You are on page 1of 16

Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

Introduction
In Six Sigma, catapults (“Statapults”) are a standard training tool used to teach process control
and to create experimental design ideas to test and observe a phenomena. These catapults are
made from wood with a retractable arm, a front pin, a stop pin, and a screw-eye which attaches a
rubber band to the arm. The catapult itself will shoot a rubber ball (the projectile) a certain
distance with the aid of the rubber band. For this project, we will apply six sigma measurement
tools through the define, measure, analyze, improve, and control (DMAIC) phases to improve
the quality of this catapult.

Part 1: Define
● Scope: Importance of the Project
○ This project focuses on improving the quality of the catapult we were provided
through the Six Sigma management tools
○ We will be using the DMAIC format to create an experiment that will allow us to
shoot the catapult with the most accuracy
○ It is necessary to perform this project in order to improve our knowledge of the
processes of Six Sigma, work together to create a process improvement project,
and form knowledge that can be utilized in similar situations in the real world
● Defect
○ Accuracy: A defect in accuracy would be a shift between the point at which our
ball actually hit compared to the target distance given in the competition
○ Precision: A defect in precision would be a high variation in the 10 shots
attempted (calculated in actual distance - target distance)
○ Overall: A defect is inaccurate or imprecise shooting of the catapult
● Parameters of the Catapult
○ What could be adjusted to determine the distance the ball travels:
■ How far back the arm is drawn: we used angles between 160-185 degrees
■ Position of the front pin: there were 4 different, vertical positions
■ Position of the stop pin:
● This pin was used to stop the arm of the catapult
● There were 5 different positions for this pin on the base of the
catapult
○ What could NOT be adjusted:
■ Position of the screw-eye attaching the rubber band to the arm--this was
placed in the position closest to the top of the catapult arm
■ Position of the cup that held the ball on the catapult
● Goals
○ General goal: To “characterize” our catapult by understanding the performance of
the machine, its variability, and the effects of its key parameters (in order to make
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

the shooting distance as accurate and precise as possible and hitting the provided
distance with as low variation as possible)
○ Goals of the Project:
■ The goal of this project is to be able to develop a process that will allow us
to shoot the catapult to any distance between 80-130 inches
■ The goal of this project is to be able to shoot the catapult to the correct
distance while maintaining a low standard deviation and mean square error
over 10 shots
■ Our goal is to be able to shoot the catapult in an in-control fashion
between 80-130 inches
■ We hope to improve our process continuously and observe accurately
where variation is occurring so we can make changes
● Timeframe
○ We were given roughly three weeks to create and test our process/method before
the competition on December 11, 2018
● Team
○ Our team consisted of five individuals
○ During our testing phases, two individuals held and shot the catapult, two
individuals watched to see at what distance the ball landed, and one individual
recorded the distance shot and the configuration of the catapult into Excel for
modeling and analysis purposes
● Resources
○ Catapult (“Statapult”)
○ Rubber ball
○ Tape measure
○ 2 pieces of black poster board: used to record where the ball landed
○ Baby powder: to powder the ball for a more accurate reading on the poster board
○ Plastic cup: used to hold the baby powder and powder the ball
○ Masking tape: used to hold the poster board and catapult down
○ Laptops/Excel: for modeling/analysis purposes
Part 2: Measure
● Shooting the Catapult: Elasticity and Movement of the Arm
○ Figures 1 and 2 below provides a representation of what our catapult looked like.
FIgure 1 is the catapult itself, and Figure 2 is the angle measurement system at the
base of the catapult.
○ The elasticity of this catapult comes from the rubberband, which is attached to the
front pin of the catapult, the arm of the catapult, and an immobile pin on the front
base of the catapult.
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

■ We positioned our rubber band so that only one end was attached to the
catapult arm, creating a triangle shape.
○ The movement of the catapult arm is restricted by how far we pull back on it
(pull-back angle), positioning the stop pin at any of the positions on the base of
the catapult, and the positioning of the front pin. The ability to vary these three
metrics gave us a wide range of movement and shooting distances for our
catapult, both vertically and horizontally.

Figures 1 and 2: Catapult and Angle Measuring System

● Shooting the Catapult: Measuring


○ We used the measuring tape to measure 140 inches from the front base of the
catapult. The measuring tape remained on the ground for the experiment to
measure our successive shots.
○ The black paper was placed on the ground from 75-135 inches
○ The ball was to be powdered before each shot. When the ball landed on the paper,
it created a white marking at the distance it landed, and this distance was recorded
in Excel along with the front and stop pin positions and angle of the shot.
However, in our measurements, in order to save time and resources, we did not
powder the ball before every single shot.
● Shooting the Catapult: The Experiment
○ We began shooting shots focused on the manipulation of the front pin and the stop
pin. Our goal with this strategy was to find a working combination of front and
stop pin positions that allowed us to shoot at any angle on the base and be within
our given range of 80-130 inches. Just moving the pins while keeping the angle
constant would probably give us less precise/diverse measurements and more
distance groups.
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

○ After shooting using various pin combinations, we ruled out the use of the
following pins which gave us extreme distances not within our required range of
80-130 inches. These pins either positioned the arm in a way that the ball barely
traveled from the cup or traveled too high rather than a certain distance:
■ Stop Pins: 1, 2, 3, and 5
■ Front Pins: 3 and 4
○ This led us to focus on the top two front pins (1 and 2) and the stop pin set at 4.
Once we realized that front pins 1 and 2 and stop pin 4 would give us the best
results, we started to manipulate the pull back angle.
○ After a few trials, we saw that front pin set 1, stop pin set 4, and pull back angles
from 160-185 were in target range. We also saw that front pin 2, stop pin set 4,
and pull back angles from 165-185 were in target range.
● Regression
○ Using these two combinations (front pin 1, stop pin 4; front pin 2 and stop pin 4),
we created two separate regressions and regression equations that included shots
at multiple angles.
○ Graphs 1 and 2 show our regression output for the two different models we used.
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)


Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

○ Regression Equation for 1, 4:


Distance = 2.0915 *angles + (-255.2205)
○ Regression Equation for 2, 4:
Distance = 1.7109*angles + (-198.8705)
○ By inputting a desired distance into the y variable, our equations were able to give
us the angles at which to pull the catapult back at either combination in order to
achieve this distance.
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

○ Combination 1-4 was able to go across all distances (80-130 inches), but
combination 2-4 did not go beyond 115. For that reason, we chose to use 1-4 for
distances beyond 115, and 2-4 for distances below 115. The 2-4 combination was
more accurate and showed less deviation when we experimented it against the 1-4
combination.

● Control Charts
○ In order to test the capability of our process before the competition, we decided to
create a few control charts based on a different distances that could be potential
distances in the competition and that provided us with a better idea of the
accuracy of our catapult at a range of distances

Test 1: Front: 1, Side: 4, Degrees: 185


Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

Fig. 2.1

We decided on these first settings because–from what we saw– after a few tries, it was shooting
in the last block of measurements we wanted to get (115-130). We have 30 subgroups of 3 data
points. In the x bar chart, the first block of 10 trials is in control. Between subgroups 11 and 20
our throws are out of control and then it goes back to being a little more stable, however not
entirely within the limits. There is a significant gap between shooting above 130 and around 120
from one group of shoots to another, so the shift can be associated with us modifying the number
of seconds that the arm was held for (from a higher distance to a low one) or fixing the rubber
band. Another factor that could have affected this variable set of data points is the fact that we
measured these observations in separate days. The first day we measured subgroup 1-9 and the
second day 10-30. The most “in control” test was the first day and there was a different shooter
on the second day. These factors altogether may have caused all the variability we see in this
chart.

As far as the R chart or Range chart, test one was is in control with the exception of 2 outliers. In
this case, there isn’t much of a difference between the 2-day blocks. This chart suggests we
didn’t experience much variability when it came to the measurement ranges. Our data kept
varying the approximately the same amount throughout both tests.

Test 2: Front: 2, Side: 4, Degrees: 170


Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)


Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

Fig. 2.2
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

Test 3: Front: 1, Side: 4, Degrees: 170


Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

Fig 2.3

For the second and third charts we recorded 30 observations of each and once again subgroups
of 3. We kept the same settings in both (S:4 and D: 170) with just changing the front pin between
1 and 2, the same way we did our regression equations.

Test 2’s X bar chart was in control and didn’t have much fluctuation (the first 4 subgroups are
very constant). The R- chart had narrow control limits (0-5.41 and the process maintained and
the ranges stayed very close to the mean, with 80% of the subgroups being just 0.1 units below
the centerline.

Test 3’s X bar chart was majorly in control with the exception of one outlier subgroup at the end.
It seems like the measurements were slightly increasing as we kept shooting, and this could be
due to common variation such as holding the degrees a little bit off the exact measurement once
we got comfortable and started shooting faster. The R chart had wider control limits compared to
the x bar chart, but the process still stayed closer to the mean than to the control limits. Aligned
with our previous observation, it seems that the ranges started increasing slightly as we kept
shooting and this could be for the same reasons highlighted previously.

Overall, both tests proved that our process was in control in both range and mean settings. At this
point of the process our team had already figured out what the best practice to shoot was and we
accounted for many of the errors made when we did our first control chart (e.g rubber band
placing) Test 3 proved to be very in control as it was entirely within the limits while test 2 had
only one outlier in the R chart, but it was still very stable.
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

Part 3: Analyze
After looking at our data, we wanted to create a fishbone diagram to try to identify the various
factors that could be influencing our data and causing large variation in our shots for analyzation:

● Catapult
○ Material of the pins: the pins were sometimes cold when we started shooting,
which could have caused variance.
○ Stability of the catapult: our catapult arm is slightly loose
○ Stability of the pins: We noticed that the front pin was slipping back, so we
insured that we keep the pin fully pushed in.
○ Variance in the rubber band: The rubber band seemed to become warmer and
looser with successive shots due to increased friction with the pin. The difference
in elasticity caused the catapult arm to produce shots of difference length at
different times in the session at the same angle and pin locations.
● Environment
○ Air resistance: We used the same ball for each trial, but if we had lost the ball or
used another during the trial that was of a different material or size, this would
have skewed data. We also worked in a common area where the door to outside
was constantly opening, potentially causing a difference in temperature and a
slight difference in air resistance
○ Storing of the catapult: We stored our catapult in a cold car, and during our
sessions, we realized that the earlier shots we took (when the catapult was still
cold), produced different outcomes than the shots we took later in our session
(when the catapult had “warmed up”)
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

○ Floor: We worked on the same floor for the entire experiment and competition,
but not taping the catapult to the floor and the potential of powder to be on the
floor where the catapult lied could have caused movement in trials
● Shooting Method
○ Direction control of the catapult: We noticed that our catapult tends to shift to the
left, so we shift it minimally right to balance that out.
○ Holding the arm back: We noticed that results varied depending on the length of
time the catapult was held back in the shooting position. To combat this, we made
sure the shooter held the arm back for 5 seconds and called out each of these
seconds once they were set up at the pull back angle
○ Pressing the ball into the cup: If you press the ball in with force in the cup, it can
cause the ball to release from the cup later, making it shoot less far.
○ Holding the catapult: Instead of taping the catapult to the floor, we chose to hold
the catapult for the vast majority of our trials. The strength of those holding the
catapult to the ground could impact shifting in successive trials
○ Time between shots: Taking too much time between shots led to a cooling down
of the rubberband and sometimes a loss of focus among team members
● Measuring
○ Use of powder: We did not use powder for all of our trials. Therefore, we may not
have received as accurate of a measurement simply via observation with the
human eye
○ Human observation: Although we did have two observers for our data, because
we did not use powder everytime, they would occasionally disagree with the
measurements seen. Even when powder was used, there were sometimes varying
conclusions. In general, the human eye is not as accurate as a more robust or
technology-based form of observation
○ Position of observers: We noticed that the angle at which our person was
measuring our shots mattered. To combat variation due to inaccurate
measurement, we ensured that the same person recorded each angle and that they
were at a “middle” position when measuring--not looking on from a side angle.
● Shooter
○ Incorrect positioning of the ball: If the ball was positioned by the shooter
incorrectly on the arm of the catapult, this could result in an inaccurate shot or a
complete mistrial overall
○ Shooter strength and stability: Stronger shooters will be able to hold the catapult
arm down longer and with more stability. Even when the same shooter is used
each time to reduce shooter to shooter variation, the strength and focus of this one
shooter could deplete after multiple trials
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

○ Length of time the catapult is held back: that the shooter had to make sure they
started the count of the catapult arm at the same time each instance. To ensure
this, we made sure the same person shot the catapult each time so they could get a
feel for it.
Part 4: Improve
● Improvements Made to Our Shooting Process for the Competition
○ Use of warm-up time: We made sure to utilize our warm-up time during the
competition so that all of the components of our catapult could warm up to
produce more accurate shots to what we had recorded in our trials. Our rubber
band became looser and the metal pins were warm as well. We also used this time
to confirm our regression equations and determine a strategy for potential target
measurements.
○ Taping down the catapult: In our trials, we did not tape down our catapult, but for
the competition, we decided to do so in order to reduce variation that could come
from the floor or from our team members holding the catapult down
○ Small angle and holding adjustments made during the trials: Once we were given
the number we were supposed to hit in the competition, we used our regression
equation to determine the pin positions and the angle that we were to use. After
identifying the variations in our process, we knew that this could have skewed
some of the points in our data, so our regression may not be 100% accurate.
However, from our analysis, we knew the certain components of our process that
produced variation. If the combination that our regression gave us did not shoot
the ball in the correct position, we made small adjustments to either the amount of
time we held the ball down or the angle at which the ball was shot at in order to
try and hit our target. From successive trials, we knew how to work with the
variations in our catapult to use them to our advantage.
○ Use of baby powder for every trial: We used baby powder for all of our warm-up
shots and trials during the competition to ensure more accurate measurements.
● Side Effects of Improvements
○ Overcorrection/Undercorrection: Even though the adjustments that we made from
our regression equation were based off of knowledge of our previous trails, they
can sometimes lead to overcorrection or undercorrection which would lead to a
greater variation in our final data. If a greater than expected variation occurred,
we were often left shifting between many slight variations. It was stressful to us
and hurtful for the process as a whole to have so many minor adjustments.
○ Inexperience with Powder: Because we did not powder our ball for every trial in
our experimentations, we iftern did not powder the ball enough during the
competition, which led to confusion and potentially inaccurate measurements by
our judges.
Rohan Gupta December 20th, 2018

Statapult Project (Six Sigma)

Part 5: Control
● Control Plan for the Competition
○ Use of different regressions/pin combinations depending on the measurement
given: From our experimentation, we knew that positioning the front pin in the
second position and the stop pin in the fourth position resulted in more accurate
trails compared to target measurements. At its greatest angle, this combination
could only hit up to 115 inches. Placing the front pin in the first (top) position and
the stop pin in the fourth location gave us greater variation, but we were able to
angle the arm to shoot at a distance of over 130 inches. Thus, in competition,
based on the measurement we were given, we decided to go with a certain
position and stuck with that position. For any number from 80-115, we used the
2-4 position, and for any measurement greater than that, we used the 1-4 position.
○ Use of the same shooter/shooting method: We decided to use the same shooter
that we had used in all of our trials for the competition. We also employed the
same shooting method of holding down the arm for a baseline of five seconds and
calling out those five seconds verbally.
○ Start from regression equation output and make very small adjustments from
there: We decided before the competition started to make adjustments to the
holding time and angle of the catapult arm if we saw the provided angle from our
regression equation was not accurate (this is taking into account all the variation
in our data from experimentation). We tried this strategy in our final session
together and in our various warm-ups during the competition, and it proved to be
effective. This was confirmed as we were trying to hit our targets and we were
able to hit more accurate measures.

You might also like