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