"I believe this book will help a great deal to clarify misconceptions about Dr. Genichi Taguchi's approach to robust design, such as why dynamic signal-to-noise ratio is used and the role of orthogonal arrays in parameter design and tolerance design. The authors understand the intent of robust design is to prevent fire instead of becoming better fire fighters!"
Ñ Shin Taguchi
President, American Supplier Institute
With practical techniques, real-life examples, and special software, this hands-on book/disk package teaches practicing engineers and students how to use Taguchi Methods and other robust design techniques that focus on engineering processes in optimizing technology and products for better performance under the imperfect conditions of the real world.
The unique WinRobust Lite software included with the book, together with a number of practice problems, enables you to conduct and analyze Taguchi experiments by simplifying the tedious process of performing the many necessary computations.
The book contains complete information on the process of engineering robust products that are insensitive to sources of variability in manufacturing and customer use. You will find detailed instructions for planning, designing, conducting, and analyzing the experiments that are used to optimize a product's performance under a variety of "stressed" conditions. An entire section focuses on designing products that achieve additivity, the property that reduces negative interactions. In addition, the book offers a systematic method for optimizing cost, quality, and cycle time. It even discusses the relationship of robust design to such other quality processes as Quality Function Deployment and Six Sigma.
Numerous case studies, taken from the authors' extensive practical experience, illustrate how robust design theories and techniques actually work in the real world of product engineering. With the techniques described in this book as well as the WinRobust Lite software, you will be better able to design robust products that are high-quality, durable, and able to perform well in the marketplace.
1. Introduction to Quality Engineering.
An Overview. The Concept of Noise in Robust Design. Product Reliability and Quality Engineering. What Is Robustness? What Is Quality? On-Target Engineering. How Is Quality Measured? The Phases of Quality Engineering in Product Commercialization. Off-Line Quality Engineering. On-Line Quality Engineering. The Link between Sir Ronald Fisher and Dr. Genichi Taguchi. A Brief History - The Taguchi Method of Quality Engineering. Concluding Remarks. Exercises for Chapter 1.
I. QUALITY ENGINEERING METRICS.2. Introductory Data Analysis for Robust Design.
The Nature of Data. Graphical Methods of Data Analysis. Quantitative Methods of Data Analysis. An Introduction to the Two-Step Optimization Process. Summary. Exercises for Chapter 2.3. The Quality Loss Function.
The Nature of Quality. Relating Performance Distributions to Quality. The Step Function: An Inadequate Description of Quality. The Customer Tolerance. The Quality Loss Function: A Better Description of Quality. The Quality Loss Coefficient. An Example of the Quality Loss Function. The Types of Quality Loss Functions. Loss Function Case Study. Summary. Exercises for Chapter 3.4. The Signal-to-Noise Ratio.
Properties of the S/N Ratio. Derivation of the S/N Ratio. Defining the Signal-to-Noise Ratio from the Mean Square Derivation. Identifying the Scaling Factor. Summary. Exercises for Chapter 4.5. The Static Signal-to-Noise Ratios.
Introduction, Static vs. Dynamic Analysis. The Smaller-the-Better Type Signal-to-Noise Ratio. The Larger-the-Better S/N Ratio. The Operating Window: A Combination of STB and LTB. A Signal-to-Noise Ratio for Probability. The Nominal-the-Best Signal-to-Noise Ratios. Two-Step Optimization. A Comparative Analysis of Type I NTB and Type II NTB. A Note on Notation. Summary. Exercises for Chapter 5.6. The Dynamic Signal-to-Noise Methods and Metrics.
Introduction. The Zero-Point Proportional Case. The Reference-Point Proportional Case. Nonlinear Dynamic Problems. The Double-Dynamic Signal-to-Noise Ratio. Summary. Exercises for Chapter 6.
II. PARAMETER DESIGN.7. Introduction to Designed Experiments.
Experimental Approaches. The Analysis of Means (ANOM). Degrees of Freedom. Full Factorial Arrays. Fractional Factorial Orthogonal Arrays. Summary of Chapter 7. Exercises for Chapter 7.8. Selection of the Quality Characteristics.
Introduction. Engineering Analysis in the Planning Stage. The Ideal Function of the Design. Guidelines for Choosing the Quality Characteristic. Summary: The P-diagram. Exercises for Chapter 8.9. The Selection and Testing of Noise Factors.
Introduction. The Role of Noise Factor - Control Factor Interactions. Experimental Error and Induced Noise. Noise Factors. Choosing the Noise Factors. The Noise Factor Experiment. Analysis of Means for Noise Experiments. Examples. Other Approaches to Studying Noise Factors. Case Study: Noise Experiment on a Film Feeding Device. Summary of Chapter 9. Exercises for Chapter 9.10. The Selection of Control Factors.
Introduction. Selecting Control Factors to Improve Tunability and Robustness. Selecting and Grouping Engineering Parameters to Promote Additivity. Sliding Levels for Control Factors. Example: The Catapult. Example: The Paper Gyrocopter. Summary: The P-diagram. Exercises for Chapter 10.11. The Parameter Optimization Experiment.
Introduction. Dr. Taguchi's Parameter Design Approach. Layout of the Static Experiment. Layout of the Dynamic Experiment. Choosing the Noise Factor Treatment. Choosing the S/N Ratio. Summary of Chapter 11. Exercises for Chapter 11.12. The Analysis and Verification of the Parameter Optimization Experiment.
Introduction. The Data Analysis Procedure. An Example of the Analysis of the Parameter Optimization Experiment. Estimating the Effects of Each Factor Using ANOM. Identifying the Optimum Control Factor Set Points. The Two-Step Optimization Process. The Additive Model. The Predictive Equation. The Verification Tests. Summary: Succeeding at Parameter Design. Exercises for Chapter 12.13. Examples of Parameter Design.
The Ice Water Experiment: Smaller-the-Better. The Gyrocopter Experiment: Dynamic Larger-the-Better. The Catapult Experiment. Conclusion. Exercises for Chapter 13.14. Parameter Design Case Studies.
Introduction. Paper Handling - An Operating Window Example with Two Signal Factors. Improvement of a Capstan Roller Printer Registration. Enhancement of a Camera Zoom Shutter Design. Summary.
III. ADVANCED TOPICS.15. Modifying Orthogonal Arrays.
Introduction. Downgrading a Column. Upgrading a Column. Compound Factors. Summary of Chapter 15. Exercises for Chapter 15.16. Working with Interactions.
The Nature of Interactions in Robust Design. Interactions Defined. How Interactions Are Measured. Degrees of Freedom for Interactions. Setting Up the Experiment When Interactions Are Included. Summary of Chapter 16. Exercises for Chapter 16.17. Analysis of Variance (ANOVA).
Introduction. An Example of the ANOVA Process. Degrees of Freedom. Error Variance and Pooling. Error Variance and Replication. Error Variance and Utilizing Empty Columns. The F-Test. WinRobust Examples. Summary. Exercises for Chapter 17.18. The Relationship of Robust Design to Other Quality Processes.
Quality Function Deployment (QFD) and Robust Design. Design of Experiments and Robust Design. Six Sigma Quality Process and Robust Design. Summary.Appendix A Glossary.
Quality in products and product related processes is now, more than ever, a critical requirement for success in manufacturing. In fact, for many successful companies, such as Motorola, Toyota, Ford, Bausch & Lomb, Xerox, and Kodak, it is fair to say that quality is a corporate priority. These companies have realized that to obtain customer loyalty, their products have to be perceived as nearly flawless. In addition, to be competitive, their product development process must minimize waste, cycle time, and rework.
The practices adopted by companies that are succeeding in the quality competition vary, but two common elements can be found. Careful attention to the customer is absolutely paramount. Products must satisfy a diverse customer base, with features accurately targeted to customer requirements. Technology must serve customer needs and wants, or the latest and greatest widget will languish on the shelf. Also, continuous improvement, applied to both products and business processes, is ubiquitous.
At the Eastman Kodak Company, the authors have been participants in the ongoing effort to improve the equipment development process. The result has been a world-class process for developing products. This process features Quality Function Deployment (QFD) for capturing the voice of the customer, Robust Design (Quality Engineering) to deliver the level of quality demanded by the customer, and a disciplined engineering process for managing the business of product commercialization. Much of the Kodak Equipment Commercialization Process is described in Professor Don Clausing's book Total Quality Development (ASME Press, 1994).
Physics and engineering principles are the basis for beginning a good product design or fixing problems with a design that is already in existence. Any graduate from engineering school knows these fundamental subjects as well. They have been used effectively by many generations of engineers. However, they alone are no longer enough. The current competitive situation requires a disciplined engineering process that ties together the multitude of engineering tools currently being taught and practiced.
The need to further define the process for linking the principles of engineering and physics to commercialization inspired the writing of this book. The authors' experiences in applying Robust Design to mechanical and electrical systems, electrophotographic process optimization, and chemical process optimization at Kodak have demonstrated convincingly that Dr. Taguchi's design optimization techniques are extremely effective in reducing cycle time and rework. Every company that employs Robust Design does so in the context of their own internal culture. Only the books written by Dr. Taguchi follow his views in totality. This book is a reflection of how we have internalized Dr. Taguchi's insights and teachings into our culture at Kodak. In this industrial environment, we have found broad acceptance and a strong willingness to employ Taguchi methods when practiced in an engineering context. This, of course, is exactly how Dr. Taguchi and those who have listened to him over the years approach the topic - as an engineering process. The successes experienced at Kodak and at many other companies we have encountered are derived from Dr. Taguchi's advice T6: "Spend about 80% of your time in engineering analysis and planning and about 20% actually running experiments and evaluating the results."
Recently, engineering process improvement has been introduced into the academic arena. Courses on Robust Design, QFD, Six Sigma, and other quality processes can now be found at an increasing number of schools. Some of the leaders in this new trend include the Massachusetts Institute of Technology, Stanford University, Georgia Institute of Technology, and Michigan Technological University. Rochester Institute of Technology (RIT), where we teach and serve on the Industrial Advisory Board, recently adopted elements of a quality engineering curriculum as mechanical engineering electives. This book is largely based on our experience in teaching the Robust Design course at RIT in an engineering department. This is unique, because much of the academic attention given to Dr. Taguchi's methods has come from the statistics community as a result of Dr. Taguchi's use of empirical statistical techniques, particularly design of experiments. This has led to a misunderstanding of robust design as being statistical in nature. This book takes an entirely fresh look at robust design as an engineering process, where the emphasis is on using engineering analysis to improve product performance.
This book offers simple, yet effective, guidelines on how to practice robust design in the context of a total quality development effort. In these pages, the fundamental metrics of quality engineering are fully developed, and the rationale behind them is explained. Designing low-cost solutions is a given requirement. We discuss the impact of robust design on the cost of a design, as well as how cost and quality can be co-optimized using Dr. Taguchi's Quality Loss Function. The fundamental statistical tools (e.g., design of experiments, analysis of variance, and analysis of interactions) are explained in what we hope will be an intuitive yet mathematically precise way. A healthy balance exists between the statistical sciences and the engineering sciences. In this book, we try to introduce practical insight into the statistical side of Robust Design, while maintaining the hightest priority in basing the experimental approach on sound engineering principles. The most important element of engineering success is clear thinking, planning, analysis, and communication. For this reason, we offer this book primarily as a guide on how to invest your time efficiently in the 80% up-front engineering required, particularly as it pertains to technology development and product commercialization.
The rest of the book is presented in three major parts. The first is an introduction to Quality Engineering Metrics. It consists of Chapters 2 through 6. Robust Design is a data driven process. Chapter 2 goes through Introductory Data Analysis for Robust Design and is presented to establish a context for how data will be treated throughout the rest of the book. Chapter 3 presents the theory and derivation of the various forms of the Quality Loss Function. An application of the quality loss function to tolerance design is also included. Chapter 4 presents the fundamental knowledge behind the Signal-to-Noise Ratio. The static and dynamic signal-to-noise ratios are fully discussed with numerous examples in Chapters 5 and 6, respectively.
The second part delves into the details of the parameter design process with a special emphasis on achieving additivity.1 Additivity is a property of a design that reduces harmful interactions,1 thus simplifying the optimization process. Chapter 7 is a practical Introduction to Designed Experiments. Without the use of designed experiments, the process of optimizing a product becomes a time consuming endeavor laced with rework and unwanted surprises due to interactions. Chapter 8 is focused on a thorough discussion concerning the Selection of the Quality Characteristics. Few choices in the process of quality engineering are as critical as the selection of the physical responses to be measured during the designed experiments. Chapter 9 provides a sound basis for the Selection and Testing of Noise Factors to stress the design during the development of robustness. Constructing viable noise factor experiments is an indispensable step in preparing for credible and realistic optimization experiments. Chapter 10 completes the discussion on the selection of experimental parameters by giving strategies for the Selection of Control Factors. Chapter 11 shows how to lay out the Parameter Optimization Experiment and is followed in Chapter 12 with the Analysis and Verification of the Parameter Optimization Experiment. Quantifying the individual control factor effects on the overall design performance is highly prized information to the engineering team. In summary, Chapters 7-12 are designed to take you through a comprehensive process of planning, experimenting, and verifying optimized parameter performance.
This book is intended to be useful for teaching and learning the principles and practices of robust design. Chapter 13 demonstrates the parameter design process by covering, in detail, three examples that are actually used as workshop problems by the authors during courses in robust design at the Rochester Institute of Technology and at Eastman Kodak Company. These simple examples are good illustrations of the techniques and can be performed by the reader to practice the method. Chapter 14 demonstrates the parameter design process by presenting three actual Kodak case studies, previously unpublished. Real design problems always take on additional complexity that is intentionally avoided in heuristic examples. These case studies show how parameter design is effective at real-life problem solving. The performance improvements are significant and lasting.
The third and final part of the book is geared toward the engineering practitioner who is interested in more advanced techniques of Robust Design. Chapter 15 provides the necessary information to allow the engineering team to modify arrays to aid in the optimization of unique cases of parameter design. Working with Interactions (Chapter 16) is probably the most controversial topic among the methods of Robust Design. We have produced a balanced approach to maintaining statistical validity of experimentation while promoting the use of enineering knowledge and experience during the construction and analysis of designed experiments that may contain interactive control factors. Chapter 17 teaches the method of the Analysis of Variance, and advanced tool for analyzing designed experiments. Chapter 18 completes the book with a discussion of three special topics within the field of Robust Design. They are the relationship of Robust Design to (1) Quality Function Deployment, (2) Classical Design of Experiments, and (3) Six Sigma.
Because the empirical methods of Robust Design require statistical analysis of large amounts of data, WinRobust Lite software is included with this book. Numerous examples are provided to introduce the reader to many helpful features contained in this PC-based Windows software package. This is the first book of its kind to integrate a custom software package with the text. This union with computer-aided Robust Design techniques will provide you with a comprehensive set of tools that will simplify the tedious process of computation, thus freeing your efforts to focus on the essential engineering issues behind the functional performance you seek to optimize.
We have been very fortunate to have had the opportunity to create and teach one of the first undergraduate-level courses in the United States specifically on the topic of Robust Design. This book is, in large part, a product of our course notes. We greatly appreciate the support and encouragement of Bob Merrill, the current chairman, and Ron Amberger, the past chairman, of the Department of Mechanical Engineering Technology within the College of Applied Science and Technology at the Rochester Institute of Technology. Additional support came from the members of the Industrial Advisory Board of the Department of Mechanical Engineering Technology, and in particular from the advisory board chairman, John Shannon of the Bausch & Lomb Corporation. Thanks to each of them.
Special recognition is due to George Walgrove and Tom Foster, engineers who through their frequent and imaginative use of parameter design have helped make Kodak world-class in the Robust Design process. We would also like to thank the individuals who contributed case studies to this book: Chuck Bennett, Marc Bermel, Atsushi Hatakeyama, Shigeomi Koshimizu, Mike Parsons, Allen Rushing, Steve Russell, Markus Weber, Reinhold Weltz, and Mark Zaretsky. We wish we could list all the other contributors who have added their experiences to ours to help make this book possible, but it is impossible to be comprehensive in such a list, so we simply thank them collectively.
Finally, we would like to thank Susan Baruch for her help in preparing the manuscript, our editor Jennifer Joss, our technical reviewers, and the entire staff at Addison-Wesley for their support.