine skips
that's where the time savings account and

in this case there.

That's a seventy percent
reduction in inspection time and

even though we have
that much time savings.

We're still of over ninety
eight percent of total US.

And so
bronze going to talk about a new way.

So it's miles away for easier for
further tests but in other words.

Therefore it is very worthwhile.

And we're seeing them stay accurate.

Also right now at
the finest demonstration.

Thomas ever was are provided by what I

buy at the after machine
stage every opportunity and

can potentially that can
magically tangible effective air

force has seemed quite possible and
resulted in ways to thank you.

And that process I understand has cost

around one hundred eighty
thousand dollars so

to tackle this problem we
perform a courageous study first

to get all our data from the after
process and the final process.

So you look at a graph and
write just one features we do a great.

All the final plan is very

we identify pairs correlational
seventy percent higher.

So she can see there is
about six pairs of features.

Now we perform a relation every So

how do you perform your vocal
relation by doing the right.

Hears every known heart

is the feature they were
trying to get right.

So we

can create a process
towards using a final.

So this process times as I said
before because there are six

pairs of seven percent.

Now we have six times
of whatever machine and

now the first is expecting
her second is a concern

for expected Torres pairs.

They have hired her own way and
for a conservatory and this

sort of project for Column one like a boy

and by creating a system say two
hundred thousand dollars for

example is in two thousand and
three second only

average of nine special sessions for
you know right.

Not mine.

Again.

So thank you

for having more accurate provide
a better understanding of how the facts

thousand dollars three thousand dollars

for

my inspection

before the final of the funds but
still expect everything to go.

But before that process that's where

we had and that's where the most

cost it was allegedly but
there is going to be additional costs

that don't go up and that it's still.

Or

are

very different

because if you

really are right next to each other.

You're going to be horribly expensive

or you explain it from a different.

A much.