close

Вход

Забыли?

вход по аккаунту

?

An application of Taguchi's fractional factorial DOE technique for the hand soldering process at Hughes Aircraft Company's microwave division

код для вставкиСкачать
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type o f computer printer.
The quality of this reproduction is dependent upon the quality of the
copy subm itted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back of the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.
UMI
A Bell & Howell Information Company
300 North Zed) Road, Ann Arbor MI 48106-1346 USA
313/761-4700 800/521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
AN APPLICATION OF
TAGUCHPS FRACTIONAL FACTORIAL
DOE TECHNIQUE FOR THE HAND SOLDERING
PROCESS AT HUGHES AIRCRAFT COMPANYS
MICROWAVE DIVISION
A Thesis
Presented
to the Faculty of
California State University Dominguez Hills
In
partial
Fulfillment
of the Requirements for the Degree
Master of Science
in
Quality Assurance
by
Michael Richard Hinz
Fall 1995
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 1379646
Copyright 1995 by
Hinz, Michael Richard
All rights reserved.
UMI Microform 1379646
Copyright 1996, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Copyright by
MICHAEL RICHARD HINZ
12/ 14/ 95
All Rights Reserved
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
THESIS:
AN APPLICATION O F TAGUCHI'S FRACTIONAL
FACTORIAL DOE TECHNIQUE FOR THE HAND
SOLDERING PROCESS AT HUGHES AIRCRAFT
COMPANYS MICROWAVE DIVISION
AUTHOR:
Michael Richard Hinz
APPROVED:
Eugene Watson, Ph.D.
Thesis Committee Chair
Mr. Daniel F. Dunahay
Committee Member
Mr. William Trappen PE
Committee Member
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ACKNOWLEDGMENTS
I would like to thank my wife, Colleen, for her support and
understanding during the many years that I have spent studying for
my different degrees.
Without her by my side, I would never have
been able to accomplish my goals.
I also thank her for her love and
understanding during the preparation of this thesis.
Her
encouragement and understanding will always be greatly and warmly
appreciated.
I must also thank my two young children Natalie and
Christopher.
While they did not truly understand what Daddy was
doing all the time, they still allowed me the quiet time that was so
often needed.
Without their cooperation I would never have been
able to complete my work.
Happily, their question "Work or school
Daddy?" is now no longer answered with a "yes, work and school."
I also want to thank the different members of the DOE team that
helped set up and run the DOE. The team members were, Ed Alba,
Jeff Bertea, Lan Tso, and Dan Wilson.
In addition, I must thank Julia
Ramirez for building the actual PWBs and Bernadine Clark for
inspecting the PWBs.
Their hard work helped make the DOE the
success it was.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
LIST OF TABLES.............................. ..................................................................v i i i
LIST OF FIGURES................................................................................................... ix
CHAPTER 1
INTRODUCTION.................................................................................................1
Information about the Microwave Department............................1
How Costs can be Reduced................................................................. 2
CHAPTER 2
A SHORT HISTORY OF DOE.............................................................................8
D e fin itio n s ...............................................................................................8
Types of DOE's........................................................................................ 9
Fisher and Taguchi.....................................................................................9
Fractional Factorial Experiments.............................................................1 0
Taguchi's Signal-To-Noise (S/N) Ratio.......................................... 11
The ANOVA Table...................................................................................... 1 2
The Basic Steps For Setting Up and Running a D O E.................... 1 3
CHAPTER 3
THE SOLDERING DOE....................................................................................... 1 8
The Start of the DOE.............................................................................1 9
The Information Gathering Stage............................................................ 21
Getting the Go Ahead and the Budget............................................. 2 6
Identifying Factors And Levels To Select The Array................ 3 0
The Planning Stage............................................................................... 3 6
Designing for Robustness................................................................... 4 0
Organizing the Supplies......................................................................4 3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 4
CONDUCTING THE DOE...................................................................................4 5
CHAPTER 5
THE INSPECTION PHASE...............................................................................4 9
CHAPTER 6
THE DATA ANALYSIS PHASE..........................................................................51
The DOE Results.....................................................................................53
The Use of the ANOVA Table............................................................. 59
The Results of the Analysis.............................................................. 6 4
CHAPTER 7
RESULTS IMPLEMENTATION.......................................................................... 6 7
Changing the Departments Supplies............................................... 6 7
Monitoring the Results of the New Process................................. 6 8
CHAPTER 8
CONCLUSION.....................................................................................................69
Continuing Evaluations....................................................................... 71
WORKS CITED...................................................................................................... 7 2
BIBLIOGRAPHY.................................................................................................... 7 4
APPENDIX A ......................................................................................................... 7 6
D e fin itio n s .............................................................................................. 7 6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
vii
APPENDIX B.......................................................................................................... 81
Guidelines for Brainstorming........................................................... 81
Leaders Rules for Brainstorming.....................................................81
Team Make-Up.......................................................................................... 81
APPENDIX C.......................................................................................................... 8 2
Transferring L16 Matrix Into the L16 Array.......................................8 2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
viii
LIST OF TABLES
TABLE NUMBER AND NAME
PAGE
1
Pre DOE Defect Data Ending 6/30/95 (26 Weeks)...............................2 3
2
Selected Factors for Use in the DOE..................................................31
3
Factors, Levels, Degrees of Freedom, and Required Columns
4
L16 Orthogonal Array Matrix and Column Effects.........................3 4
5
Final Settings for Factors and Levels for each Run........................... 3 7
6
Lead Count per Part Type and Total Leads per Run............................ 41
7
Reference Designator Codes................................................................. 4 4
8
DOE Results for - Total Defects..........................................................5 4
9
Best Set of Levels Using Total Defects............................................5 5
10
Best Levels for Total Defects Using Level Averaging................. 5 7
11
ANOVA Results for Total Defects.......................................................6 2
12
Best Set of Levels for the Soldering Process.................................6 5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
33
UST OF FIGURES
RGURE NUMBER AND NAME
PAGE
1
Classic Model of Optimum Quality Costs.................................................. 5
2
New Model of Optimum Quality Costs.................................................6
3
Four Week X-Bar Chart............................................................................. 2 4
4
Four W eek R Chart..................................................................................... 2 5
5
26 Week X-Bar Chart.....................................................................................2 7
6
Tip Temperature - Solder Type Interaction......................................6 4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ABSTRACT
A design of experiments (DOE) was conducted on the hand soldering
process of surface mounted electronic components at Hughes
Aircraft Company.
The DOE was performed to determine which
factors effect the soldering process so that they could be controlled
to reduce assembly costs and increase the soldering yield.
Taguchi's fractional factorial design techniques were used for
the DOE.
The DOE was designed using five factors and a modified L16
orthogonal array.
The five factors used were: flux type and age,
solder type, soldering iron tip size, soldering iron tip cleaning
method, and soldering iron tip temperature.
The levels used ranged
from four levels for flux to two levels for soldering iron tip
temperature.
The results of the DOE were analyzed and the solder
process was changed accordingly.
results.
A proofing run verified the
A team is now monitoring the long term effects of the
changes made to the soldering process.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
CHAPTER 1
INTRODUCTION
Information About the Microwave Department
The Microwave Department at Hughes builds complex
multi-layered microwave PWBs and modules that are used in
Receiver, Transmitter, and Antenna radar units.
The units are part
of the weapons guidance and control systems for the F-15,
F-15 MSIP, F/A-18, F/A-18 RUG, and AV8B (Harrier Jump Jet)
fighter jets.
The Department also makes receiver modules for the
weapons systems on the B2 stealth bomber.
The Department is also
working hard at obtaining new programs in other fields such as
microwave communications.
The Microwave Department is very concerned with the higher
cost of production in California as compared to other areas in the
U.S.
It was only a few years ago when all the work for the different
radar programs was performed in El Segundo.
Now our Department is
the only radar production department left in El Segundo.
Presently,
the number of production employees in El Segundo is approximately
300, down from over 3,000 in the 1980s.
One of the main reasons why the Department is still in
El Segundo is the complexity, both mechanically and electrically, of
the hardware that is produced there.
However, there is no doubt that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0
if a cheaper production facility could be found that was capable of
doing this type of complex work, the remaining jobs at the
Microwave Division would also be gone.
How Costs Can be Reduced
Since soldering encompasses over 95% of the work done in the
Microwave Division, a reduction in the amount of solder defects
would result in substantial savings for the Department.
not new to the Department or to its management.
This fact is
For many years the
Microwave Department has been one of the leading Departments in
the company in regard to employee involvement techniques.
The
Department has been, and still is, a strong supporter of the work
cell and product action team concepts.
The Department presently
has over 12 active employee involvement teams.
While these teams
have been working on problems with the soldering process, they are
also busy working on other production related problems.
Because of
this, it was determined that a "special team" should be put together
to work exclusively on solder defects.
It was also determined that
this team would provide a ten times reduction in soldering defects.
In May 1995 a solder yield team was put together to study the
problems with the soldering process and to find ways to meet the
defect reduction goal.
After several meetings, the team established
a three prong approach to attack the solder defects.
The three
prongs are: (1) "Assembler Awareness and Accountability" which
consists of a training program, workcell activities, operator testing
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3
and certification, and operator defect accountability, (2) "Hand
soldering Process Improvement" which consists of a hand soldering
design of experiments (DOE) and training, and (3) "Solder Process
Enhancement" which consists of the implementation of a mass
reflow process.
Besides reducing defects and saving costs, the Department is
also hoping to become known as the "center of excellence" in
complex hand soldering.
If the Department were able to accomplish
both of these goals there would be two potential benefits.
Not only
could the Department prevent work from being moved out of state,
there would be a good chance of increasing the work base for the
Department in the future.
To help obtain these goals, the Department is investigating new
processes like mass solder reflow techniques.
This includes vapor
phase and thermal conductivity methods of mass solder reflow.
Even
with these new processes, the yield for conventional hand soldering
must still increase to reduce the over all cost of production.
Quality data reports show that there are 4,200 solder defects
per year across the entire Department.
While Hughes has a Cost of
Quality system (COQ), exact cost figures could not be obtained by the
team.
Even without the exact number, the team knows that the
Department's low inspection yields and the high volume of related
rework results in high failure costs for the Department.
To analyze
the Department's cost position the team used the cost of quality
cost curve.
Because of the low inspection yields and the high failure
costs, the team believes that the Department is not at the optimum
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
point on the "Cost of Quality Curve" (Campanella, 1990). This fact is
true for both the classic and the new model curves.
The classic
model is shown in figure 1. The team believes that there is not
enough money being effectively spent on prevention costs in the
Department.
The team also believes that the money that is being
spent is not producing the desired results -
fewer defects.
In the
same respect, the team believes that there is too much money being
spent on, or related to, failure costs.
inspection costs, and re-test costs.
prevention spending,
This includes rework cost, re­
The high failure costs and low
support the team's conclusion that the
Department is to the left of the optimum point on the classic model.
It should also be pointed out that the team believes that the
"New Model of Optimum Quality Costs" is actually the correct or
"real life" model when it comes to plotting actual quality costs.
The
team believes that by effectively increasing prevention spending,
failure costs will decline at a much faster rate.
is lower total quality costs.
The result of this
This also moves the optimum point all
the way to the right where total quality costs are also at a
minimum.
This phenomenon is illustrated by the new model of
optimum quality costs.
shown in figure 2.
The new model of optimum quality costs is
While the team believes in the new model, the
team used the classic model to explain the problems with the
Department because it helps make the point clearer and easier for
the reader to understand.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 1
Classic Model of Optimum Quality Costs
TOTAL
QUALITY
COSTS
u.
FAILURE
COSTS -
ILI
CL
COSTS OF APPRAISAL
PLUS PREVENTION
DEPARTMENTS
PRESENT
^
POSITION
OPTIMUM
POINT
100
QUALITY OF CONFORMANCE, %
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6
Figure 2
New Model of Optimum Quality Costs
o
13
Q
O
CC
a.
u.
O
=>
a
O
O
(3
cc
iu
0.
H
(0
O
O
TOTAL
UALITY
COSTS
NOTE:
THE
OPTIMUM
POINT
MOVED
RIGHT TO
100% QOC
FAILURE
COSTS
COSTS OF APPRAISAL
PLUS PREVENTION
QUALITY OF CONFORMANCE, %
100
The team believes that the DOE will help push the Department
towards the right on the quality cost curve.
The DOE should push the
Department past the optimum point on the classic curve and all the
way to the right on the new model curve where it truly belongs.
While the team was unable to obtain all the cost of quality
information it wanted, it was able to obtain most of the
Department's quality data reports relating to inspection defects.
The team reviewed these reports and they showed that the top four
soldering defects are burnt flux, inadequate wetting, disturbed or
fractured solder, and excess solder -
heel fillet.
The fifth highest
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
defect was "other" which is an assortment of many different defects
which do not have their own categories.
Using this data the team
decided that the most effective way to cut costs for the Department
would be to focus the DOE on both the over all quantity of defects
and on the four highest defects themselves.
After implementation of the DOE's findings, the team believes
that the resulting savings from reduced defects will allow the
Department to remain competitive and compete for additional work
in the future.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
s
CHAPTER 2
A SHORT HISTORY OF DOE
Before covering the details of the actual DOE which was
conducted for this thesis, a brief review of the history of DOEs will
be given.
The following information is not intended to cover all the
different areas in the DOE field.
It is intended to cover the aspects
of DOE that were used in this thesis.
D efinitions
While this paper is intended to be read by someone that is
familiar with the DOE concept, it has been written in a manner
which will allow less experienced readers to better comprehend the
material.
However, to truly benefit from this paper, the reader must
be able to understand the basic concepts of DOE.
In addition, the
reader must understand the many different definitions that relate to
the various terms used in the DOE field.
To help the reader, these
terms and their definitions have been gathered together and can be
found in Appendix A.
at this time.
The reader is encouraged to review Appendix A
To future benefit the reader, the appendix was
compiled to cover many different aspects of DOEs.
This should help
the reader continue to advance in the DOE field.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9
Types of DOEs
There are many different types of DOEs, each used for different
reasons and in different circumstances.
The different DOEs can be
classified into one of three main categories.
These categories are
Classical, Taguchi, and Shainin (Bhote, 1991).
The classical methods
range from fractional factorials to evolutionary optimization
(EVOP).
Classical methods include Plackett-Burman, Box-Behnken,
and Central Composite.
The Taguchi methods use orthogonal arrays
(inner and outer) in "tolerance design," employing analysis of
variance and signal-to-noise for statistical evaluation.
The Shainin
methods use multi-vari, components search, paired comparisons,
variable search, full factorials, B versus C, and scatter plots.
All three of these methods are far superior to conventional SPC.
They also are far superior to the old-fashioned experiments in which
one variable at a time was varied while holding all others constant.
The main problem with the old-fashioned experimental method is
that it did not take into account the interaction between variables.
Because of these omissions the other methods were needed and
developed.
Fisher and Taauchi
Experimental design techniques developed by R. A. Fisher, date
back to the 1920s (Evans & Lindsay 1993).
However, Fisher’s
traditional methods of experimental design were not widely used.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
Fisher's traditional methods were not used because they required a
‘very large number of runs to account for all the variables and their
interactions.
In the late 1970s and early 1980s, Professor Genichi
Taguchi, a Japanese engineer, recognized the problems with Fishers
techniques and developed new approaches to designing experiments.
The method he developed is known as the Taguchi method.
By
concentrating on the critical factors and de-emphasizing their
interactions, Taguchi was able to greatly reduce the number of runs
required.
Taguchi did this by using what is called the fractional
factorial design and orthogonal arrays.
This DOE method saves both
time and money without reducing the reliability of the results.
Fractional Factorial Experiments
The fractional factorial design using orthogonal arrays is a
compromise between the use of all the information, which relates to
the cost of the experiment, and the value or confidence in the
experiment's results.
This can be done by designing the experiment
so that the effects and interactions that are believed to be
unimportant to the experiment are eliminated before the experiment
begins.
This reduces the number of runs required in the experiment
from the start.
For example, a full factorial experiment with 15
factors at 2 levels each would require ( 2 15) test runs which equals
32,768 runs.
A fractional factorial experiment with the same
number of factors and levels using an orthogonal array requires only
16 runs. This can be done because it has been proven that the higher
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
11
order interactions are unlikely to have any engineering or statistical
significance and can usually be eliminated.
However, it must be
pointed out that if the wrong effects (factors) or interactions are
discarded the entire experiment could provide wrong or misleading
results.
Because of this, it is very important to select the correct
inputs for the DOE.
A major feature of Taguchi's utilization of orthogonal arrays is
the flexibility and capability for assigning a number of variables to
them.
An even more important feature, however, is the
"reproducibility" or "repeatability" of the conclusions drawn from
small scale experiments (Taguchi & Wu, 1987).
Taguchi's Siqnal-To-Noise fS/NI Ratio
Another key contribution that Taguchi made was the concept of
signal-to-noise (S/N) ratio.
In this concept the effect or signal is
measured by its mean value, while the variability of the signal is
measured by the standard deviation and represents the effect of
noise.
This means that the S/N ratio is the ratio of the mean to the
standard deviation.
The advantage of this measurement is that it
takes into account both the controllable and uncontrollable factors.
The S/N ratio addresses both the mean and the variation and
therefore can give you more information in one number.
Taguchi's
S/N ratio is used on DOEs that have repetitive results for the same
run.
The bottom line is the higher the S/N ratio the better.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
Taguchi's methods have been proven successful in many different
disciplines.
For example, ITT Avionics Division, a leading producer
of electronic warfare systems, has successfully used Taguchi
methods (Nordwall, 1987).
It is estimated that ITT conducted over
2,000 Taguchi experiments between 1984 and 1986 to improve
production processes.
Xerox and Ford have also been using Taguchi
methods since the early 1980s.
Design Of Experiments, and related
statistical modeling methods such as Sources of Variation Analysis
and Multiple Correlation provide a systematic approach to speeding
up the needed technical solutions.
Product cycles have been
compressed by factors of 3 to 4 with their use (Whitman, 1994).
The ANOVA Table
The ANOVA (analysis of variance) table, also developed by
Fisher, is used after the different experimental runs designed by the
DOE have been conducted. The ANOVA table is used to analyze the
results of a DOE.
The term analysis of variance describes a
technique whereby the total variation is analyzed or divided into
meaningful components.
The ANOVA results indicate which factors
or interactions are significant and depending on the results which
ones should be controlled,
if the ANOVA results in no significant
factors or interactions, the DOE was either designed with the wrong
inputs, the experiment was not conducted correctly, or
other" column is significant.
the "all
If the latter is the case, the
experimenter must determine what in the "all others" is significant
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and redesign the DOE around these factors.
While this result is not
desirable it does not mean the DOE was a total failure.
There are
things that can be learned from what is not significant as well as
from what is.
conducted.
This may however.require additional DOEs to be
The extra DOEs and the extra cost they represent can be
prevented through proper designs before the DOE even starts.
The Basic Steos For Setting U p and Running a DOE
It cannot be over emphasized that the most important step in
any DOE is in the organization or design of the DOE itself.
The
following is a summarization of the basic steps that need to be
followed when setting up and conducting a DOE(Bendell & Pridmore,
1 9 89).
1)
Brainstorming
-
The gathering together of information using the relevant,
knowledgeable, and expert contributors.
This step or
concept should be made a subset for all the other steps.
2)
Developing a statement of basic objectives Insures that everyone understands what is wanted.
Don't
confuse goals (the ultimate result of a task) with
objectives (a statement of how the task will be
accomplished).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
3)
Collection of relevant process information and hypotheses Helps collect ideas, concerns, and sets up the hypotheses
for the DOE.
4)
Formulation of detailed objectives Define the objectives.
Include who the customer is and how
the results are to be measured.
5)
Identifying and distinguishing factor types List and identify the types of Noise and Control Factors.
Select the factors appropriate and discard factors that are
unlikely to have any significance.
6)
Identifying Noise Factors List and identify the inter-noise factors that are
"between - product noise."
List and identify any
outer-noise factors that are "post production."
7)
Identifying Control Factors List Signal factors that effect the "level of the response."
List the Control factors that effect the "variability of the
response."
8)
Identifying
Interactions
-
Determine and list which factors will effect or interact
with the other factors.
the
Select the main interactions from
list.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9)
Selecting the orthogonal array Select the appropriate array that will fit the factors and
their interactions.
This selection must be a compromise
between cost and confidence in the DOE's results.
The
bigger the array the more confident the result, but it also
means a higher cost for the DOE.
1 0 ) Conducting the DOE Take the time to do it right and monitor the tests to insure
the results are true to the DOE.
1 1 ) Analyzing the results Analyze the results using tables, plots, S/N ratios, ANOVA
tables.
Do not forget common sense.
1 2 ) Interpreting of results Consider trade-offs.
Consider trial and error between
optimum levels of factors.
1 3 ) Confirming the results Run a confirmatory experiment using the factors and levels
chosen. This can be done by running another DOE or by more
runs at the selected factor levels.
The additional DOE
however is the preferred way.
1 4 ) Reporting on the DOE's results Cost out the effect of the changes.
Provide a detailed
report with the basic finding and as much background
information as possible.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15)
Acting on the DOE's results Insure that the findings are made known and acted upon.
Without the action all the work is totally wasted.
These steps can be summarized as follows.
Items 1 through 8
are part of what is normally called the planning stage which is the
first stage of the DOE.
Item 9 is the second stage which involves
selecting the appropriate orthogonal array for use in the DOE.
third stage consists of item 10.
DOE.
The
This is the actual running of the
Items 11 and 12 make up the fourth stage which consists of
analyzing the data.
The fifth stage is the conformation stage and
this is covered in item 13.
This stage normally consists of a
proofing run using the selected set of factors and levels.
and final stage is the documentation stage.
of the DOE are documented and archived.
all the efforts are recorded for future use.
The sixth
In this stage the results
This is done to insure that
It is very important to
record and document both the successes and the failures of the
different stages and the DOE itself.
The failure information is
important because it can be used later to prevent the same type of
failures from reoccurring.
Another important aspect in this final
stage is that in this stage the results of the DOE are also put into
action.
Item 14 covers the documentation and item 15 covers the
implementation for this final and very important stage.
All of the stages are important and they all need to be followed
in one form or another to achieve a successful DOE.
A team with a
basic understanding of the concept of DOE, and a working knowledge
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of the process being studied should be able to design a DOE by
following the steps just described.
The time spent in the design
stage of the DOE will more then pay off during the DOE and in the end
results of the DOE.
It has been said that five minutes of planning
equals thirty minutes of work without planning.
This fact is
particularly true for DOEs.
It is also important that all of the members of the DOE team
understand why the DOE is needed, how the DOE is going to be
conducted, and what type of results the DOE is going to provide.
The
team must also know who the DOE is being conducted for (who is the
customer) and what are the benefits that the customer is expecting
from the DOE.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 3
THE SOLDERING DOE
Now that the basic concepts of a DOE have been covered we can
now turn our attention to the main topic of this paper, the soldering
DOE that was conducted at Hughes' Microwave Division in the
summer of 1995.
As explained earlier a solder yield team was established in May
1995 to analyze the soldering process that was being used in the
Microwave Department.
The solder yield team decided that a DOE
was needed because no one truly knew the cause of the soldering
defects or why there was so much variation in the defects.
The
team knew that to understand the process and make any logical
changes to it these questions had to be answered first.
As will be
explained later, the DOE team was tasked to provide 20% of the
overall reduction in soldering defects.
This would allow the solder
yield team to achieve its assigned ten times reduction in soldering
defects.
The team felt that a DOE would reveal the best set of inputs for
the soldering process.
The team hoped that by knowing the best set
of input factors they would be able to reduce the variation in the
defects.
The team also felt that by reducing the variation, the over
all quantity of defects would also reduce.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Before the DOE was started, the team found that different
assemblers were using different materials to do basically the same
job.
The different types of materials being used were flux type, flux
age, solder type, solder iron tip type, solder iron tip temperature,
and solder iron tip cleaning method.
No one knew which set of
materials were correct or which produced the best product.
Interviewing the different operators only revealed that every
operator felt that the set of materials he or she was using was the
best.
One thing that the team knew was that all of the different
methods only added to the variation in the solder process which
added to the variation in the solder defects.
The team was
determined to find the best method for the hand soldering process
that was being used in the Department.
The Start of the DOE
After the team was organized, the final decision to conduct a
DOE was made.
The team followed the basic steps in organizing a
DOE as described earlier.
The first step was to brainstorm.
The
team used brainstorming to work on a statement of the soldering
problem and the opportunity that a solution would present.
The team
also worked on the goal for the DOE.
Describing the problem was easy.
The problem is that the cost
of building products must be reduced in order for the Microwave
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Department to survive.
A slight modification to this fact resulted in
the development of the following problem statement.
PROBLEM STATEMENT:
Hand soldering defects are the largest and most costly
defects in the Microwave Department.
Since hand soldering defects are the largest defects in the
Department, a reduction in them would result in cost saving.
From
this fact, the statement of opportunity was generated.
STATEMENT OF OPPORTUNITY:
By reducing the quantity of soldering defects, the
Microwave Department can reduce its costs and set itself
up as the hand soldering "center of excellence" at Hughes.
In
addition to the savings, this could result in more work for
the Department and increased job security for its
employees.
While the goal for the DOE was easy to determine, reduce the
number of soldering defects, determining a numeric goal was not.
Even though an overall numeric goal was given to the team, achieve a
ten times reduction in soldering defects, the team had to find out
how to obtain it.
By using the three prong approach, the team was
able to break the overall goal into different percentages relating to
each prong.
The team then went back and forth adjusting the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
21
percentages between the different prongs.
Finally, the team was
able to agree on a set of numbers and to agree that the numbers were
realistic and obtainable.
The team decided that the DOE prong should
be able to reduce the defects by approximately 20%.
The team was
then able to establish the following goal.
THE GOAL OF THE DOE:
After implementation of the findings and the
corrective action associated with the soldering DOE,
soldering defects will be reduced by 20%.
The Information Gathering Stage
The next step for the team was to analyze the soldering process
itself.
The members of the team represent many different
disciplines related to the soldering process.
The team consists of
Process Engineers, Quality Engineers, Production Operators, and
Production Management.
The manufacturing experience of each team
member was an important aspect and resulted in a well rounded and
stable team.
The experience of the individual team members ranged
from five to fifteen years.
The combined experience of the group
was well over fifty years.
The diversity in the team's experience
helped the team during this stage.
The team identified the types of soldering defects that were
going to be included in the DOE and studied the history of these
defects.
The team accomplished this by using the quality data
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
reports that are generated by the COQ system.
These reports
consists of the results of all hardware inspections performed by the
Quality Assurance Department on the hardware built by the
Microwave Department.
The quality data showed 1,800 solder
related defects for a six month period ending in June of 1995.
data is shown in table 1.
its defect code.
This
The table shows the name of the defect and
The table has been sorted by defect quantity from
highest to lowest.
One of the rules for the use of a DOE to study a process is that
the process must be in control.
To verify that the soldering process
is in control, the team generated some X - Bar and R - charts.
The
team first grouped the overall data into groups of four week
intervals.
This grouping is one method of generating the "range data”
needed for the R - chart.
The four week X - Bar chart is shown in
figure 3 and figure 4 shows the four week R - chart.
Using the
charts and the data, coupled with a working understanding of the
process, the team concluded that the process was in control.
The team also recognized that the process is in a state of
change due to the work that is being done by the other two prongs of
the soldering defect reduction plan.
The team understood the
changes and determined that the decreasing trend in defects shown
by the X - Bar chart and the decreasing range of the defects shown
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
23
Table 1
Pre DOE Defect Data Ending 6/30/95 (26 weeks)
DEFECT CODES AND DESCRIPTIONS
lli
Li­
U.
» to
u.
DEFECT
CODESDXX
1/6/95
1/13/95
1/20/95
1/27/95
2/3/95
2/10/95
2/17/95
2/24/95
3/3/95
3/10/95
3/17/95
3/24/95
3/31/95
4/7/95
4/14/95
4/21/95
4/28/95
5/5/95
5/12/95
5/19/95
5/26/95
6/2/95
6/9/95
6/1 6 / 9 5
6/23/95
6/30/95
Totals
21
08
04
Ui
49
22
09
u.
u.
u.
u.
20 05 03
06 02 43
28
33
52
17
13
42
35
60
106
183
34
53
61
12
24
12
12
29
27
12
15
100
12
69
31
19
34
62
18
14
30
12
18
24
73
25
58
59
89
82
109
176
86
55
64
12
16
29
12
16
20
43
10
19
101
53
25
10
10
14
15
13
13
515 325 196
187
168 126
109 60 28 24
15 13 10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
36
40
16
1800
Figure 3
Four Week X-Bar Chart
CONTROL CHART (X-BAR) SOLDER DEFECTS
120
100
a
m
LD
LO
LO
LO
S
'T
^
i-
O)
(O
CM
ca
CM
CM
r-
*-
t —
CM
CT
■<*
W
<0
a>
a)
o)
O)
o»
w
a>
FOUR WEEK ENDING DATE
------ ■ —
Defects
“ ““ “ ““ “ MEAN
“ ------------- UCL(X) - ------------- LCL(X)
a 33.11
= 110.55
( X ) * 71.82
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
Figure 4
Four Week R - Chart
CONTROL CHART (R CHART) SOLDER DEFECTS
180
T
160
— — — — — — — — -
to
05
to
05
to
05
in
05
IO
05
IO
05
CM
CM
CM
CM
o>
▼
—
to
y—
CM
CO
to
to
FOUR WEEK ENDING DATE
------ ■ ---
Range
. R=
80.17
___ —
— UCL(R)
> 160.66
—
“ “ — LCL(R)
>0
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
by the R - chart was acceptable.
Using this information, the team
decided that the process was indeed in control and that the DOE
could be conducted. The team then proceeded to generate a X - Bar
chart for the entire range of data.
That is, a X -
Bar chartusing all
26 weeks of data independently and not grouped.This chart is shown
in figure 5.
The team believes that this chart gives a better and
more accurate view of the true process.
The chart also supports the
team's conclusion that the process is in control.
Getting the Go Ahead and the Budget
Now that the process was determined to be
in control,the
team's next step was to get final approval for the DOE from the
Department manager.
This approval also results in a budget being
set up which allows the DOE to proceed.
The team decided that the
best way to sell the DOE would be by showing the Return On
Investment (ROI) that it would provide.
The team knew that if a
savings in money could be shown, which is what management is truly
interested in, it would be able to "sell" the DOE.
The team knew that
it had to talk in the language of management ~ Money.
Since the
team was given a requirement for a ten times reduction in solder
defects, the defect improvement was broken down into the three
different prongs established to accomplish the reduction.
The first
prong which is Assembler awareness and accountability was
targeted for a 30% reduction.
The solder enhancement prong which
is the process change to mass reflow is expected to result
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
27
Figure 5
26 Week X - Bar Chart
CONTROL CHART (X - BAR) SOLDER DEFECTS
ALL WEEK ENDING DATES
2 00
-T___________________________________________________
IO
cn
in
cn
IO
IO
IO
a>
at
at
IO
cn
IO
CO
o
CM
CO
CO
---CO
rT »
▼
—
■r—
▼“
CM
CM
CO
at
CO
CO
IO
cn
'T “
IO
at
IO
cn
to
cn
CM
CO
CM
00
CM
•a
ymm
IO
cn
to
cn
at
CO
CM
CO
IO
IO
CO
WEEK ENDING DATES
Defects
— - - — MEAN —
(X) =. 69.23
—
UCL (X)
=■196.22
m t
i [_CL (X)
=0
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
28
in a 50% reduction in defects.
The last prong which consists of the
DOE is targeted to provide a 20% reduction.
The team then used the
20% reduction target to determine the ROI.
By using the 20% reduction figure and the previously established
number for defects per year, it was determined that the DOE would
result in an annual cost savings of $84,000.
This savings was
determined by multiplying the company's established cost for a
defect by the average number of defects occurring per year.
To find
these values the team used the company's cost of quality system to
determine the cost of a defect.
The COQ system at Hughes produces many different reports for
management.
One of these reports is a report detailing the cost of
different types of defects.
for a solder defect is $100.
This report shows that the average cost
Therefore, the savings is based on an
average cost of $100 per defect and an average of 4,200 defects per
year.
(4,200 defects per yr. X $100 per defect X 20% reduction in
defects = $84,000).
Then determining the ROI, the cost of the DOE also needs to be
known.
The cost of the DOE includes both the cost of labor and the
cost of material.
The cost for the DOE was estimated by the team to
be around $12,500.
This value is the sum of material costs ($2,000)
and both assembly and inspection labor costs ($10,500).
By
subtracting the costs of the DOE ($12,500) from the savings
($84,000) an ROI of $71,500 was obtained.
The team also realizes that there are many hidden failure costs
that are not represented in the COQ's failure cost data reports.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Because of this fact, the team strongly believes that the true ROI
will even be larger.
Examples of these hidden costs which would be
saved are the costs of rework, operator touch-up time, material
handling time, test troubleshooting time, re-test time, and
re-inspection time.
Another important cost is the cost of latent
defects that are not found until the product has advanced through its
production life cycle.
Finally, the most important hidden cost that
can be reduced is the hidden and potentially devastating cost of
unhappy customers.
Countless dollars could be saved if solder
defects and their related failures could be eliminated before they
reach the customer.
After the team had all of the facts and figures together a
meeting was held with the Department manager.
The team presented
the ROI data to the Department manager along with the DOE plan.
After some discussion the team received the go ahead for the DOE.
Organizing the information and collecting data was the next
step for the team.
While the team had already gathered some data
and had established the basic ideas for the DOE, the team was not
able to spend the required amount of time and money on the DOE until
the budget was established.
After receiving the funding, the team
again reviewed the selected soldering defect codes and agreed that
they were the correct defects to attack.
The next step was to
determine the correct input factors and levels for the soldering
process so that the team could select the proper array for the DOE.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30
Identifying Factors And Levels To Select The Array
The team went out to the production area and recorded the
different materials that were being used in the soldering process.
The team also brainstormed the process and the different materials
being used.
In addition, the team contacted different vendors to see
what they recommended.
The result of all of the investigation was
the establishment of the factors, the levels, and the main
interaction for the DOE.
During the investigation stage, the team
selected five factors for use in the DOE.
The five factors that were
selected are shown in table 2.
The team then reviewed the factors to determine the important
interactions.
The team decided that there was one main interaction,
this was the interaction between the solder type and the soldering
iron tip temperature.
This interaction was picked because the team
felt that these two factors and their interaction were main
contributors to burnt flux which is the largest defect for the
soldering process.
The levels required for each of the five factors were also
determined.
Flux type and age was set at four levels, soldering iron
tip cleaning method was set at three levels, soldering iron tip size,
solder type, and soldering iron tip temperature, were all set at two
levels.
The different levels are also shown in table 2.
Determining the size of the array was the teams next move.
When selecting an array, two things have to be considered.
The size
of the array is directly proportional to the amount of confidence
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 2
Selected Factors for Use in the DOE
g.9dS
F)
Name and levels
Flux type and age.
1) a lm it-n e w
2 ) alpha-new
3 ) a lm it-o ld
4 ) alpha-old
C)
Soldering iron tip cleaning method.
1) sponge
2 ) b rillo
T)
3 ) rotary
Soldering iron tip
size.
1) X25 (small tip)
2 ) X38 (large tip)
S)
Solder type.
1) a lm it
2 ) kester
D)
Soldering iron tip temperature.
1) 600 degrees
2 ) 700 degrees
that can be placed in the results.
And the size of the array is
directly proportional to the number of runs required and to the cost
of the DOE.
The team took some time determining the correct array
because there was some debate on the number of interactions that
existed between the factors.
While the team had already determined
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32
the main and most important interaction, the team wanted to include
as many interactions as possible.
The major problem was that when
more interactions were added, the size of the array grew to an
undesirable size.
The factors, the levels, the degrees of freedom, and the number
of columns that are needed for the selected factors and interactions
are shown in table 3.
In order for this DOE to fit into an L16 array
the three level factor, soldering iron tip cleaning method, was
modified by temporarily giving it an extra level.
in the “columns needed" column of the table.
This is shown
This modification
allowed the team to reduce the number of runs required by using an
L16 array.
The table also shows that there are 15 columns required for the
selected interaction and factors.
This tells the team that an L16
array should be large enough for the DOE.
To verify that a L16 will
truly accommodate the requirements of this DOE the factors and
interactions must be assigned to the columns of a L16 matrix.
Column assignments for the L16 array are shown in table 4.
The matrix in table 4 uses the same codes for the factors and the
interaction as table 3.
In order for this DOE to use the L16 array
each level of each factor and interaction must be assigned to a
separate column.
In addition, any interactions between these levels,
must be unique to themselves.
This is done by giving each level the
same code as its respective factor and then assigning each one to a
separate column.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
33
Table 3
Eactors. Levels. Degrees of Freedom, and Required Columns
FACTORS
LEVELS
DEGREES OF COLUMNS
CODE
1
NAME
2
3
4
FREEDOM
NEEDED
FLUX TYPE
AND AGE
3
3
(F) ALMIT NEW ALPHA NEW ALMIT OLD ALPHA OLD
SOLDERING
IRON TIP
SPONGE
BRLLO
ROTARY
2
2+1-3
CLEANING
(C)
SOLDER
ALMIT
KESTER
1
1
(s )
TIP SIZE
02S
038
1
1
fTl
TIP TEMP
600
1
700
1
INTERACTIONS
D 'S
D 'S
1
1
ALL
OTHERS
ALL
5
5
To explain how the column assignments work, flux type and age
coded "F" and the factor soldering iron tip temperature coded "D"
will be used as an example.
Since there are four levels for "F" there
are three dfs requiring three columns.
Because the first "F" was
placed in the first column (it could have gone anywhere) the
last two "Fs" must be placed in relation to the first "F."
shows all of the interactions for column 1.
row was used.
The matrix
In this example the first
This results in the other two "Fs" being assigned
columns 2 and 3.
Now that "F" is assigned in columns 1, 2, and 3
none of the other interactions shown in columns 1, 2, or 3 can be
used.
The next factor to be assigned is "0."
column 7.
It was assigned
The interactions used for "D" were columns 10, and 13.
Therefore none of the other factor’s levels can be assigned columns
10 or 13.
Now that "D" is assigned column 7 with interactions 10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
34
Table 4
L1.6. Orthogonal Array Matrix and Column Effects,
L16 ORTHOGONAL ARRAY MATRIX
COLUMN EFFECTS
FACTOR NAME________________________OCDE LEVELS
df
COLUMNS
NEEDED
FLUX TYPE AND AGE
SOLDERING IRON TIP CLEANING
SOLDER
TIP SEE
TBUP
DXS
ALL OTHERS
3
2
1
1
1
1
S
3
2+1-3
1
1
1
1
5
c
DXS ALL ALL
13 1 4 15
4
3
2
2
2
1
1
(F?
Is )
(T)
(D)
DXS
ALL
L16 ORTHOGONAL ARRAY MATRIX
COLUMN ASSIGNMENTS
F
F
F
C
1
2
3
4
| 2 X 3 | 1 X 3 | 1 X 2 | 1X5
T
5
ALL
6
D
7
C
8
ALL
9
S
10
ALL
11
12
1X4 1X7
1X6
1X9 1X8 1X11 1X10 1X13 1X12 1X151X14
4X7 2X6 2X7 2X4
2X5
2X10 2X11 2X8 2X9 2X14 2X15 2X122X13
5X6 3X7 3X6 3X5
3X4
3X11 3X10 3X9 3X8 3X15 3X14 3X13 3X12
8X1l | 3X12| 8X13 8X14 8X15 ||4X1 2|4X13 4X14 4X15|4 X 8 1 4X9 4X104X11
9X10 9X13 9X12 9X15 9X14
5X13 5X12 5X15 5X14 5X9 5X8 5X11 5X10
112X1312X1412X15 10X14 10X1510X1^1 0X1 3 | 6X14 6X15 6X12 6X13 6X10 6X11 6X8 6X9
|l4X1513X1513X14 11X1511X1411X13 11X12 7X15 7 X U |7 X 1 3 FtxTT 7X1117X1 p|17X9 7X8
4X5
6X7
8X9
10X11
4X6
5X7
8X10
9X11
and 13, when the first "F" was put in column 1 the interactions
of row 3 which is 6 and 7 or row 6 which is 12 and 13 could not have
been used.
If they were used, they would have "crossover" into "Ds"
interaction and the array would not have worked.
In the same
manner since "F" was assigned columns 1, 2, and 3 the first 3 rows
of interactions in column 7 for "D" could not have been used.
If they
were they would have "crossover" into column 1, 2, or 3 which is
already assigned to "F."
As can be seen in the table, the interactions
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
35
between the different levels of each factor do not intermingle.
This
means that the L16 array is large enough to hold the factors and the
interaction for this DOE.
If the interactions did "crossover" into
more than one set of levels for any factor, the selected array would
not work.
The array would have had to be redesigned so that the
factors would fit.
If they could not, a larger array would have had to
be used or the number of factors or levels would have had to be
reduced.
As can be seen from the examples, the first few factors and
their interactions are easy to assign.
much harder.
It is the last few that are
Sometimes a matrix must be redesigned a few times in
order to get everything to fit.
The matrix in table 4 also shows that
all remaining or unused columns are assigned the code "ail."
are five such columns in this DOE.
There
These columns represent the
"within" variability or the "between product noise" of the process.
Other names for this are uncontrolled factors, unknown factors, or
all others.
These columns represent all the other factors and
interactions which effect the process that are not shown in the
assigned columns.
The assignment of columns in this matrix only shows that the
array selected is large enough to accommodate the DOE.
mean that it is the smallest array that could be used.
It does not
Smaller
arrays and matrixes should be tried if it seems that they may work.
The thing to remember is; the larger the array, the more runs and the
more runs, the higher the costs.
usually preferred.
This is why the smallest array is
Using this information and the book "Orthogonal
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
36
Arrays and Linear Graphs" (Taguchi & Wu, 1987) the team selected
the L16 orthogonal array for use in this DOE.
After the array has been selected, the runs must be set up.
This
is done by transferring the factors and the levels from the matrix
into the array itself.
Since the Soldering iron tip cleaning factor
coded "C" was modified to fit into the L16 matrix, an extra step, a
reduction step, is required when placing it into the array.
factors with multiple levels will also be reduced.
The
The matrix in
table 4 shows that the factors use a total of nine columns.
the reduction, the array only has five columns.
After
One for each factor.
The levels for "C" were also modified to show its three levels.
same thing was done for "F" to show its four levels.
The
The
modification steps were done during the reduction steps in this
conversion phase.
These steps are shown in detail in Appendix C.
The final product after the conversion steps is shown in table 5.
The table shows the run order as well as the levels for each factor
determined by the L16 array format.
The Planning Stage
After the run order and levels for each run were determined, the
next thing that the team had to do was to set up the planning for the
DOE.
The planning stage is a very important stage in the design of a
DOE.
If the planning is done incorrectly, the results of the DOE
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
Table 5
Final Settings for Factors and Levels for each Run
RUN
ORDER
1ST
2ND
3RD
4th
5th
6th
7th
8th
9th
10th
11th
12th
13th
14th
15th
16th
FLUX
TYPE AND AGE
ALM IT-NEW
ALMIT-NEW
ALMIT-NEW
ALMIT-NEW
ALPHA-NEW
ALPHA-NEW
ALPHA-NEW
ALPHA-NEW
ALMIT-OLD
ALMIT-OLD
ALMIT-OLD
ALMIT-OLD
ALPHA-OLD
ALPHA-OLD
ALPHA-OLD
APLHA-OLD
FACTORS
CLEANING
TIP
METHOD
SIZE
SPONGE
X25
BRILLO
X25
ROTARY
X38
SPONGE
X38
SPONGE
X25
BRILLO
X25
ROTARY
X38
BRILLO
X38
SPONGE
X38
BRILLO
X38
ROTARY
X25
ROTARY
X25
SPONGE
X38
BRILLO
X38
ROTARY
X25
ROTARY
X25
TIP
TEMP
600
600
700
700
700
700
600
600
700
700
600
600
600
600
700
700
SOLDER
TYPE
ALMIT
KESTER
ALMIT
KESTER
KESTER
ALMIT
KESTER
ALMIT
ALMIT
KESTER
ALMIT
KESTER
KESTER
ALMIT
KESTER
ALMIT
r~
m
<
m
r~
cn
would be unreliable, making the entire DOE a waste of time and
money.
In the planning stage, the team had to decide what materials and
supplies were going to be used.
This included but was not limited to
the type and size of the PWB, the types of components, and the skill
level of the operator that was to be selected to build
the PWBs.
made.
There were numerous other decisions that also had to be
Each decision had to be weighed against its effects on the
results of the DOE, the cost impact of the DOE, and the practicality
of implementing the decision.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
The team also had to decide on the different controls that would
be used during the DOE.
This included the controls for the different
levels as well as the controls for the overall DOE.
Some of the
controls for the factors are; solder dwell time, the amount of flux to
be used, the amount of time between solder tip cleanings, and how
the operator will tin the soldering iron tip.
Some of the controls that the team looked into for the overall
DOE are; the tinning and plating thickness of the PWB circuit pads, if
rework would be allowed on the solder joints (an operator review),
what scope magnification levels would be used for assembly and
inspection, what the specific gravity of the flux should be when
determining its age, how many different squawks would be allowed
on the same solder joint, and if the inspection squawks were final or
would they be reviewed to determine their accuracy by a second
party.
The quantity of solder joints per run was another very important
decision that needed to be made.
The decision on the quantity of
solder joints was based on two things.
First, as the quantity of
solder joints increase, the accuracy or reliability of the DOE results
increase.
Second, as the quantity of solder joints increase, the time
and cost of the DOE increases.
Since the teams first priority was to
find a way to reduce solder defects, the team felt that they had to
insure that the DOE was reliable and provided accurate results.
Because of this, the team's first attempt in determining the correct
amount of solder joints was based solely on the DOE's accuracy and
reliability.
By using the Department's average defects per unit and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
the average size of the units built, the team decided that to get
reliable results, each run should have around 1,500 solder joints.
The team however, understood that this decision would also have to
stand up against the cost factor.
The team worked hard at finding a
compromise between the cost and reliability aspects.
When the team first received permission to conduct the DOE the
team gave management an estimate of two weeks for the assembly
phase and two weeks for the inspection phase.
The team needed to
find out how long it would take to solder 1,500 solder joints per run.
To find out, the team used Hughes' Industrial Standards Department's
time standards for assembly.
The time standards at Hughes allow
for 15 seconds per solder joint.
soldering and cleaning.
This time includes placement,
With this information the team was able to
determine the amount of time the DOE would take to complete.
The
team multiplied the standards by the number of solder joints in the
DOE.
The team then divided the answer by the number of seconds in
an hour to determine the total time.
hours.
This resulted in a total of 100
(15 seconds X 1,500 joints X 16 runs / 3,600 seconds = 100
hours.)
This equals two and one half weeks.
While recognizing that
this number was on the high side of the estimate the team felt that
it was very close to where it should be.
After the team was satisfied with the lead count and build time,
the next step was to decide on the part type and part count for each
run.
The team also knew that the selection of the part type and
count would also result in a fine tuning step for the amount of
solder joints.
While the team would have liked to increase the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
quantity of joints, the team was satisfied with 1,500 per run.
The
concern the team had was that management, looking at cost alone,
would want to decrease the count.
After playing with the numbers
for a while the team came up with a final decision.
The total
number of solder joints and part types per run was broken down and
is shown in table 6.
came to 1,504.
100 hours.
estimated.
The total number of solder joints for each run
The time required for 1,504 joints per run equals
This total is only 20 hours more then that was first
The team decided that this value was a very good
compromise between the value of 1,500 solder joints that it wanted
and the estimated value of two weeks given to management.
team decided to use these values for the DOE.
The
The team reviewed
the numbers with management and was again told to proceed with
the DOE.
Designing for Robustness
One of the goals of the DOE team was to insure that the DOE
would result in a process that would be robust enough to handle the
different products built by the Microwave Department.
To do this,
the team selected six different component part types for use in the
DOE.
These part types were resistors, inductors, capacitors, 14 and
16 leaded IC's, and fine pitch IC's.
Besides the different component
part types, the team selected ten different PWB part numbers for
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
41
Table 6
Lead Count per Part Type and Total Leads per Run
Part TvDe
Lead Count
14
1C
1C
16
Resistor
2
CaDacitor
2
Inductor
2
144
Gate array
Total leads per run
use In the DOE.
QTY
20
60
32
7
21
1
Leads Per Run
280
960
64
14
42
144
1504
Like the components, the PWB part numbers
represented a cross section of all of the different PWBs being used
on the production lines.
While obtaining robustness was very important to the team, it
also resulted in a lot more work for the team.
Because the DOE was
using ten different PWB part numbers, each run required its own
separate set of planning.
The team set up the runs so that each run
used a cross section of all of the different PWBs.
The DOE was also
designed so that each run used the same quantity of the different
part types.
This was done to prevent the different types of
components and PWBs from becoming a factor or variable in the DOE.
The team also went to great efforts to make the assembly planning
easy for the operator to follow.
The team felt that the "user
friendliness" of the planning can be directly related to the number of
errors made in the DOE.
The ease of planning can also directly relate
to the amount of time required to complete the DOE.
The team
continued to work on the premise that time spent in the planning
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
stage can save many times that amount of time in the
implementation or working stage.
Generating inspection planning was the next step for the team.
In order to insure that the inspection results were not biased by the
different runs or the run order, the team decided that the assembly
planning could not be used by the inspector.
Because of this, a
completely different set of inspection planning had to be generated.
The team also had to guarantee that the inspector would not know
which solder joint was from which run.
However, for data analysis,
the team also had to insure that each solder joint could be traced
back to the run in which it was made.
This meant that the part
identification, known as the reference designator, had to hide the
different runs from the inspector while still revealing the required
information to the DOE team.
The team accomplished both of these goals by generating 19
different inspection pages.
Each PWB had its own inspection page
identifying all of the parts.
The inspector was instructed to record
each defect using the PWB part number and serial number along with
the reference designator for the part in question.
This insured that
only the team members knew which part went with which run.
The team developed a coding scheme for the reference designators
for use in the DOE.
The codes used for the different components can
be seen in table 7.
By giving each of the different part types
different reference designators, the defects can be sorted in many
different formats.
The letter symbol in the reference designator
relates to the part type.
The location of the part can be determined
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
from the remaining numbers used in the reference designators.
first number relates to the row the part is on.
The
The second and third
numbers relate to the location or position the part holds in the row.
This information can be used to help the team determine if part type
or part location correlates to any of the defects found in the DOE.
Organizing the Supplies
In order for the assembly phase of the DOE to run smoothly the
team organized all of the supplies that were needed.
up 16 different kits, one for each run.
The team set
The operator was given only
one kit at a time to insure that there was no mixing of supplies.
While this step seems easy, any mistake here would invalidate the
results of the entire DOE.
After both the assembly and inspection
planning was completed and verified to be accurate, the team was
ready to start the DOE.
The team selected an operator for the DOE that they felt was
an average operator.
The team did not want the best operator for the
DOE because the best operator may be able to make a bad process
look good.
Using the same logic, the team did not want to use the
worst operator either.
If this was done even a good process could
end up looking bad.
Additionally, the team wanted to find the best
process for the average operator.
By doing this, the better operators
could still use the DOE’s results to even further improve their skills.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
Table 7
Reference Designator Codes
Part Tvoe
14 PIN IC’S
16 PIN IC’S
FINE PITCH ICS
RESISTORS
CAPACITORS
INDUCTORS
Reference Desiqnator Code
CKXX
uxxx
HXXX
RXXX
CXXX
LXXX
The below average operators could also use the results to improve
themselves to become better operators.
And the results would still
be targeted at the largest section of operators, the average
operators.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
CHAPTER 4
CONDUCTING THE DOE
When the DOE was being conducted, one member of the DOE team
was with the operator at all times.
This was done to insure that the
operator followed all of the instructions.
It was also done so that
any questions or problems that might occur would be dealt with
without delay.
Even though setting up coverage would seem to be a
simple task, it turned out to be difficult.
Since each team member
had different meeting schedules, it was very hard to set up a
schedule that did not interfere with all of the different meetings.
In
fact, even after the schedule was established, unplanned meetings
and other emergencies caused the team to have to scramble around
in order to back each other up.
While it was a struggle, the team
managed to provide complete coverage during the DOE.
Another way in which the team tried to help the operator was to
develop a list of operator guidelines and instructions.
The team felt
that these instructions would help minimize any problems or
confusion.
The following is the list that was given to the operator
at the start of the assembly phase of the DOE.
1.
It is important to work in a normal manner.
Do not do any
thing special for the DOE. Operate as you do on an every day
basis.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
46
2.
Only use the supplies and materials given to you by the DOE
team members.
3.
Use the same tools (tweezers, picks, pliers, scope, soldering
iron, etc.) for the entire DOE.
4.
Do not touch up any solder joints.
soldering."
5.
Try to perform "one touch
Do not perform an "operator review."
Do not try to make defects or try extra hard to prevent them.
Defects will occur and they will be related to the different
process that are being used for the different runs.
6.
Do not clean the board until it is reviewed by one of the
members of the DOE team.
7.
If a defect occurs do not rework it.
Notify the DOE member
that is with you to determine if anything should be done.
8.
The defects are being caused by the changes in the different
processes being used.
Defects are required to analyze the
process and are not a result of the operator.
9.
If you notice something different between or during the runs
or differences in the materials being used, notify the DOE
team member.
This type of information is very important
when analyzing the results.
10.
Only use enough flux for the solder joints that are being
soldered at that time.
11.
"Do not pre-flux" the board.
Solder dwell time is to be kept as close to 3 seconds as
possible.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
47
12.
Clean the soldering iron tip as follows.
a. When using the sponge wipe the tip three times.
b. When using the brillo pad wipe the tip four times.
c. When using the rotary cleaner rotate the tip while
leaving the tip in the rotary cleaner for three seconds.
No matter which cleaning method is being used, clean the
entire tip (all four sides).
Notify a DOE team member if the
cleaning does not seem to be working.
13.
Only tin the solder tip with the same type of solder that is
being used for the run.
14.
Wipe off all of the solder from the soldering iron tip when
changing solder type between runs.
15.
The operator is the expert and input from the operator is
very valuable to the DOE team members and to the success of
the DOE itself.
16.
Take it easy, do not worry about the results.
best.
Just do your
The team is reviewing the different processes being
used not the operator.
Besides trying to make the DOE go smoothly, it was important
for the team not to forget about the time element.
The team needed
to keep the time for the assembly phase of the DOE to the forecast
of 100 hours.
While management was very interested in the results
of the DOE, the cost was also a very important issue.
Up to this
point, the team was able to save cost by working on DOE issues
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
48
between other tasks and during the weekly soldering yield team
meetings.
However, there was no way to work around the cost
issue during the assembly phase.
While there were many eyes on
this phase of the DOE, the team knew that all of the eyes did not
belong to supporters of the DOE concept.
Even though there were a lot of things that could have gone
wrong, the assembly phase of the DOE went very smoothly.
While the operator started out slower than expected, the
assembly pace increased after the operator became familiar and
more comfortable with the DOE requirements.
The operator was
able to complete the assembly phase in 88 hours.
This was 12 hours
less than the 100 hour estimate given management during the
planning phase.
The team was also very satisfied by the fact that
there were no planning errors or parts problems during this
phase. This helped speed up the DOE and keep the assembly phase
under its estimate.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
49
CHAPTER 5
THE INSPECTION PHASE
The inspection phase was not started until the operator had
finished building all of the PWBs.
The team held off on the
inspection phase because the team did not want anything to effect or
influence the assembly phase.
inspection for burnt flux.
The only exception to this was the
Since in some cases burnt flux can be
cleaned off during the cleaning process, the team's process engineer
inspected the PWBs after each assembly point.
The burnt flux
defects were recorded by the process engineer so that they could be
entered into the data base during the inspection phase.
The team selected one of the best inspectors to inspect the
PWBs.
The
team wantedthe best inspector to insure that all of the
defects were found.
The inspector was told that the inspection was
for an engineering related program.
The inspector was instructed to
squawk the PWBs using the same criteria that is used on production
assemblies.
The only difference was that the inspector was
instructed to squawk all of the defects, even if some of them were
on the same solder joint.
The inspector was not told about the
different runs or that different materials were used.
While the
inspector was "kept in the dark" about as much of the DOE as
possible, like the assembler, the inspector was supported by the DOE
team during the entire inspection phase.
The inspector used the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
same inspection station that is normally used for the inspection of
production hardware.
The same defect criteria, defect codes, and
data collection system was also used.
This allowed the inspector to
work in the same environment so that the inspection results would
not be influenced by any changes in working conditions.
The
inspector recorded all defects in the quality data collection system.
The burnt flux defects that were found by the process engineer were
also added to the data collection system at this time.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 6
THE DATA ANALYSIS PHASE
The data analysis phase of the DOE was not started until the
conclusion of the inspection phase.
Just like the assembly phase,
the team did not want the inspection phase to be affected or
influenced in any way by the data analysis phase.
The team did not
want the inspector to think that the inspection results were being
reviewed.
If this occurred, it could have changed the results.
The first step in the analysis phase was to gather all of the data
that was entered into the data collection system during the
inspection phase.
The team first reviewed the data to determine the
total defect quantity per run.
After this review, the team broke the
data up into different reports.
One report was sorted by defect
types per run.
Another report sorted the defects per run number and
per part type.
A third report sorted the defects per run number and
then per part location.
The team wanted to analyze the results from
every angle to insure that no information was missed.
One very important aspect of this DOE is now that the data has
been gathered and archived, it is available to anyone interested in
using it.
With the use of the reference designator system, the data
could be sorted in numerous report formats.
to be analyzed in many different ways.
This allows the results
This means that the results
will remain useful for many different projects and for many years.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
52
The effects of everything from part type to part location, from PWB
material, to circuit pad orientation can be studied.
All of this information and more can be obtained through the
analysis of the data stored in the quality data system.
The team's
long term plan is to analyze the results of the DOE in many different
ways.
The team wants to see how the factors effect the different
types of defects.
The team will look into the factor's effects on
burnt flux, inadequate wetting, disturbed — fractured solder, and
insufficient solder -
heel fillet.
The team wants to look at these
defects separately and in different combinations.
The team hopes
that by looking at all of the different results a common process can
be found that will help reduce these defects.
The team understands
that this is an aggressive goal and that it will not be accomplished
overnight.
The team also knows that the Department cannot wait the
amount of time that would be needed to find the answer for all of
these questions before corrective actions are taken to improve the
solder process.
Therefore, the team decided that the best thing to
do at this time was to use the DOE results for total defects.
The
team therefore used the initial results for total defects for the
first analysis and for this report.
By doing this, the Department will
be able to realize some positive results quicker.
The team will then
work on the rest of the results at a later date, so that continuous
improvement can be obtained.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
53
The DOE Results
The results of the DOE are shown in table 8.
the total defects for each run.
This table shows
Once the team had the information
shown in table 8 the team was able to begin to analyze the results.
One of the quickest ways at analyzing DOE results is by finding the
run that produces the best value in respect to its output
characteristic.
In this case, since the results represent defects, the
best value is the value with the lowest number.
However, if the
results were a measure of the time-to-failure for some type of
system, the run with the largest value would be the best.
As can be
seen in table 8, run number 5 has the lowest number of total defects
with 20.
This means that out of all of the different runs, this run
was designed with the levels that produced the lowest number of
defects.
The best set of levels from table 8 are listed in table 9 in
an easy to read format.
It would however be a mistake to stop the analysis here and use
the levels of run number 5 to build PWBs.
While using the levels of
run 5 would be better than using the levels of any of the other runs,
there is no guarantee that run 5 provides the very best levels.
The
only thing that the results from this analysis method really means
is that run 5 provides the best set of levels out of the 16 sets used
in the DOE.
These results do not mean that the levels in run 5 are
the absolute best.
One reason for this is that this total defect
method also does not take into account all of the effects of the
interactions between the factors and their levels.
There could still
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
Table 8
DOE Results for - Total Defects
TOTAL DEFECTS
RUN
FLUX
CLEAN
TIP
CBS
SOLDER
TOTAL
NUMBER
TYPE
METHOD
SIZE
TEMP
TYPE
DEFECTS
1ST
ALMIT-NEW
SPCNGE
X25
600
ALMIT
72
2ND
ALMIT-NEW
BRILLO
X25
600
KESTER
75
3RD
ALMIT-NEW
ROTARY
X38
700
ALMIT
282
4th
ALMIT-NEW
SPONGE
X38
700
KESTER
123
5th
ALPHA-NEW
SPONGE
X25
700
KESTER
20
6th
ALPHA-NEW
BRILLO
X25
700
ALMIT
70
7th
ALPHA-NEW
ROTARY
X38
600
KESTER
115
8th
ALPHA-NEW
BRILLO
X38
600
ALMIT
67
9th
ALMIT-OLD
SPCNGE
X38
700
ALMIT
250
10th
ALMIT-OLD
BRILLO
X38
700
KESTER
1 03
11th
ALMIT-OLD
ROTARY
X25
600
ALMIT
95
12th
ALMIT-OLD
ROTARY
X25
600
KESTER
163
13th
ALPHA-OLD
SPONGE
X38
600
KESTER
69
14th
ALPHA-OLD
BRILLO
X38
600
ALMIT
46
15th
ALPHA-OLD
ROTARY
X25
700
KESTER
172
16th
APLHA-OLD
ROTARY
X25
700
ALMIT
151
TOTAL
1873
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
Table 9
Best Set of Levels Using Total Defects
FACTOR
BEST LEVEL
FLUX TYPE
ALPHA-NEW
CLEANING METHOD
SPONGE
TIP SIZE
X25
TIP TEMPERATURE
700 DEGREES
SOLDER TYPE
KESTER
be another combination of levels that were not used in the DOE
which could give better results.
While this method does not result
in an absolute answer it does provide an insight on what may be good
levels and bad levels.
This method of selecting the best set of
levels by using the best run is a very rough estimate for what the
actual best set of levels may be.
This method is only used to get an
idea of which levels may be the best.
designer into the ball park.
This method only gets the DOE
It does not provide any guarantee that
the results are accurate or that they should be relied on.
It is just
the best set out of the given sets that were tried.
Since the total defects method of selecting the best set of
levels, does not guarantee that its results are the absolute best,
another method must be used in parallel with it to help pin point the
best levels.
This method involves the use of level averaging or
average effects (Roy, 1990).
Level averaging is done by finding the
average value of the defects for each level used.
This method is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
56
shown in table 10.
Level averaging is accomplished by separately
adding up the DOE run results for each of the different levels for
each factor.
The total for each level is then divided by the number
of levels that were added together.
To help explain the method, the
levels for the factor flux type and age will be used.
Since the first four runs in this DOE use level 1 for flux age type
and age, the results for the first four runs are added up.
the first four runs is 552.
The sum of
This value is then divided by 4 since level
1 for the factor flux type and age was used four times.
The result is
an average of 138 defects for level 1 of the factor flux type and age.
This answer agrees with the results shown in table 10.
This
procedure is repeated for all the different levels of each factor.
After all of the averages have been found for all of the levels, the
averages of the different levels are compared to each other to find
the best one.
In this case the lowest averages are the best.
As can be seen in table 10, the level averaging method provides
its own set of best levels.
For the level averaging method,
the best set of levels are; alpha-new for flux, brillo for the cleaning
method, X25 for tip size, 600 degrees for the tip temperature, and
kester for solder type.
Like the levels in run 5 for the total defect
method, this new set of levels developed from the level averaging
method is only another possible set of best combinations.
The levels
averaging method however, provides a much higher level of
confidence in its results.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
Table 10
Best. Levels for Total Defects Using Level Averaging
DOE RUN LEVELS AND TOTAL DEFECTS RESULTS
RUN NUMBER
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
FLUX
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
CLEANING
1
2
3
1
1
2
3
2
1
2
3
3
1
2
3
3
SIZE
1
1
2
2
1
1
2
2
2
2
1
1
2
2
1
1
TIP TEMP
1
1
2
2
2
2
1
1
2
2
1
1
1
1
2
2
SOLDER
1
2
1
2
2
1
2
1
1
2
1
2
2
1
2
1
72
75
282
1 23
20
70
115
67
250
1 03
95
163
69
46
172
151
LEVEL AVERAGES USING TOTAL DEFECTS RESULTS
LEVELS
1
2
3
4
FLUX
CLEANING
SIZE | TIP TEMP
1 0 6 . 8 0 1102.2511 8 7 . 7 5
138.00
I 6 8 . 0 0 II 7 2 . 2 0 11131.881 1 4 6 . 3 8
152.75
163.00
109.50
SOLDER
1 129.13
II 1 0 5 . 0 0 I
BEST LEVELS USING TOTAL DEFECTS RESULTS
FACTOR
FLUX TYPE AND AGE
CLEANING METHOD
TIP SIZE
TIP TEMP
SOLDER TYPE
LEVEL
2
2
1
1
2
NAME
ALPHA-NEW
BRILLO
X25
600
KESTER
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I
So far we have looked at the total defects data using two
different analysis methods.
It should be understood that the results
of these two methods do not always give the same results.
One
reason for this is that the fractional factorial, as its name implies
does not include all of the different possible combinations of levels.
Therefore, the averaging of levels method can result in a
combination of levels that were not used for any of the runs in the
DOE.
If this occurs it is obvious that there is no way that the two
different methods can give the same results.
When there is a
difference between these two methods the results from the level
averaging method should always be considered the more reliable and
more accurate.
In this case, the two different methods resulted in two
differences between their sets of best levels.
The difference is that
the overall defect method selected the sponge for the cleaning
method and 700 degrees for soldering iron tip temperature while the
level averaging method selected the brillo for the cleaning and 600
degrees for the soldering iron tip temperature.
All of the other
levels in this case were the same for both methods.
Because the
two methods can give different results, a third method should also
be used when determining the best levels.
This method involves the
use of the ANOVA table.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
59
The Use of the ANOVA Table
The ANOVA is a standard statistical technique which is
routinely used to provide a measure of confidence.
ANOVA
determines the variability (variance) of the data.
From this
variance, confidence is measured.
information in a standard form.
The ANOVA table displays this
The ANOVA table is a very good tool
to use when analyzing DOE results. The ANOVA table provides
information on the sum of squares, the mean squares, and the F ratio.
By using the ANOVA table you can also find out how significant each
factor is in relation to the other factors and how much effect each
factor has on the results.
In order to help analyze the results of the DOE the team decided
to use a computer software program specifically designed to analyze
DOEs.
The team reviewed three DOE software packages which were
available at Hughes before making the final selection.
The team
reviewed a program called "Jump" and one called "BBN/Catalyst."
Both of these software packages included both design and analysis
programs.
However, the software program that the team finally
selected was called, Experimental Analysis Based on the Taguchi
Method (C-P-C Engineering and EDS 1987).
The team selected the
C-P-C package because it was the only software package that allows
the user to modify the columns in the array.
Since this option was
required for this DOE, the C-P-C software was the logical choice for
the team.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
60
The software package was very easy to use.
very easy on screen step by step instructions.
The program had
First, the names and
levels of the factors were entered into the program.
Interactions were entered.
Then the
After listing the name and type of
results that were going to be studied, the program provided the
option of selecting the type and size of the array to be used.
It is in
this step that the program allows for the array to be modified.
The
array was modified to match the factor and level conditions shown
in table 5.
After the array was modified the only thing left to do
was to input the results of the different DOE runs.
After the results are entered the program generates an average
table, an ANOVA table, and average graphs.
The program can even
provide its own set of best levels through a recommended option
command.
The program also has the capability of running signal to
noise analysis on the results when applicable.
By analyzing the ANOVA table the team was able to determine
which factors have the most effect on the DOE results.
To find out
how much confidence can be placed in the conclusion that a factor
effects the results, the F ratio from the ANOVA table is converted to
an alpha value.
This is done by using an F distribution table.
The
significance, or confidence level is then found by subtracting the
alpha value from 1.
In order to obtain the F ratio the mean squares variance must be
compared to the residual or error mean squares variance, also known
as all others.
To develop a residual mean square variance the
unknown factors, referred to as all others in the ANOVA table, are
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the first to be combined.
The factors that have a small contribution
to the total of the sum of squares are also combined with the
residual sum of squares as needed.
This process of disregarding the
contributions of a selected factor is known as pooling (Roy, 1990).
Pooling is usually accomplished by starting with the smallest sum
of squares and continuing with the ones having successively larger
effects.
Taguchi recommends pooling factors until the error or
residual degrees of freedom (DOF) is approximately half the total
DO F (Taguchi, 1987).
After the ANOVA table has been pooled the factors can be
analyzed in regard to their significance.
The ANOVA table results,
after pooling, for the DOE are shown in table 11.
The alpha level and
significance levels are also included in table 11.
As can be see the
table was pooled using only all others.
This resulted in a 6 for the
error DOF which is approximately half the total DOF of 15.
The
factor solder type could have been pooled due to its low sum of
squares value compared to the other sum of squares values.
The
team initially pooled it to see if it affected the ANOVA results.
Since it did not, it was not pooled so that its significance level
could still be seen.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
62
Table 11
ANOVA Results for Total Defects
SIGN.
LEVEL
ANOVA TABLE FOR TOTAL DEFECTS
00L
FACTOR
1 FLUX TYPE & AGE
z CLEANING METHOD
3 TIP SIZE
4 ALL OTHER
5 TIP TEMP
6 ALL OTHER
7 SOLDERTYPE
8 ALL OTHER
9 TEMP X SOLDER
1 0 ALL OTHER
1 1 ALL OTHER
A LL OTHERS (error)
TOTAL
SUM OF
DCF SQUARES
3
16705.188
2
23251.338
1
3 510.562
VAR.
VAR RATIO
a
1 - a
F
MEAN SOS
5568.396 9.77012 .01005 .98995
11625.67
20.398 .00310 .99690
3510.563 6.15952 .04850 .95150
(1 )
1 13747.563
( 1)
1
2328.063
13747.56
2 4 . 1 21
.00370 .99630
2328.623
4.0 8 57 2
.09310 .90690
(1 )
1
14220.563
14220.56
24.9509
.00350 .99650
3419.648
77182.922
569.9414
■ Pooled factors
(1 )
( 1)
6
15
Now that the ANOVA table is complete the factors can be
reviewed for significance.
The higher the significance level the
more confidence is placed on the fact that the factor has an effect
on the results.
values.
The last column in table 11 shows the significance
As can be seen by table 11 cleaning method has the largest
significance value followed by the tip temperature and solder type
interaction.
Soldering iron tip temperature, flux type, solder iron
tip size, and solder type follow in that order.
Now that the significance of the factors are known, the next
step in the analysis is to go back to the results from the level
averaging method.
The ANOVA table is used to support the finding in
the level averaging method.
Since the ANOVA table resulted in very
high significance for cleaning method, tip temperature, and flux
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
63
the best levels from the averaging method should be used for these
factors.
The significance level for tip size fell 5 points to 95.1%
and solder type was all the way down to 90%.
compared to the other factors.
This is very low
This means that these factors are
not as significance or important as the other three.
The next step is to check the interaction between the factors
tip temperature and solder type.
The significance for this
interaction given by the ANOVA table was 99.65%.
This is the
second highest.
To see the effect of the interactions, an additional
step is required.
Level averaging is done on the interactions.
Since
the interactions are at two levels, there are only four different
settings that can occur between them.
For this DOE, each one of
these settings are used in four different runs.
of these settings are averaged and recorded.
The results for each
The results of the
averaging is shown in figure 6 along with the graph of the
interaction between tip temperature and solder type.
The
interactions need to be reviewed when making the final decisions
on the best set of levels.
The graph of figure 6 shows that when using solder level 1 the
defects very from 70 to 188 when the soldering iron tip
temperature is changed form level 1 to level 2.
The graph also
shows that the solder type level 2 only changes from 105 defects
to 104 defects when the tip temperature is changed.
This must be
taken into account when the final best set of levels are selected.
The question is whether or not the large variation in defects
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
64
FIGURE 6
Tip Temperature - Solder Type Interaction
TIP TEMPERATURE LEVEL VS SOLDER LEVEL
INTERACTION
SOLDER
LEVEL 1
SOLDER
LEVEL2
TEMP SOLDER AVG
1
1
70
2
1
105
2
104
2
1
188
2
TIP TEMPERATURE LEVELS
that results between tip temperature levels 1 and 2 are acceptable
or offset by the low number of defects when solder and tip
temperature levels are set at 1.
The Results of the Analysis
By analyzing the results using the three different methods
described in this paper and after analyzing the interaction of
temperature and solder, the team was able to determine the best set
of levels to use for the entire soldering process.
best levels are shown in table 12.
The selected set of
These levels are the over all best
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
65
TABLE 12
Best Set of Levels for the Soldering Process
THE BEST LEVELS FOR THE HANDSOLDERING PROCESS
FACTOR
NAME
LEVEL
FLUX TYPE AND AGE
2
ALPHA * NEW
CLEANING METHOD
2
BRILLO
TIP SIZE
1
X25
TIP TEMP
1
600
SOLDER TYPE
1
ALMIT
set of levels and they are the levels that should be used for the hand
soldering process in the Microwave Department.
All three methods support the decision for the flux type and age
and the soldering iron tip size
The team selected the brillo because
the ANOVA table placed high significance on the cleaning method and
the averaging method selected the brillo as the best level.
There is
no question that the brillo is best, even though the over ail method
selected the sponge.
As stated before, the best run method just
gives an idea of what the best set of levels might be.
It does not
have a lot of support behind its results because it does not take into
account any interactions between the levels of the run.
The same
logic was applied to the selection of 600 degrees for the soldering
iron tip temperature.
It also needs to be noted the low level of
significance that the ANOVA table gave to the factor solder type is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
66
an important issue.
At 90% significance, the solder type might not
seem to matter or effect the results.
However, the interaction that
solder type has with tip temperature changes that assumption.
The
lowest defects occur when solder type and tip temperature are both
at
level 1.
However, if the tip temperature can not be held at level
1, over all defects can be reduced if the solder type and tip
temperature were grouped together into sets.
Solder type level 2
and tip temperature level 2 provides the best combination when the
700 degree tip has to be used.
This could be set number 1.
Solder
type level 1 and tip temperature level 1 provides the best
combination when the 600 degrees tip must be used.
This would be
set number 2.
While the team has decided on using the almit solder at this
time, it is also looking into the feasibility of grouping the kester
solder with the 700 degree tip temperature and the almit with the
600 degree.
This may be required because the team understands that
there are unique PWBs that require the 700 degree tips due to large
ground planes.
The team must review the cost of such a plan.
For
now, since the team is concentrating on overall defects the best
decision is to keep both solder type and soldering iron tip
temperature at level 1.
This decision is subject to change when the
team reviews the DOE results based on the different defect types.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
CHAPTER 7
RESULTS IMPLEMENTATION
Changing the Departments Supplies
As stated before, the DOE is not complete until the
recommendations of the DOE are put into action.
The first thing that
the team did after the results were determined, was to verify the
results with a proofing run.
The team had an operator, that was
identified as having excessive soldering defects build an assembly
using the selected levels from the DOE.
The team reviewed the
assembly and felt that it had no soldering defects.
The assembly
was then inspected and the inspection results were studied.
inspection data showed only two defects.
The
The operators solder
defects were reduced 18 percentage points using the new supplies.
The team felt very confident in these results and started to
implement the new process across the Department.
Before the new
supplies were given to the operators, the team held training classes
with all of the operators to explain why the changes were being
made.
A kit that contained the soldering supplies that matched the
DOE findings as shown in table 12 was given to all of the operators.
All of the old supplies were also collected at that time.
The team
provided support for the operators during the transition from the old
supplies to the new supplies.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
68
Monitoring the Results of the New Process
The next step for the team was to monitor the results of the
new process.
In the short period of time that the new process has
been in place the results have been positive.
The operators like the
new supplies and the quantity of defects has continued to decline.
The team, however, understands that the true results will not be
seen until the operators have used the new supplies for a few
months.
The team plans to review the data after a six month period
to see how the changes effected the Department's defects.
The team
plans to do this by comparing six months of post DOE defect data
with the six months of pre DOE defect data that is shown in table 1.
This comparison will provide a good idea on how the process has
been effected by the implementation of the DOE findings.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
69
CHAPTER 8
CONCLUSION
The DOE accomplished its goal.
It provided the Microwave
Department with scientific data and figures on the hand soldering
process.
The results of the DOE allowed the Department to make
logical changes to the hand soldering process.
These changes will
improve the soldering process and reduce the Departments overall
costs.
While the results of the DOE did not totally surprise the team
members, the team was still very satisfied with them.
At the
beginning of the DOE most of the team members felt that the smaller
soldering iron tip size (X25) and the lower soldering iron tip
temperature (600°) were really the best combination.
before, this was only a opinion.
But as stated
Before the DOE there were no facts
or figures that could back up this opinion.
This meant that any
changes made to the soldering process before the DOE would have
really been only a form of tampering.
Changes made using only
options might have helped the process but they might have hurt it
even more.
Now that the DOE has been completed, the results can be
used to make logical changes to the process.
The Department now
has hard facts and figures to use and to stand behind when changes
to the soldering process are made.
If someone questions the changes
the data will backup the decision.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
70
The results of the DOE will also allow the Department to save
money in more ways than by just reducing the quantity of solder
defects.
After implementation of the DOE's findings, the Department
no longer has to buy two different types of solder and flux.
In
addition, only one type of soldering iron tip cleaner is required.
will reduce ordering costs for the Department.
This
Additional savings
includes savings in inventory and stocking costs.
Another benefit from the DOE is that the PWBs that were built
during the DOE are now being used to help train operators and
inspectors. The PWBs are being used as examples of what a good and
a bad solder joint should look like.
The DOE has provided the
Department with examples of all of the different defects which can
be used for comparison with production assemblies.
For the first
time both Inspection and Production have "real life" examples which
will remain in the Department for years to come.
This will save the
Department money by helping the Department make better decisions
relating to potential non-conformities in the future.
The results of
the DOE and the actual PWBs have provided very important
information to the Department.
Because of this information the
Department can now make educated decisions, backed up by facts and
data, when determining which items are truly defective requiring
reworked and which items are acceptable.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Continuing Evaluations
While the initial set of findings have been implemented, the
team is continuing to study the results of the DOE to see if there are
any other changes that can be made to further improve the soldering
process.
The team has decided to continue to review the data based on
the top four defect codes.
This will allow the team to see if there
are different best levels for different defects.
This type of
information is important in understanding what is occurring in the
soldering process.
The team believes that it will be able to continue
to use the information form this DOE for a long time to come.
If and when the team comes up with different information from
this type of continued analysis of the data, the team will make
additional adjustments to the process.
The team also understands
that it may be necessary to conduct additional, but most likely
smaller, DOEs on the soldering process in the future.
The team also
will run proofing runs before any changes are made to the process.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
72
WORKS CITED
Bendell, A., Disney, J. & Pridmore, W. (1989). Taguchi Methods
Applications in World Industry, (pp. 3 - 1 1 ) New York: IFS
Publishing.
Bhote, K. R. (1991). World Class Quality Using Design of Experiments
to make It Happen, (pp. 49 - 54) New York: AMACOM.
Campanella, J. (1990). Principles of Quality Costs. (2nd ed.,
pp. 11 - 15)
Milwaukee, Wisconsin: ASQC Quality Press.
C-P-C Engineering and EDS (1987). Experimental Analysis Based on
the Taguchi Method Version 1.11 [Computer software].
Evens, J. R. & Lindsay, W. M. (1993) . The Management and Control of
Quality. (2nd ed., p. 477) St. Paul, MN.: West Publishing Company.
Nordwall, B. D. (1987, May 11) . ITT Uses Process Control Methods to
Increase Plant Productivity. Aviation Week & Space Technology,
pp. 69 - 74.
Roy, R. (1990) . A Primer on the Taauchi Method, (pp. 48 - 51). New
York: Van Nostrand Reinhold.
Taguchi, G. (1987). Systems of Experimental Design, (pp. 293 - 295).
New York: UNIPUB, Kraus International Publications.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
73
Taguchi, S. & Wu, Y. (1987). Orthogonal Arrays and Linear Graphs.
(p. ii). Dearborn, Michigan: American Supplier Institute, Inc.
Whitman, C. I. (1994). Speeding Technical Solutions in Particulate
Technology with Design of Experiments and Related Statistical
Methods The International Journal of Powder Metallurgy.
3 Q .-U ), 31 - 45.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
74
BIBLIOGRAPHY
Aguayo, R. (1990). Dr. Demina : The American Who Taught the
Japanese About Quality. New York: Simon & Schuster.
Barker, T. (1985). Quality By Experimental Design.
Marcel Dekker, Inc.
New York:
Bhote, K. "A More Cost Effective Approach to D.O.E. than Taguchi,"
1990 - ASQC Quality Congress Transactions - San Francisco.,
pp. 857 - 862.
Catello, F. & Chalmers J. "A D.O.E. Network in a Manufacturing
Operation," 1989 - ASQC Quality Congress Transactions Toronto, pp. 28 - 33.
Crosby, P. (1979). Quality Is Free.
Company.
New York: Mcgraw-Hill Book
Green, T. & Launsby, R. "Using DOE to Reduce Costs and Improve the
Quality of Microelectronics Manufacturing Processes," IS H M
33.94 .proceedings, pp. 60 - 65.
Johnson, M., Jones, K., & Liou, J. "The Comparison of Response and
Taguchi Methods For Multiple - Response Optimization using
Simulation," 1992 IEEE/CHMT Inf! Electronics Manufacturing
Iachnfltoqy-SymppsiMiiL pp. 15 - 18.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
75
Rustagi, J., Singh, V. & Wong, K. "Statistical Methods In
Manufacturing," 1993 IEEE/CHM T Int'l Electronics Manufacturing
Technology Symposium, pp. 215 - 218.
Taguchi, G. (1989). Introduction to Quality Engineering Designing
Quality into Products and Processes. White Plains, New York:
UNIPUB/Quality Resources.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
76
APPENDIX A
D efinitions
ANOVA-
Analysis of variance. A standard statistical
technique which is routinely used to provide a
measure of confidence. ANOVA determines the
variability (variance) of the data. From this
variance, confidence is measured.
ANOVA TABLE -
A table of information that displays the
contributions of each factor.
Coding -
A mathematical technique used to force the
treatment combinations in a design into
orthogonality.
Control Factors -
The variables one wishes to learn about that
can be adjusted easily. Its level is controlled
in the experiment.
Correlate -
To relate so that each member of one set or
series corresponds to a member or series of
another set.
COQ System -
Cost of Quality System. A system that tracks
and records quality related costs for the
company.
DOE-
Design of Experiment.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
77
Efficient Design -
An experiment that derives the required
information with the least expenditure of
resources.
Experiment -
Leads to understanding via a structured set of
coherent tests that are analyzed as a whole.
Always information oriented.
Factors -
An item to which you make deliberate changes
in the response variable. Process inputs. The
control factors in an experiment.
Goal -
The end result of a study applied to get results.
Interaction -
A result of the non-additivity of two or more
factors on the response variable. The situation
where the change in a response differs
depending on the level of the other factor or
facto rs.
Level -
The value assigned to the changes in the factor.
Noise -
The uncontrollable variation that effects the
responses.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
78
Noise Factors -
The variables which are known to influence a
response which cannot or are chosen not to be
controlled in actual applications. There are
three types.
Outer Noise -
Consists of environmental
conditions.
Inner Noise - The deterioration of machines,
tools, parts, etc.
Between
Product Noise - The variation from piece to
piece.
Orthogonal
Array -
Sets of tables that determine the smallest
amount of runs and their conditions for the
experiment. Often abbreviated OA.
Objective -
A statement of how the task will be
accomplished.
Orthogonal -
Independent mathematically, "balanced",
"separable" or "not mixed."
Orthogonal
Design-
An experimental design constructed to allow
independent analysis of each single factor and
the interactions between all factors.
Printed wiring boards.
PWB -
Reproducibility
-
The ability to product the same result with the
same inputs.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
Replicate -
A repeat of a set of experimental conditions.
Not a re-reading of a value, but an entire new
run.
Repeatability-
The ability to perform the same task
repeatedly.
Required
Information -
That is necessary and sufficient to accomplish
our goals.
Resources -
What we have to derive the required
information. Time is the most valuable and
costly.
Response -
A quantitative value of the measured quality
characteristic. The resulting change in the
output due to a change in an input.
Robust -
Strongly formed or constructed.
Runs -
The different conditions of factor levels that
make up the DOE. Also called Trials.
Signal -
The response itself.
Signal Factor
A factor that influences the average value, but
not the variability in response.
S/N -
Taguchi's Signal to Noise ratio.
SPC-
Statistical Process Control.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
80
Squawk -
A term used at Hughes when an item is
identified as being non conforming. The term
is used when an item needs to be replaced or if
it requires some type of rework action.
Taguchi Design -
A methodology to increase quality by
optimizing system design, parameter design,
and tolerance design.
Test -
One shot go, no-go.
T re a tm e n t
Combination -
Usually success oriented.
The levels of all of the factors at which a test
run is made or simply the set of conditions for
a test in an experiment.
Trials -
The different conditions of factor levels that
make up the DOE. Also called Runs.
Variability -
The common uncontrollable changes or
differences in the response. Also called
variatio n.
Variables -
An item to which you make purposeful changes
in the response variable. Process inputs. The
control factors in an experiment.
Variance -
The distribution around the mean. The
difference between the target value and the
result. Known as variation.
Variation
-
The difference between responses when all
input and conditions are the same.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
APPENDIX B
Guidelines for Brainstorming
Suspend judgment - There are no right or wrong ideas.
Strive for quantity - The more the better.
Generate wild ideas - The "wild" one is often the best one.
Build on the ideas of others - Each idea enhances the next.
Leaders Rules for Brainstorming
Be enthusiastic - Make the Enthusiasm spread.
Capture all the ideas - All ideas have value.
Make sure you have a good skill mix - Results in different views.
Push for quantity - The more the better.
Strictly enforce the rules - This results in more participation.
Keep intensity high - Keep the pace fast and interesting.
Get participation from everybody - Anyone could have a cure.
Ieam_Make.-Up
The Leader - Someone that can take charge but not dominate.
Experts - People that know the system or product or process.
"Semi" experts - People that understand what the experts know.
Implemented - People that can "make it happen."
Technical staff who will run the experiment -
People who do it.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
APPENDIX C
Transferring L16 Matrix Into the L16 Array
STEP #1
TRANSFERRING THE DATA
1)
Only the factors and the factor's levels are copied from the
matrix of table 4 into the array.
from table 4 are also used.
The same column numbers
The interactions and the
"all others" are not transferred.
L16
FACTOR CODES
RUNS/COLUMNS
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
F
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
F
2
1
1
1
1
2
2
2
2
1
1
1
1
2
2
2
2
F
3
1
1
1
1
2
2
2
2
2
2
2
2
1
1
1
1
C
4
1
1
2
2
1
1
2
2
1
1
2
2
1
1
2
2
T
5
1
1
2
2
1
1
2
2
2
2
1
1
2
2
1
1
6
1
1
2
2
2
2
1
1
1
1
2
2
2
2
1
1
D
7
1
1
2
2
2
2
1
1
2
2
1
1
1
1
2
2
C
8
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
9
1
2
1
2
1
2
1
2
2
1
2
1
2
1
2
1
S
C
10 1 1 12 13 14 15
1
1
1
1
1
1
2
2
2
2
2
2
1
1
2
2
2
2
2
1
2
1
1
1
2
2
1
1
2
2
1
1
2
1
1
2
2
2
2
2
1
1
1
1
1
1
2
2
1
1
2
2
1
2
2
1
2
1
2
1
1
2
2
1
2
1
1
1
2
2
1
2
1
2
1
1
2
2
1
2
2
1
1
2
1
2
1
2
1
2
1
1
1
2
2
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
83
STEP #2
REDUCING COLUMNS
1) Columns 1, 2, and 3 are combined into a new "F" column.
A) Each group of similar levels for columns 1, 2, and 3
are given the same level number in the new "F" column.
B) Since the factor "F" has four levels, four separate sets
are assigned.
2) Columns 4, 8, and 12 are combined into a new "C" column.
A) Each group of similar levels for columns 4, 8, and 12
are given the same level number in the new "C" column.
B) Since the factor "C" only has three levels, only three
sets of new levels are assigned in the new "C" column.
The forth set of similar levels are re-assigned one
of the existing three levels.
The level assigned for this
forth set depends on the judgment of the DOE designers.
C) The other factors that only use one column are
not changed.
L16 ARRAY
FACTORS
RUNS'jCOLUMNS
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
F
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
F
2
1
1
1
1
2
2
2
2
1
1
1
1
2
2
2
2
F
3
1
1
1
1
2
2
2
2
2
2
2
2
1
1
1
1
F
1 ,2 ,3
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
C
4
1
1
2
2
1
1
2
2
1
1
2
2
1
1
2
2
C
8
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
C
12
1
2
2
1
1
2
2
1
1
2
2
1
1
2
2
1
C
4 .8 .1 2
1
2
3
1
1
2
3
2
1
2
3
3
1
2
3
3
T
5
1
1
2
2
1
1
2
2
2
2
1
1
2
2
1
1
D
7
1
1
2
2
2
2
1
1
2
2
1
1
1
1
2
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
S
10
1
2
1
2
2
1
2
1
1
2
1
2
2
1
2
1
84
STEP #3
REDUCING THE ARRAY SIZE.
1)
The array is reduced leaving only
the active columns.
L16
FACTORS
RUNSVCOLUM
1ST RUN
2ND RUN
3RD RUN
4TH RUN
5TH RUN
6TH RUN
7TH RUN
8TH RUN
9TH RUN
10TH RUN
11TH RUN
12TH RUN
13TH RUN
14TH RUN
15THRUN
16TH RUN
F
1 ,2 .3
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
C
4 .8 .1 2
1
2
3
1
1
2
3
2
1
2
3
3
1
2
3
3
T
5
1
1
2
2
1
1
2
2
2
2
1
1
2
2
1
1
D
7
1
1
2
2
2
2
1
1
2
2
1
1
1
1
2
2
S
10
1
2
1
2
2
1
2
1
1
2
1
2
2
1
2
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
85
STEP #4
COMPLETING THE ARRAY.
1) The active columns are re-numbered.
L16
FACTORS
RUNS\COL
1ST RUN
2ND RUN
3RD RUN
4TH RUN
STHRUN
6TH RUN
7TH RUN
8TH RUN
9TH RUN
10TH RUN
11TH RUN
12TH RUN
13TH RUN
14TH RUN
1STHRUN
16TH RUN
F
1
1
1
1
1
2
2
2
2
3
3
3
3
4
4
4
4
C
2
1
2
3
1
1
2
3
2
1
2
3
3
1
2
3
3
T
3
1
1
2
2
1
1
2
2
2
2
1
1
2
2
1
1
D
4
1
1
2
2
2
2
1
1
2
2
1
1
1
1
2
2
S
5
1
2
1
2
2
1
2
1
1
2
1
2
2
1
2
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
STEP #5
CONVERTING THE ARRAY FROM CODES TO NAMES.
1)
The different letter codes and numbers are replaced with
their corresponding names.
RUN
ORDER
1ST
FLUX
TYPE
ALMIT-NEW
FACTORS
CLEAN
TIP
METHOD
SIZE
SP0N3E
X25
2ND
ALMIT-NEW
BRILLO
3RD
ALMIT-NEW
4th
DBG
TEMP
600
SOLDER
TYPE
ALMIT
X25
600
KESTER
ROTARY
X38
700
ALMIT
ALMIT-NEW
SPONGE
X38
700
KESTER
5th
ALPHA-NEW
SPONGE
X25
700
KESTER
6th
ALPHA-NEW
BRILLO
X25
700
ALMIT
7th
ALPHA-NEW
ROTARY
X38
600
KESTER
8th
ALPHA-NEW
BRILLO
X38
600
ALMIT
9th
ALMIT-OLD
SPONGE
X38
700
ALMIT
10th
ALMIT-OLD
BRILLO
X38
700
KESTER
11th
ALMIT-OLD
ROTARY
X25
600
ALMIT
12th
ALMIT-OLD
ROTARY
X25
600
KESTER
13th
ALPHA-OLD
SPONGE
X38
600
KESTER
14th
ALPHA-OLD
BRILLO
X38
600
ALMIT
15th
ALPHA-OLD
ROTARY
X 25
700
KESTER
16th
APLHA-OLD
ROTARY
X25
700
ALMIT
This is the final format for the L16 array.
LEVELS
It shows the run
order and the factor levels for each run in the DOE.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Документ
Категория
Без категории
Просмотров
0
Размер файла
3 318 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа