Very hard here.
First of all thank you all for coming out.
I see a lot of people I recognize see you
tonight in the semester a lot of people
don't recognize like periods
in your design your.
So thank you all for coming into the
family members friends who have to be here
since a lot of you might not know senior
designed this is basically the last
course that the people they work and it's
where they get to get to form teams and
go out to where you know there were
some real problems spending thousand
fifteen hundred hours over the course of
the semester and really fill up with.
Nice results based on all the things
that hopefully they've learned here
everybody refers to that and
we have presentations by the finalists
we started with twenty three teams
in the winter and the top four.
Teams from two teams to work with
North Southern Railroad berthing
Yes every year and
that's the the randomizer are there
so we're going to do is do that either
guys are to give a little introduction to
their team for the first teams advisor
Professor out to be here tonight.
So I'm going to introduce them.
So when you told me I could
be part of a couple of things
he said all those things you would expect
the finalists in that they were really
great to work with it were a bit hard
working all these nice things but
was really was really boosted interesting
to me is that the Institute for
operations research and
management sciences really rude.
Has just heard of the right
of a woman from India or
more interested in these really hard
problems that rolled her off and
so the problem that they came up with
the thing is really different of
the homes that everyone is really
almost easier to simplify a version
of exactly what the list was so
unfortunately the beauty of it through
your bust and are there of course
will work for the wires around the world.
The real root of the.
It's all about them.
So we all go.
Thank you Reza Good evening everyone.
We have an offer to southern Israel team
and meeting face to ferry sank was due.
They were Chen's MAPI pointing yet
on who tormented men and
myself about the level of our project
with nothing sudden was focused on
the optimization of their feeling policy
through the course of the semester we had
Dr Phil Gramm of the survivor not
examine is the fourth largest freight
railroad company in the United States they
operate with over ten billion in annual
revenues across about twenty two states
in the eastern seaboard of the country
they cover about twenty one thousand
miles of track within this sector
are projects specifically focuses on
five hundred routes of their systems
which are operated by about five
thousand three hundred locomotives and
their stations are we're looking to
work around our About twenty twenty five
stations are working around which comprise
about thirty five fueling stations
are our project derives
from the fact that it's not a southern
country has a sub optimal feeling policy
in place in our head dress this we start
off by setting the current operations
looking at the data that we have and we
then build a simulation based optimization
model based then provided them with
the optimal fuel rules fueling policy.
For them to utilize we estimated now your
value added of by eight million dollars
And in order that more folks are than
may continue to use this tool and
few should update.
These people rules or
the feeling policy with one of them.
It's all so
going into the problem in more detail that
the Southern currently has a static
feeling policy in place within the not for
certain system they have what's
defined as a fuel rule for
every station in the system basically
a fuel rule is a fill up to level for
every locomotive that the parts of
the tickler station so Station the A had
a fuel rule of four thousand five hundred
gallons don't mean that every locomotive
that departs that station leaves a four
thousand five hundred gallons of fuel
in the time one of the reasons or
not the Southern does this is because
they plan to have all the planes come into
a station they pull off the locomotive
heads a separate fueling yard fuel it over
there then move those locomotives back to
the trains and it's not known which
train each look what is going on and
then departs the station combining this
not the sub and also there's not take
into account the fuel prices of different
stations the grapple with here shows you
the fuel prices of five consecutive
days of the week across the system.
What's very clear is that certain
stations are always more expensive than
the others even though they
all change by the same degree.
Given that Knoppix up and
misses out on the opportunity of
fueling more of cheaper stations and
feeling less or more expensive stations
and this is exactly what we're
trying to come to address over here
in our project and this is a reduced
profitability is not the sort of country
to be and I heard over the magic board and
we will continue to explain
the methodology we had this project.
So our methodology revolves
around a reimbursable one of
the first space station reanalysis
of history here for six.
So it was the fewest plants.
On the schedule which tells us
the probability of a train going from one
station to the next as well as burn rate
which is the amount you will earn on
station and then this goes into
a single it's not recession.
Gives us a solution just a few rules for
how we look at the birth rates for
twenty four thousand visual of voters
over two months time period and
founders are actually normally distributed
with mean two point four seven gallons to
a mile and a standard deviation
of zero point eight six.
Given the magnitude of this burn
rate is a fairly large variation.
So we want to look at factors that may
be affecting this very very action.
We want to know why we can't just
use one standard burn rate for
everybody in the system.
So all of the factors such as how much
will the tonnage of the trains hold
actual physical length of the train as
well as the equipment number which is
that you need as ignition for
over the motor and
the amount of time which is a combination
of a mile here and the number of parcels
of little depended on these were
very significant factors making for
a great over what we call
the origin nation destination.
Here was in fact here.
So what we would do is compare the burn
rate from station to station the with
station B. the station and
of these pairings seventy six
percent were soon in doing so.
We concluded that this was
a significant factor will
be there on time for
our simulation which comprises these three
star National Design strategy was
to follow an optimisation model and
decide what fuels we would put in place
at each station to minimize the total
fuel expense for
not fix the problem we ran into it best.
Is that we couldn't calculate what Norfolk
sellers' total fuel expense what you'll
expand.
Is the amount of fuel you dispense the
station times the station fuel prices that
just talked about how we got
the station fuel prices and
the amount that we fuel at each station
is the fuel roll which is our fill up to
level minus what is in the tank
of the locomotive already.
And what's in the class of all of them are
ready is dependent on where the locomotive
was previously and then whatever.
Burn in route to get to the station but
because we don't know where locomotives
come from or from the many different
stations they come to we can't accurately
determine how much fuel is dispensed there
are four determine what our fuel expense
is so you get around this issue we
developed a simulation based optimization
model by which we generate our own routes.
And so by generating our own routes we
know where locomotives have been and
where they will go and
therefore can calculate fuel expense.
So we generated thirty three hundred
routes sequences which comprises thirty
thousand stops which represents about
a month's worth of Norfolk Southern three
months from there we
generated fuel rules for
which we run all these routes over and
then calculate an expected.
Total cost using a lower bound such
that the fuel provides enough fuel for
any train departing that station
to reach its next station.
We have ninety nine trillion possible
fuel combinations to look at
which is clearly a number too
large to examine all of them.
So we develop a curious to search to
search over a smaller subset of that.
The simulation work by running all
these routes sequences over a fuel and
calculating the expected
total cost in the expected.
Fuel dispensed the station
from there we choose
our optimal fuel set to be the fuel roll
with the lowest total expected cost.
So this is the solution that we found.
The color of every station indicates
the fuel that you can see and
we conducted sensitivity analysis
on that solution to the robots for
Norfolk Southern Scurrah variation and
schedules fuel price and burn amount and
we also looked at the impact on inventory
Norfolk Southern will see because
as a NAT Norfolk Southern will consume
the same amount of fuel every year.
It just redistributes how much
fuel is used the station.
Gave them how the demand would change for
us and David will now talk
about the savings from this.
And so so running the current force other
operating rules which is to take
vast you know forty five gallons or
simulation came up with an expensive total
cost of six hundred seventy eight billion
and now if we ran the fuels from
the previous slide which is
something that Norfolk Southern
could implement immediately.
These are static fuels at
every station every train or
that station is filled up to that.
Fuel that will be expected to cost
in six hundred seventy million.
Therefore we arrive at our
anyone expected savings.
Now we went ahead and
looked at if we could give north
from Southern a recommendation for
if they were able to track their loading
orders at least toward the future station.
So in this example if you are at Station
eight and you are going to station B.
if the price at Station eight is greater
than where you put the station B.
then you want to fuel just the minimum
amount needed to do that but
if the cost eight is cheaper than B. We
recommend an awful sudden fuel to the tank
past your two hundred
gallons at that station.
So this is dynamically changing for every
single locomotive within the station and
allows for an awful silence to
have increased settings based
on if they were able to travel using
this one station look it had halted
the expected total cost to six
hundred forty nine million.
Therefore the dental savings from using
the recommendation which one was on would
have to make changes in order
to travel to one station and
their expect to save twenty nine million.
And so we delivered in
the course of that we've given
the analysis of the system in your various
which is the gallons per mile that trains.
Burn from station to station.
We've also given them a solution that they
can implement immediately the static fuel
solution I was on the map and also the
inventory impact because of the changes
within the fuels of each station
affects the inventory needed for
that station and
we've given them a troll which.
Has written Excel runs on macros and
creates a few holes allowing them to make
changes for any current additions if
they want to change the schedule if they
want to change the locomotive numbers if
they want to change price the station.
Of course comes with a user's manual
which will come in the store and
you've got a future changes they
want to make as well if they want to
have any future operational
characteristics that they want and
also the recommendation of
a one station look at all.
You basically we want to just know for
Southern was able to track to one station
how they can field and fueling stations.
So this concludes our presentation
of the life of a doctor I met
he wasn't here with us today for
his direction and
we're now open to any questions that
we're going to change their
now so that the lower bound on all
the fuel roles provides enough fuel for
every train departing a station
treats its next destination.
So it provides enough fuel for ever.