Você está na página 1de 12

Quality Engineering, 18:431442, 2006

Copyright # Taylor & Francis Group, LLC


ISSN: 0898-2112 print=1532-4222 online
DOI: 10.1080/08982110600719415

A Six Sigma# Case Study: G.E.P. Boxs Paper Helicopter


ExperimentPart B
J. A. Johnson, S. Widener, and H. Gitlow
School of Business Administration, University of Miami, Coral Gables, FL

E. Popovich

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

Sterling Enterprises, Inc., Boca Raton, FL

This article presents an application of the Six Sigma


DMAIC model to G.E.P. Boxs famous paper helicopter
experiment. The Improve and Control Phases are presented
here. The Define, Measure, and Analyze Phases were presented in an earlier paper. The intent of this article is to
present the reader with a case study for structuring a Six
Sigma project.
Keywords

Case study; Control; DMAIC model; Improve;


Paper helicopter experiment; Six Sigma management.

INTRODUCTION
This article presents the Improve and Control Phases
of an application of the Six Sigma DMAIC model to
G.E.P. Boxs famous paper helicopter experiment
(Box, 1992; Johnson et al., in review). The intent of this
article is to present the reader with a case study for structuring a Six Sigma project. Similar work has been done
with this type of case (Rasis et al., 2002).
In the previous article (Johnson et al., 2006), which
contained the first three phases of the DMAIC model
(Define, Measure, and Analyze Phases), the Lilliputian
National Army (LNA) was moving toward a siegebreaking technique based upon aerial assault, instead
of traditional land-based attack methods. By following
the first three phases of the model, the process to be
improved was determined to be the Helicopter

Construction Process, and the CTQ in question was


identified to be the flight time in seconds.
The Lilliputian scientists and engineers studied the
effects of the seven variables, or Xs, on the flight time, or
CTQ, in the Improve Phase. The Xs are paper type (X1),
body length (X2), body width (X3), wing length (X4),
paper clip (X5), body tape (X6), and wing joint tape (X7).
IMPROVE PHASE
The Improve Phase optimizes the CTQ(s) by
manipulated the Xs and their interactions. It has three
steps: (1) design and run experiments to isolate the significant Xs and their interactions, (2) interpret the
results of these experiments to determine the settings
of the Xs which optimize the spread, shape, and center
of the CTQ(s), and (3) conduct a pilot study to test the
settings of the Xs.
The deliverables of the improve phase are:
(1) the levels of the Xs that optimize the CTQ(s)
(2) the optimized process map with the optimized settings of the Xs
(3) the results of a pilot study of the revised process
(4) the estimated capability (sigma level and=or
DPMO) of the revised process
(5) the statistical gap between the original and the
revised system capabilities
Design of Experiments

Six Sigma is a registered trademark of the Motorola


Corporation.
Address correspondence to H. Gitlow, Professor of
Management Science, School of Business Administration,
University of Miami, Coral Gables, FL 33124, USA. E-mail:
hgitlow@miami.edu

Designing the Experiment


The designed experiment contained seven Xs, which
were identified in the Analyze Phase of the project. The
experiment was initially viewed as a full factorial design

431

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

432

J. A. Johnson et al.

with two replications per trial, resulting in 27 trials run


twice, which would have required 256 independent trials.
Manufacturing each prototype requires approximately 60 man-days just to build the machine if all of
the components are available and ready to go. Flight testing setup requires an additional day to prepare all of the
scientific equipment. Although the actual test is brief, the
work put into making each test a success is quite extensive. This comes to a head in considering the amount
of time available to actually run the experiment. As
stated in the Define Phase, war with the Lilliputian Freedom Fighters (LFF) is imminent, and spending over six
months in testing is not an option. Therefore, the time
must be cut. However, this is not the only constraint,
as the budget is also limited.
Realizing that the total material budget for the
experiment is 15,000,000 gold florins, and that the costs
of production are essentially fixed for retooling machines
and changing the setups, the LNAs financial analysts
have concluded that each prototype, including overhead
expenses, cost about 100,000 gold florins and the testing,
including overhead expenses, was an additional 10,000
gold florins. Considering that these figures included the
overhead, which covered salaries of the workers and
staff, essentially the entire budget could be dumped into
this experiment. This was the AACs (Army Air Corp)
motivation for allocating funds away from helicopter
purchases and into development. As a result of this,
quick arithmetic showed that 128 total runs were possible, with a small pool of leftover money in the budget
to help offset any potential bumps along the way. It is
worth noting that these figures do not include the
3,300,000 gold florins that were allocated to the pilot test.
128 total runs for the experiment was compatible
with the time constraint, and as a result, it was the
number of trials chosen for the experiment. The problem became that of experimental clarity. Running
128 individual trials in the full factorial design, which
would then be replicated, requires a total of 256
individual trials. This will fully show all interactions
and relationships between the variables, but it requires
too much time. Therefore, reducing the design to a half
fractional factorial, requiring 64 randomized individual trials, plus a randomized replication, for 128 total
individual trials is a better solution. However, the cost
of saving the time and money of the trials is the inability to distinguish between the effects of certain interactions. This is known as confounding.
Potential Confounding Problems
The half-fractional design reduces the number of
runs, but at a loss of information due to confounding.

This 27-1 design has a resolution of VII. In this case,


this means that the main effects and two-way interactions will not be confounded with each other. The
main effects are confounded with the six-way interactions and the two-way interactions are confounded
with the five-way interactions. This confounding is
acceptable because field expertise dictates that highorder interactions are safely ignored. The resolution
VII confounding description continues throughout
the design, so that the three-way interactions are confounded with four-way interactions.
Given that five-way and six-way factor interactions
are extremely rare, the chances of significantly confounding the main effects or the two-factor interactions
are very small. The problem lies in the three-factor
interactions being confounded by the four-factor interactions. Since three-factor interactions do happen,
although not very often, and four-factor interactions
are more uncommon than three-factor interactions,
the chances of these effectively clouding each other is
small, especially when weighted against the costs of
running a full factorial to ensure that all of the effects
are isolated. Therefore, the Lilliputians have not backed
themselves into a corner through confounding as it
might seem, but instead have taken a calculated risk
to minimize the cost and duration of the experiment.
Gearing Up for Running the Experiment
Two aircraft of each of the 64 prototypes were
manufactured. This ensured that two separate aircraft
were available for testing, which served two purposes.
First, this helped to determine within-prototype variation. Second, in the event of a catastrophe, a second
unit was available for the experiment.
The aircraft were manufactured using the same
processes as they would see in real-time production.
The Flight Testing was done in the laboratory instead
of the usual facilities. However, the inspection was still
performed to ensure that the aircraft, were it standard,
would be fully functional, operable, and ready for use
in the field.
Running the Experiment
Each aircraft was marked with a number corresponding to its position on the design layout. It is
worth noting that the design layout was randomized,
to curb the effects of lurking variables. Each aircraft
was individually dropped and the times were recorded
per the directions laid out in the Measure Phase. Next,
team members obtained the results of the experiment.

GEP Boxs Paper Helicopter ExperimentPart B

Interpretation of the Experiments


Preliminary Interpretation
Figure 1 shows a Pareto chart of the Xs, and their
interactions, that significantly impact flight time.
The bars in Figure 1 that extend past the dotted
line (5% level of significance) represent the Xs or interactions of potential interest. It can be seen that the
main effects, with the exception of D, are the most
important in the explanation of the Xs impact on the
flight time. Furthermore, Figure 1 shows that there
are potentially four two-way interactions, FG, AE,
BF, and AB, that could be working in this design.

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

In-Depth Analysis
Team members studied the Analysis of Variance
(ANOVA) output for the experiment to identify which
factors and interactions significantly effect flight time
(see Table 1). The second part of Table 1, Analysis
of Variance for Flight numerically reiterates the
impact of the two way interactions and the main effects
that was illustrated in the Figure 1. This gives some
further insight as to what will actually be found in
the first part of Table 1, Estimated Effects and Coefficients for Flight.
The first part of Table 1, Estimated effects and
coefficients for flight, shows the p-values for each
main (factor) effect and interaction effect. All effects
that are statistically significant are indicated by a
p-value less than 0.05. The statistically significant main
effects and interactions have been condensed into
Table 2.

Figure 1.

Pareto chart of the standardized effects.

433

Figure 2 shows that wing length is the most significant X. Paper, clip, width, and length are of secondary
importance. Body taping and joint taping are the least
significant of all the factors.
All interactions supercede all main effects to determine what value of the individual design factor should
be set at. With this in mind, it would be wise to start at
the bottom of Table 2, and work up, analyzing each
significant interaction. The first interaction is
Length! Wing, and going to Figure 3, it can be seen
that the longer wing, 4.75 inches, and the shorter
length, 3.00 inches, optimize the flight time, as that is
the highest point within the box where the Length
and Wing columns or rows intersect. Both combinations in the table will yield the same result, as the
two plots are inverses of each other. The next interaction is Clip! Length, and using the same methods to
find the highest point within the relevant boxes on
Figure 3 shows that the length should be 3.00 inches
and that no paper clip should be used. Continuing
up Table 2, the next interaction is Paper! Width, and
using the relevant boxes in Figure 3, shows that the
width of the helicopters body should be 1.42 inches
and that the helicopter should be made with white
paper. The final interaction is Paper! Clip, and using
Figure 3, it can be found that white paper should be
used with no paper clip. These results are then compiled and summarized in the Table 3.
Note that in Table 3 there is no indication for the
factors labeled Body Tape or Joint Tape. This is
because there were no interactions for these factors,
and as a result, no known information can be
displayed.
With all of the interactions from Table 2 having
been analyzed, the remaining effects, known as main
effects, or single factor effects can be analyzed. The
analysis for these is found in Figure 2, and the first
of these effects from Table 2 is Wing. Consulting
Figure 2, the optimal wing length is 4.75 inches. In this
case, it can be seen that this is consistent with the wing
length that was determined by analyzing the relevant
interaction. If there is a conflict between the analyses
of the interactions and the main effects, in general,
the interaction affect is given consideration first.
The next two items found in Table 2, Length and
Width, show that the analysis of the main effects for
best results is consistent with the interaction analysis.
However, the next item in Table 2, Body tape, must
be considered.
Body Tape is listed as one of the unknowns in
Table 3; therefore the main effect is relevant to determining the level for the factor in the final design of
the aircraft. Consulting Figure 2, it can be seen that

434

J. A. Johnson et al.
Table 1
Original ANOVA output from Minitab
Estimated effects and coefficients for flight (coded units)

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

Term
Constant
Paper
Clip
Bodytape
Jointtape
Width
Length
Wing
Paper! Clip
Paper! Bodytape
Paper! Jointtape
Paper! Width
Paper! Length
Paper! Wing
Clip! Bodytape
Clip! Jointtape
Clip! Width
Clip! Length
Clip! Wing
Bodytape! Jointtape
Bodytape! Width
Bodytape! Length
Bodytape! Wing
Jointtape! Width
Jointtape! Length
Jointtape! Wing
Width! Length
Width! Wing
Length! Wing
Paper! Clip! Bodytape
Paper! Clip! Jointtape
Paper! Clip! Width
Paper! Clip! Length
Paper! Clip! Wing
Paper! Bodytape! Jointtape
Paper! Bodytape! Width
Paper! Bodytape! Length
Paper! Bodytape! Wing
Paper! Jointtape! Width
Paper! Jointtape! Length
Paper! Jointtape! Wing
Paper! Width! Length
Paper! Width! Wing
Paper! Length! Wing
Clip! Bodytape! Jointtape
Clip! Bodytape! Width
Clip! Bodytape! Length
Clip! Bodytape! Wing
Clip! Jointtape! Width
Clip! Jointtape! Length

Effect

Coef

SE Coef

0.14734
"0.12797
"0.05828
"0.00516
"0.17797
"0.16391
0.49297
"0.04484
"0.02641
"0.01828
"0.05172
"0.04078
"0.01328
0.01953
"0.01859
0.01984
"0.04984
0.03016
"0.02141
"0.01797
0.01672
0.03359
"0.01797
0.01672
"0.00703
"0.03047
0.00078
"0.05516
0.04141
0.00078
"0.00391
"0.00297
0.01641
"0.03953
0.01703
0.00734
"0.00766
"0.01672
0.03234
0.00297
0.03453
"0.01609
"0.02766
0.03641
"0.01141
"0.01297
"0.03422
"0.01516
0.00078

2.10523
0.07367
"0.06398
"0.02914
"0.00258
"0.08898
"0.08195
0.24648
"0.02242
"0.01320
"0.00914
"0.02586
"0.02039
"0.00664
0.00977
"0.00930
0.00992
"0.02492
0.01508
"0.01070
"0.00898
0.00836
0.01680
"0.00898
0.00836
"0.00352
"0.01523
0.00039
"0.02758
0.02070
0.00039
"0.00195
"0.00148
0.00820
"0.01977
0.00852
0.00367
"0.00383
"0.00836
0.01617
0.00148
0.01727
"0.00805
"0.01383
0.01820
"0.00570
"0.00648
"0.01711
"0.00758
0.00039

0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089

193.39
6.77
"5.88
"2.68
"0.24
"8.17
"7.53
22.64
"2.06
"1.21
"0.84
"2.38
"1.87
"0.61
0.90
"0.85
0.91
"2.29
1.39
"0.98
"0.83
0.77
1.54
"0.83
0.77
"0.32
"1.40
0.04
"2.53
1.90
0.04
"0.18
"0.14
0.75
"1.82
0.78
0.34
"0.35
"0.77
1.49
0.14
1.59
"0.74
"1.27
1.67
"0.52
"0.60
"1.57
"0.70
0.04

0.000
0.000
0.000
0.009
0.814
0.000
0.000
0.000
0.043
0.230
0.404
0.021
0.066
0.544
0.373
0.396
0.365
0.025
0.171
0.329
0.412
0.445
0.128
0.412
0.445
0.748
0.166
0.971
0.014
0.062
0.971
0.858
0.892
0.454
0.074
0.437
0.737
0.726
0.445
0.142
0.892
0.118
0.462
0.209
0.099
0.602
0.553
0.121
0.489
0.971
(Continued)

435

GEP Boxs Paper Helicopter ExperimentPart B


Table 1
Continued
Estimated effects and coefficients for flight (coded units)
Term
!

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

Clip Jointtape Wing


Clip! Width! Length
Clip! Width! Wing
Clip! Length! Wing
Bodytape! Jointtape! Width
Bodytape! Jointtape! Length
Bodytape! Jointtape! Wing
Bodytape! Width! Length
Bodytape! Width! Wing
Bodytape! Length! Wing
Jointtape! Width! Length
Jointtape! Width! Wing
Jointtape! Length! Wing
Width! Length! Wing

Effect

Coef

SE Coef

0.00391
0.04109
0.02172
0.00328
"0.00422
"0.00641
"0.01391
0.00766
"0.03359
0.01734
0.02078
"0.04234
"0.00516
0.00703

0.00195
0.02055
0.01086
0.00164
"0.00211
"0.00320
"0.00695
0.00383
"0.01680
0.00867
0.01039
"0.02117
"0.00258
0.00352

0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089

0.18
1.89
1.00
0.15
"0.19
"0.29
"0.64
0.35
"1.54
0.80
0.95
"1.94
"0.24
0.32

0.858
0.064
0.322
0.881
0.847
0.770
0.525
0.726
0.128
0.429
0.343
0.056
0.814
0.748

Analysis of variance for flight (coded units)


Source
Main effects
2-Way Interactions
3-Way Interactions
Residual Error
Pure Error

DF

Seq SS

Adj SS

Adj MS

7
21
35
64
64
Total

10.9781
0.6043
0.5369
0.9707
0.9708

10.9781
0.6043
0.5369
0.9707
0.9708
127

1.56830
0.02878
0.01534
0.01517
0.01517
13.0900

103.40
1.90
1.01

0.000
0.026
0.473

in the Body tape box, the line is tipped slightly up on


the left side. Reading the top shows that the high side
is associated with no taping, therefore the main effect
indicates that there should be no tape on the body.
This value can then be added to the table for the final
design.
Continuing along with the analysis of Table 2, two
more factors are found, Clip and Paper, yet both of

these were determined by interaction. A summary of


the analysis of Table 2 to achieve a final design is
shown in Table 4.
The problem that remains is that there is still no
solution to the Joint Tape problem, which is listed as
an Unknown in Table 4. Since this was not listed
as a significant factor in either the interactions or
the main effects, the level of this factor is essentially

Table 2
Significant effects from Table 1
Term
Constant
Paper
Clip
Bodytape
Width
Length
Wing
Paper! Clip
Paper! Width
Clip! Length
Length! Wing

Effect

Coef

SE Coef

"0.14734
"0.12797
"0.05828
"0.17797
"0.16391
0.49297
"0.04484
"0.05172
"0.04984
"0.05516

2.10523
"0.07367
"0.06398
"0.02914
"0.08898
"0.08195
0.24648
"0.02242
"0.02586
"0.02492
"0.02758

0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089
0.01089

193.39
"6.77
"5.88
"2.68
"8.17
"7.53
22.64
"2.06
"2.38
"2.29
"2.53

0.000
0.000
0.000
0.009
0.000
0.000
0.000
0.043
0.021
0.025
0.014

436

J. A. Johnson et al.
Table 3
Factors determined through analysis of interactions
Variable name
Paper
Clip
Bodytape
Jointtape
Width
Length
Wing

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

Figure 2.

Variable

Optimizing value

Paper type
Paper clip
Body tape
Joint tape
Body width
Body length
Wing length

White paper
No clip
Unknown
Unknown
1.42 inches
3.00 inches
4.75 inches

Main effects plot on flight times for the helicopters.

irrelevant to the design. Ideally, the AAC would like to


eliminate costs, and as a result, not taping the joint
would save some labor and extra material. Furthermore, this decision is justified by the main effect shown
in Figure 2 related to Joint Tape. The line in that box is
almost horizontal indicating no effect, hence its lack of
significance. (If there is another CTQ not studied here
that is impacted by Joint Tape [such as time to provide
helicopter maintenance], then it is reasonable to consider Joint Tape. We do not see any reason to use Joint
Tape.) Therefore the finalized design parameters, taking into account interactions, main effects, and cost
cutting moves are shown in Table 5.
Pilot Study
The final section of the Improve Phase is taking
the optimized design and conducting a pilot study to
determine its effectiveness in improving the distribution of the CTQ. A revised flowchart can be seen in
Figure 4.
A scaled production line was assembled and the
new improved prototypes were built. It was determined

Table 4
Final design parameters based on main effect and
interaction analysis
Variable name
Paper
Clip
Bodytape
Jointtape
Width
Length
Wing

Variable

Optimizing value

Paper type
Paper clip
Body tape
Joint tape
Body width
Body length
Wing length

White paper
No clip
No tape
Unknown
1.42 inches
3.00 inches
4.75 inches

that the new helicopters required 62 man-days to build,


versus the previous 60 man-days, due to complications
with the longer wing structures. As each helicopter
rolled off the line it was subjected to the same tests
as a full-scale version. The results of those flight times
are shown in Figure 5.
The new process resulted in a stable process with a
lower control limit above the lower specification limit of
2.60 seconds, shown in Figure 1. Recall, it is acceptable
to place specification limits on an Individuals chart.
However, it is also of note that the variation has
increased as a result of this upward shift in flight time,
as R-bar has increased from Figure 1s 0.05885 seconds
to Figure 4s 0.08138 seconds. When the variation
increases as the mean time increases this is often the case

Table 5
Design parameters for new aircraft
Variable name

Figure 3.

Interaction plot on flight times for the helicopters.

Paper
Clip
Bodytape
Jointtape
Width
Length
Wing

Variable

Optimizing value

Paper type
Paper clip
Body tape
Joint tape
Body width
Body length
Wing length

White paper
No clip
No tape
No tape
1.42 inches
3.00 inches
4.75 inches

437

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

GEP Boxs Paper Helicopter ExperimentPart B

Figure 4.

Improved LNA AAC helicopter construction flowchart.

in skewed time distributions. However, the skewness in


time distributions is usually to the right (greater values)
side. In other words, higher values may appear as part
of the new process but that is acceptable as the goal
was to exceed 2.6 seconds. The flight times have significantly improved. Figure 6 compares the distributions of
the former Prototype 1 units with the newly developed
units.

CONTROL PHASE
The Control Phase establishes the ongoing
controls necessary to sustain the benefits from the
Six Sigma project. It comprises four steps: standardization, mistake-proofing, documentation, and
giving control of the process back to the process
owner.

438

J. A. Johnson et al.

Figure 6. Comparing the distributions of the new helicopters and prototype 1.

(1) employees not being educated about the new wing


length
(2) employees folding the body in the wrong directions, i.e., folding at positive 180 degrees for both
folds or vice versa
(3) employees folding the wings in the same direction,
i.e., folding both wings at positive 90 degrees or
vice versa
(4) employees folding the wings at angles other than
=" 90 degrees
(5) process owner tampering with the process

Figure 5. Flight times for helicopters built on the improved


flowchart.

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

The deliverables from the control phase are:


(1) an implemented standardization and mistake
proof control plan
(2) documentation of the revised process
(3) demonstration of significant and sustainable
improved performance
(4) a lesson-learned document
(5) a project replication=leverage plan
(6) a project financial benefit document
(7) return of the revised process to the control of the
process owner for continuous turns of the PDSA
cycle
Standardization
Standardization was not viewed to be a significant
problem given that the process already operates from a
single flowchart. Reasonable care will be taken to
ensure that the new process flowchart is incorporated
on the production floor and further problems will be
addressed in the rest of the Control Phase.
Mistake-Proofing
Team members identified potential risk elements
that could be carried through the process by using
Failure Mode Effects Analysis, or FMEA (see Table 7).
These risk elements are:

Each of the Risk Elements is subjected to a rating on a


1 to 10 scale on Likelihood of Occurrence, Severity,
and Likelihood of Detection, with high numbers being
more problematic. The Risk Priority Number (RPN) is
calculated as the product of those three scales.
Risk Element 1Employee Awareness of the New
Wing LengthRPN 360
Employee education will alleviate many of the
employee awareness problems concerning the new
wing length by teaching the employees exactly how
their job relates to the operation and structure of the
entire helicopter. This awareness can be instilled
through a short one-day training course.
Another measure that could be implemented to
avoid confusion, and potential mistakes, about the
wing length would be to lock in the calibrations of
the War Paper Machine so only one wing length could
be cut. The war paper machine is discussed fully in
Risk Element 4, and can be seen later in Figure 9. This
would reduce the RPN to 100. Furthermore, a completed mock-up of the helicopter can be kept on nearby
display for quick visual comparison and as a guide.

Table 6
Process capability with the new design
Yield
CTQ
Flight time

DPMO

Sigma

Current

Desired

Gap

Current

Desired

Gap

Current

Desired

Gap

99.90888600%

99.90000000%

"0.00888600%

911.14

1000.00

"88.86

4.62

4.59

"0.03

150
6
5

6
1
3

10
1
10

8
2
2

100
4
5

439

Risk Element 5Tampering with the Process


RPN 336
The process owner will be thoroughly educated
with regards to the impact of process tampering performing the famous funnel experiment (Gitlow et al.,
1995). The RPN should drop to 150 based on the
educational program.

5
336
8
7

2
135
3
9

1
10

3) Incorrect wing Wings facing the


folds by
same direction
direction
4) Incorrect wing Unbalanced wings
folds by angle
5) Process
Process spiraling
tampering
into chaos

20

Redesign of template
and warning
sign with picture
ofcorrect fold
Redesign of template and
integration of wing
tolerance tool
Integration of wing
tolerance tool
New process flowchart
and control plan
2
6

24

5
9
8

360

Helicopter education
seminar

Risk Element 4Incorrect Wing Folds by Angle


RPN 135

Incorrect wing
1) Employee
length
awareness
of wing length
2) Incorrect body Single edge folds
folds

Action
Likelihood of
Likelihood of
Occurrence Severity
detection
RPN
Failure mode
Risk elements

Table 7
Failure mode effects analysis (FMEA)

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

Likelihood of
Likelihood of
occurrence
Severity
detection
RPN

GEP Boxs Paper Helicopter ExperimentPart B

Lilliputian workers cannot reliably fold at a 90degree angle. This problem can easily be circumvented
using the wing tolerance tool. This tool requires that
the angle of the wings be within one degree of specification, 90 degrees, for the tool to fit onto the helicopter. The RPN drops to a value of 6 given use of the
wing tolerance tool.
Problem 2Incorrect Body FoldsRPN 24
Requiring employees to check that the body has
two edges can circumvent the problem of folding the
body in the wrong directions (see Figure 7).
Figure 7 will be posted on the production floor
and will be included in the training manual. In addition
to this commonsense circumvention measure for Risk
Element 2, there is an additional precaution that incorporates instructions into the design (see Figure 8).
Employees should be able to read the word Outside and should not be able to read the word Inside
on a completed helicopter. This is not a substitute for
the first control measure because this precaution is not
capable of preventing the body folding problem.
While it may seem a simpler solution to have text
on both sides of the paper, the War Paper Machine
does not replicate designs that require objects on both
sides of the paper, and forcing it to do so would create

Figure 7.

Correct and incorrect body folds.

440

J. A. Johnson et al.

Figure 8.

New war paper design.

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

an endless cycle of recalibration that is otherwise unnecessary. Figure 9 shows a picture of the War Paper
Machine. The RPN drops to 8 given these precautionary measures.
Risk Element 3Incorrect Wing Folds by Direction
RPN 20
This risk is catastrophic, but it is blatantly obvious
because it will appear as though the helicopter is missing a wing. This has not been a problem in the past due
to ease of detection; however, reducing the risk is quite
simple. The design of the helicopter template has been
modified to make it easier to determine folds, as shown
in Figure 8. The words Bottom Side should only be
visible from below the helicopter and vice versa for
Top Side. The new design should be communicated
to all relevant employees.
The wing tolerance tool that was developed for
Risk Element 4 provides an additional control for Risk
Element 3 because it is nearly impossible for incorrect
wing folds to go undetected. Therefore, the RPN
drops to 10.

Figure 9.

The war paper machine

It is worth noting that all of these items do not


directly impact the new flowchart for the Helicopter
Construction Process that was generated in the
Improve Phase, but are instead supplemental, in that
they impact other processes flowcharts. For example,
the redesigned templates would change the Helicopter
Tool and Die Flowchart, and the Helicopter Construction Employee Training Flowchart would have to be
updated to account for teaching the new body folding
mechanisms as well as a thorough understanding of
how the helicopter works.
Documentation
The primary piece of documentation required is
the new flowchart that was generated in the Improve
Phase. This flowchart needs to be inserted into all relevant training manuals, as well as into any standard
operating procedures kept on the production floor.
Additionally, copies of all relevant control charts
should be kept on the production floor to monitor
the process. ISO documentation needs to include the
new Standard Operating Procedure.
Demonstration of Significant and Sustainable
Improved Performance
Team members observed 100 trials of the new helicopter design in the field (see Figure 10). The distribution of flight times for the new design was stable and
normally distributed with a mean and standard deviation very close to the pilot test mean and standard
deviation. Team members were confident that the
new helicopter design provided sustained performance
significantly better than the original helicopter (see
Figure 1 in Box, 1992) design.

Figure 10. Sustained improvement in flight time.

GEP Boxs Paper Helicopter ExperimentPart B

The most significant lesson learned from this project was the value of Six Sigma methods in improving
and innovating the LNA to better fight the LFF. Consequently, the training arm of the LNA developed Six
Sigma courseware and offer it on demand to any division of the LNA desirous of such a training program.

between the first burr caused by a dull knife should be


recorded to determine the optimal number of linear feet
between changing or sharpening the knife blade. This
information should be analyzed using a control chart.
This concludes the project. The improved process
is turned over to the process owner with control plans
for future turns of the PDSA cycle. The team disbands
and celebrates their success.

Project Replication/Leverage Plan

SUMMARY

A project replication=leverage plan was developed


and put into play by contacting all of the top-ranking
officers of the LNA (Lilliputian National Army) and
making them aware of the benefits and availability of
Six Sigma theory and methods training programs. It
was a real success in that many generals signed up
for the program.

The purpose of this paper was to demonstrate an


application of the DMAIC model. It is hoped that this
paper will be used as a model for Six Sigma projects.

Lesson Learned Document

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

441

Project Financial Benefit Document


The financial analysis discussed in Box (1992) did
need to be updated in that it accurately represented the
benefits and costs of the new helicopter design. The
new design is a big winner for the LNA and AAC.
Returning Process Control to the Process Owner
The final step of the Control Phase is returning
control of the process back to the process owner,
which requires a control plan so the process improvements are not lost. A big piece of this control plan is
the documentation described in the previous section,
plus relevant control charts. A copy of the revised
flowchart must go to the process owner for future
planning for the iterations of the Plan-Do-Study-Act
cycle known as the PDSA or Deming Cycle.
These are traditional items in a control plan, however; in this case, there are two special issues that were
identified in the FMEA. These two issues were that of
the knife-induced defects and the calibration of the War
Paper Machine. The first issue that the process owner will
have to deal with is that the design of the War Paper
Machine is not robust enough to maintain calibrations.
Samples must be taken that measure the dimensional
accuracy of the War Paper Machine. This information
should be recorded in case future investigative work with
regard to the flight time control charts in needed.
The second issue that the process owner will need to
deal with is dull knives. Dull knives create burrs which
negatively effect flight time. The number of linear feet

APPENDIX 1: RESULTS RELATIONSHIP TO


THE G.E.P. BOX EXPERIMENTS HELICOPTER
DESIGN
The helicopter created in this project through the
designed experiment is actually much different than
the helicopter generated by Box (1992). According to
Box, who used eight factors instead of seven, the wing
was the chief effect related to the flight time of the helicopter, as was seen in this experiment. From there, the
similarities end.
The key thing to remember when comparing the
two experiments is that Box used a 16th fractional factorial, which heavily confounds all of the interaction
terms. However, the main effect terms are only confounded with third-order interactions, and from his
stated goal of trying to teach experimental design, this
was a great route to take.
Another major difference between the two experiments is that the Box experiment used another 4 feet of
drop, as his flight was over 12 feet instead of the 8 used
here. This led to a linear series of time effects, as they
related to the flight time. However, given that the drop
was significantly further, comparing these numbers
with the effects generated in this experiment is not a
fair representation of the data. However, Boxs experiment can be compared to itself.
One difference was that Box found that a heavier
bond paper promoted a longer flight time. A variety
of reasons could be attributed to this finding, the most
probable being that a heavier paper would be more
rigid, and would have less of a tendency to disintegrate from its shape in flight. It is possible that with
the 8-foot drop, the bond paper never had the chance
to fully utilize the benefits of the heavier paper.
However, this accounted for only 0.13 seconds of
additional flight time according to Box, and in this

Downloaded By: [Popovich, Edward] At: 19:25 21 December 2009

442

J. A. Johnson et al.

experiment, the paper type was actually determined by


an interaction.
Another difference according to Box was the paper
clip. Box claims that the paper clip added another 0.05
seconds to the flight time from 12 feet, whereas in this
experiment, it was found to add excess weight and drag
the helicopter faster than without the clip. Engineers in
this experiment found that the non-clipped helicopters
opened up a bit and acted more as a parachute on the
way down. Given the shorter drop, this prevented a
potential death-spiral, as the paper did not have time
to leave its lazy fall, and turn sideways to plummet
to the ground. Box did not have this luxury, as his helicopters had an extra 4 feet to drop, which could have
given the helicopter more time to roll over and plummet to the ground. However, this clipping decision
was also accounted for by an interaction, and the main
effect was disregarded.
The third differing result was that of the joint taping. Boxs information shows that there is a 0.17 second
gain by taping the wing joint. Given the heavier paper
used, this may have caused the wings to not fold up
as much during the drop, but this is purely speculation.
The final difference was that Box chose the wider
body helicopter, as it adds 0.02 seconds to the flight
time. In this experiment, the body width was determined by an interaction.
What do these differences mean? Given that these
were two separate experiments, with different drop
heights, it suggests that these are two different, yet correct, solutions to the same problem. The experiment ran
here was geared for a short time drop. However, Boxs
experiment was for a much longer period, and as a
result, the helicopter had to behave differently. Keeping
in line with the aircraft theme, saying these helicopter
designs should be the same would be like saying that
a carrier-based fighter and a cargo plane should have
the same design, which is far from the truth, but the differences maximize the capabilities of both.

ABOUT THE AUTHORS


Adam Johnson graduated from Auburn University
in December of 2000 with a Bachelor of Science in
Mechanical Engineering. He is currently a candidate
for a Master of Business Administration at the University of Miami. He plans on pursuing a Ph.D. in Statistics.
He is a Six Sigma Black Belt.
Scott Widener graduated from Iowa State University in May of 2001, with a Bachelor of Science in Ceramic Engineering, and again from the University of
Miami in May of 2003 with a Master of Business

Administration in Quality Management. He is a Six


Sigma Black Belt.
Dr. Popovich is Vice President of Process Excellence at Boca Raton Community Hospital. Formerly,
he was a consultant to many large manufacturing companies, financial services companies, medical services
organizations, state and county government, foreign
armed services, and overseas organization. Clients
have included GE, Motorola, Lockheed Martin,
Johnson & Johnson, Intuit, Sony, Bell South, First
Data, Singapore Air Force, Hanley Hazelden Center
at St. Marys, NCR, [P Morgan, Wachovia, A&P,
and Samsung, among others. Dr. Popoivch earned
his Ph.D.] in statistics (1983), his Masters degree
(1979) and his B.S. in mathematics (1977) at the
University of Florida. He has taught at the University
of Florida, University of Central Florida, Florida
Atlantic University, Nova Southeastern University
and guest-lectured at several other universities.
Dr. Howard S. Gitlow is Executive Director of the
Institute for the Study of Quality and a Professor of
Management Science, University of Miami, Coral
Gables, Florida. He received his Ph.D. in Statistics
(1974), M.B.A. (1972), and B.S. in Statistics (1969)
from New York University. Dr. Gitlow is a Six Sigma
Master Black Belt, a Fellow the American Society for
Quality, and a member of the American Statistical
Association. He has consulted on quality, productivity,
and related matters with many organization, including
several Fortune 500 companies. Dr. Gitlow has coauthored 10 books and over 50 academic articles. While
at the University of Miami, Dr. Gitlow has received
awards for Outstanding Teaching, Outstanding Writing, and Outstanding Published Research.

REFERENCES
Box, G. E. P. (1992). Teaching engineers experimental design
with a paper helicopter. Quality Engineering, 4(3):
453459.
Gitlow, H., Oppenheim, A., Oppenheim, R. (1995). Quality
Management: Tools and Methods for Improvement, 2nd
ed. Boston, MA: Irwin McGraw-Hill.
Johnson, A., Widener, S., Popovich, E., Gitlow, H. (2006).
A six sigma black belt case study: G.E.P. Boxs Paper
Helicopter Experiement Part A. Quality Engineering,
18(4):413430.
Rasis, D., Gitlow, H., Popovich, E. (2002). Paper organizers
international: A fictitious six sigma green belt case study
Part 1. Quality Engineering, 15(1):127145.
Rasis, D., Gitlow, H., Popovich, E. (2002). Paper organizers
international: A fictitious six sigma green belt case study
Part 2. Quality Engineering, 15(2):259274.

Você também pode gostar