System Design Optimization for Product Manufacturing

System Design Optimization for Product Manufacturing Masataka Yoshimura System Design Optimization for Product Manuf...

Author: Masataka Yoshimura

245 downloads 2276 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

System Design Optimization for Product Manufacturing

Masataka Yoshimura

System Design Optimization for Product Manufacturing

123

Dr.-Ing. Masataka Yoshimura Kyoto University Graduate School of Engineering Dept. Aeronautics & Astronautics 606-8501 Kyoto Japan [email protected]

ISBN 978-1-84996-007-6 e-ISBN 978-1-84996-008-3 DOI 10.1007/978-1-84996-008-3 Springer London Dordrecht Heidelberg New York British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Control Number: 2010921598 c Springer-Verlag London Limited 2010 Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms of licenses issued by the Copyright Licensing Agency. Enquiries concerning reproduction outside those terms should be sent to the publishers. The use of registered names, trademarks, etc., in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and therefore free for general use. The publisher makes no representation, express or implied, with regard to the accuracy of the information contained in this book and cannot accept any legal responsibility or liability for any errors or omissions that may be made. Cover design: eStudioCalamar, Figueres/Berlin Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

This book explains and discusses broad areas concerning product manufacturing technologies from the standpoint of optimization. Manufacturing products of one kind or another is one of the most important and fundamental activities that people do, and the various processes involved directly, or indirectly, affect the daily life and economic wellbeing of countless people around the world. Product manufacturing methods have undergone dramatic development and change during the past 100 years. Now, successful manufacturing requires that a host of factors be taken into account when products are designed and developed, not just to match or surpass keen competition concerning product functions, performances, qualities, and manufacturing costs, but also to meet increasingly stringent requirements concerning product safety, reduced impact upon natural environments, recycling of resources, and satisfaction of subjective factors that are important to the people who use the products. Product manufacturing under these circumstances requires skillful decision-making in scenarios that are more complex and demanding than ever before. Unsophisticated methods that rely on conventional improvements and optimization within specific narrow regions must give way to more preferable and speedy decision-making based on logically precise methods capable of considering a broad range of related factors. To achieve this, the use of optimum system technologies is essential. Optimization concepts and methodologies are often explained by means of mathematical descriptions, but in engineering research and practical use in industries, mathematically strict descriptions are less important than grasping fundamental ways of thinking about complex optimization problems, and employing useful methodologies. The term “optimization” is often easily used despite a lack of clarity in its meaning, such as when a more preferable result is obtained or selected from a range of choices, but the true optimality of the result may be doubtful. However, when fundamental concepts and descriptions of optimization are adopted in practical scenarios, optimized results that are truly optimal are more likely to be achieved. I have often heard from graduates who are active in industrial or economic sectors that the optimization concepts they learned at our labora-

vi

Preface

tory are very useful in their daily work, and that their basic understanding of optimization is in fact more useful than specific optimization tools. As explained above, to generate more preferable designs and freshly attractive products, a harmonious balance must be attained while considering numerous factors, and this requires mastery of optimization concepts and technologies. Furthermore, new product manufacturing technologies are being steadily developed, so a manufacturer that wishes to maintain a competitive advantage must be thoroughly skilled in the use of optimization technologies, and thus able to understand improvement trends and anticipate future directions for development and practical application. In this book, important technologies pertaining to future product manufacturing techniques are included as much as possible. In the first chapter, the product manufacturing paradigm changes that have occurred during the past 100 years or so are explained. These discussions also indicate promising directions for future development of product manufacturing features, particularly the concepts, methodologies, and technologies that enable more preferable product manufacturing. Chapter 2 explains key optimization technologies and the criteria for judging the aptness and quality of the factors being considered. The relationships between criteria and optimization processes are then discussed. The problem of related criteria that often have complex conflicting interrelationships is explained, as is the use of multiobjective Pareto optimum solutions as a tool to cope with the multitude of product optimization details that must be dealt with. In Chap. 3, fundamental concepts and strategies for innovating product manufacturing are described, namely (1) generation from the conceptual design stages, (2) concurrent engineering, and (3) collaboration. The roles that people play are especially important when product manufacturing innovation is a goal, and Chap. 4 describes the relationships between product manufacturing and characteristics of special importance to people. To support decision-making personnel as they grapple with various aspects of product manufacturing, supporting technologies and systems are indispensable, and Chap. 5 clarifies these and explains specific key technologies, such as the important roles that information network systems play. To carry out the best possible decision-making when using a supporting system, suitable optimization methodologies must be chosen. Chapter 6 explains the present state of optimization technologies for product designs, and presents the fundamental methodologies and strategies, including system design optimizations. Achieving more preferable product manufacturing depends on obtaining more preferable decision-making results. In Chap. 7, the basic methodologies for decision-making in product design and manufacturing are explained. Effective collaborative optimization in product design and manufacturing is one of the most promising methodologies for obtaining preferable product design solutions. The details of product manufacturing directly affect the success and sustainability of a broad range of industries, industries which have considerable cultural impact in countries across the world. Chapter 8 discusses advanced product design optimization methodologies that aim to maximize the benefits of creative and collaborative optimization in product design while addressing important and pressing

Preface

vii

social and environmental issues, as well as the cultural impact of product manufacturing. Most of the material included in this book was developed in the course of my work at the Design and Manufacturing Information Systems laboratory and the Optimum Systems Design laboratory at Kyoto University, Japan. I would especially like to thank my Kyoto University colleagues, Prof. Shinji Nishiwaki, Dr. Kazuhiro Izui, and Dr. Masakazu Kobayashi (now at the Toyota Institute of Technology, Nagoya, Japan), as well as the many student members of the above laboratories who worked with me over the years. I am also grateful to Mr. John E. Goodman who edited the English manuscript, and Ms. Hiromi Ishizuka who drew the illustrations included in this book. I am particularly grateful to my wife Machiko, who has given me unwavering support and encouragement over the past decades. Finally, I am grateful to Mr. Anthony Doyle (Springer-Verlag, Ltd., London, UK) for his valuable advice and kind help. Kyoto, Japan October 2009

Masataka Yoshimura

Contents

1 Progression of Product Manufacturing Technologies ....................................1 1.1 Introduction to Product Manufacturing.......................................................1 1.2 Historical Changes in Product Manufacturing Methodology Paradigms....4 Exercises............................................................................................................7 2 Evaluative Criteria for Product Manufacturing and Optimization Fundamentals.........................................................................................................9 2.1 Evaluative Items and Criteria in Product Manufacturing............................9 2.1.1 Product Quality and Product Performance ........................................10 2.1.2 Manufacturing Cost...........................................................................10 2.1.3 Product Demand, Lead Time, Inventory, and Delivery.....................11 2.1.4 Items Pertaining to Production Method.............................................13 2.1.5 Flexibility in Manufacturing..............................................................15 2.1.6 Processing Capability ........................................................................17 2.1.7 Safety and Reliability ........................................................................19 2.1.8 Natural Environment and Natural Resources ....................................20 2.1.9 Mental Satisfaction Level..................................................................21 2.2 Criteria Requirements ...............................................................................22 2.3 Relationships Between Criteria and Optimization ....................................26 Exercises..........................................................................................................32 References........................................................................................................33 3 Technologies for Product Manufacturing Innovation..................................35 3.1 Generation of Better Products from Wider Feasibilities...........................35 3.2 Generation from Conceptual Design Stages .............................................36 3.3 Concurrent Engineering ............................................................................37 3.4 Collaboration ............................................................................................46 Exercises..........................................................................................................55 References........................................................................................................55 4 Involvement of People in Product Manufacturing .......................................57 4.1 Roles of Individuals in Product Manufacturing ........................................57 4.1.1 Human Abilities ................................................................................57

x

Contents

4.1.2 Relationships Between Customers and Manufacturers ..................... 59 4.2 Kansei Engineering................................................................................... 60 4.3 Ergonomics............................................................................................... 69 4.4 Collaboration Circumstances.................................................................... 74 Exercises.......................................................................................................... 77 References ....................................................................................................... 78 5 Product Manufacturing Support Technologies ............................................ 79 5.1 Representative Supporting Systems.......................................................... 79 5.1.1 Product Shape Description Technologies.......................................... 80 5.1.2 Technologies for Analysis of Performance Characteristics............... 83 5.1.3 Technologies that Support Generation of Product Ideas ................... 85 5.1.4 Database Technologies...................................................................... 90 5.1.5 Manufacturing Support Technologies ............................................... 91 5.1.6 Technologies to Acquire Information Concerning Customer Needs......................................................................................................... 107 5.1.7 Technologies Supporting Enterprise Management.......................... 110 5.2 Utilization of Information Technology for Product Manufacturing ....... 111 Exercises........................................................................................................ 114 References ..................................................................................................... 115 6 Optimization Technologies for Product Manufacturing............................ 117 6.1 Fundamental Optimization Technologies and Difficulties in their Application .................................................................................................... 117 6.1.1 Linear Programming Problems ....................................................... 118 6.1.2 Nonlinear Programming Problems and Local Optimum Solutions . 118 6.1.3 Multiobjective Optimization Problems ........................................... 123 6.1.4 Optimization Problems Including Discrete Variables ..................... 126 6.1.5 Genetic Algorithms (GAs) .............................................................. 129 6.1.6 Large Scale Optimization Problems................................................ 130 6.2 Fundamental Strategies for Effectively Applying Optimization Methods ......................................................................................................... 131 6.3 Fundamental System Optimization Approaches..................................... 134 6.3.1 Decision-making Sequence Applied to Task Operations and Optimization of Evaluative Characteristics............................................... 134 6.3.2 Two Stage Integrated Optimization ................................................ 138 6.4 System Design Optimization Strategies.................................................. 145 6.4.1 Features of Machine Product Characteristics and Fundamental Optimization Strategies ............................................................................. 145 6.4.2 Priority Relationships among Characteristics ................................. 148 6.4.3 Creation of Hierarchical Optimization ............................................ 152 6.4.4 Conflicting Relationships Between Characteristics ........................ 153 6.4.5 Construction of Hierarchical Optimization Procedures................... 153 6.4.6 Practical Procedures for Product Optimization ............................... 155

Contents

xi

6.4.7 Discussion Concerning System Design Optimization .....................162 6.5 Optimum Selection Method for Alternative Design Solutions ...............162 Exercises........................................................................................................167 References......................................................................................................167 7 Decision-making Methods.............................................................................171 7.1 Decision-making Difficulties and Fundamentals of Decision-making ...171 7.1.1 Decision-making Difficulties ..........................................................171 7.1.2 Fundamental Schemes to Facilitate Decision-making.....................172 7.2 Fundamentals of Decision-making .........................................................173 7.2.1 Method for Selecting the Best Alternatives when There Are Many Evaluative Factors...........................................................................173 7.2.2 Calculation of Weighting Coefficients for Attributes Using the Pair Comparison Method.....................................................................173 7.2.3 Finding the Best Alternative from Among Several Alternatives Using the Analytic Hierarchy Process (AHP) Method..............................176 7.2.4 Decision-making Using Subjective Probability Under Uncertain Circumstances ...........................................................................................177 7.2.5 Decision-making Considering Personal Preferences of a Decision-maker.......................................................................................179 7.3 Methodologies for Decision-making in Collaborative Circumstances ...181 Exercises........................................................................................................184 References......................................................................................................184 8 Design Optimization for Creativity and Balance........................................185 8.1 Creativity Optimization Based on Collaborative Effort..........................185 8.2 Cultural Impact of Product Manufacturing .............................................189 Exercises........................................................................................................190 References......................................................................................................191 Index ...................................................................................................................193

About the Author Masataka Yoshimura is now Professor Emeritus of Kyoto University. Currently, he is a visiting professor and senior researcher at Waseda University and a visiting professor at Osaka Institute of Technology. Until April of 2009 he was a professor at Kyoto University, where he participated in research and teaching for 34 years. He started his research career on chatter vibration analyses of machine tool structures and pioneered modal analysis methods. He constructed and proposed pioneering methodologies for the integrated evaluation and optimization of product design and product manufacturing factors, starting in the early 1980s. These research activities became the foundation of the concept of concurrent engineering, which achieved global recognition from the 1990s. To take advantage of synergy effects among engineers working in different engineering fields, he proposed concepts and methodologies for effective product design collaboration. He proposed practical methods that integrate aesthetic attributes with conventional objective attributes when evaluating customer satisfaction levels. He constructed and proposed system design optimization methods based on hierarchical multiobjective optimization methods, so that all factors pertaining to product optimizations can be effectively dealt with. His research results have been published in more than 200 English papers, in journals and proceedings of the ASME, IJPR (International Journal of Production Research), CERA (Concurrent Engineering: Research and Applications), WCSMO (World Congress on Structural and Multidisciplinary Optimization), JSME (Japan Society of Mechanical Engineers), and JSPE (Japan Society for Precision Engineering), and elsewhere. He is a fellow of the ASME, JSME, and JSPE, in recognition of his long-term contributions to these societies. He has also published many books concerning product design and manufacturing engineering, and product optimizations. He has influenced and stimulated progress in design engineering activities around the world. He was the ASME Design Automation Conference co-organizer and review coordinator (1994 ̶ 2000), and is a session organizer and an international liaison for the Asian area. He works on the editorial boards of the IJPR, IJPE (International Journal of Performability Engineering), and others. He has received many awards: the Best Paper Award from the Japan Society for Precision Engineering, 1976; the Best Paper Award from the Japan Society for the Promotion of Machine Tool Engineering, 1985 and 1986; the Achievement Award from the Design Engineering Division of JSME, 2004; the Excellent Design Award from the Design Engineering Division of JSME, 2004; Achievement Award from JSME, 2007; the Best Paper award from the Japan Society for Computational Engineering and Science, 2008; and the ASME Design Automation Award, 2009.

1 Progression of Product Manufacturing Technologies

The progression of product manufacturing goals and technologies is introduced through summaries of major product manufacturing developments over approximately the past century. The relationships between produced products and the customers who use them are then discussed. At present, the design scale of sophisticated products ranges from the atomic level and nanotechnology realm (billionths of a meter) to global dimensions, but the importance of preserving a central awareness of human scales is emphasized in the discussion. Next, the changes in product manufacturing paradigms that have occurred during the past century are described and the various forces driving this evolution are explained. The discussion also indicates promising directions for future development of product manufacturing features, particularly the concepts, methodologies, and technologies that enable more preferable product manufacturing scenarios.

1.1 Introduction to Product Manufacturing The primary goals of advanced product manufacturing are to develop and manufacture essential products that fulfill lifestyle needs to the highest degree possible, as well as auxiliary products that make our living more comfortable, efficient and satisfying. Figure 1.1 illustrates examples of products associated with high standards of living. The manufacturing of all products depends on various levels of technologies. In the Stone Age, early people crafted spears and stone tools so that they could kill and process game, gather edible plants, and live as securely as they were able. Such items were developed to fit human hands and operated on a correspondingly human scale.

2


Automobiles

Motorcycle Medical instruments

Trains

Cameras

Airplanes

Robots for physical assistance

Examples of products associated with high standards of living

Copying machines Elevators Escalators

Prefabricated houses Facsimile machines Personal electric products Personal computers

Cellular phones

Fig. 1.1 Examples of products associated with high standards of living

Over centuries and millennia of gradual human progress, innumerable kinds of products have been manufactured. The most advanced consumer products of today are associated with high standards of living, such as vehicles for transportation, electronic equipment for communication, business and leisure, and products for recreation and amusement. This tremendous variety of products and their associated technologies encompass a wide range of scales, from manipulation on an atomic scale, exploiting quantum effects, to monumental enterprises such as the creation of dams or a megalopolis, with the scale of the human body roughly at the center. In the course of progress, more efficient airplanes and trains are designed and built to transport increasing numbers of people to their destinations in shorter times, advanced power plants distribution grids aim to provide a more stable infrastructure, and buildings of increasing scale that incorporate more sophisticated control of materials and climate aim to provide higher levels of comfort. When considering the impact of human activity upon the natural environment and planet as a whole, it is clear that an awareness of issues pertaining to this global scale should be integrated into product design and development processes. On the other hand, it appears that most attention is focused on smaller scales, as products for personal use are increasingly miniaturized to provide greater convenience, utility, and comfort. The realm of nanotechnology is receiving increasing publicity, as researchers uncover ways to incorporate features at the scale of billionths of a meter in practical, everyday products that aim to satisfy requirements for lighter weight, superior functionality, and higher density of parts. On an even smaller scale, some areas of research focus on the atomic and molecular realms,

1.1 Introduction to Product Manufacturing

3

and certain discoveries have already led to important breakthroughs that will soon have profound social impact. Thus, as shown in Fig. 1.2, the scale of current product manufacturing covers a range from atoms and molecules, to household products, and then to cars, trains and planes, skyscrapers, space stations, and even monumental earthworks. Since the design, manufacturing, sale, and use over time of consumer products is almost always associated with rising standards of living, it is vital to preserve a strong awareness of human scales, which lie approximately at the center between the very large and the very small. Product manufacturing that ignores human needs and desires, that is, manufacturing that concentrates too strongly on one particular scale at the expense of the human scale, may turn out to have significant drawbacks or be manifestly harmful. The design and production of successful products almost always depend on a sensitive examination of the relationships of scale between these objects, the surroundings in which they are used, and the people who make them a part of their lives.

Fig. 1.2 Scale range in current product manufacturing

There are the following two major types of products: Products that consumers buy and use Industrial products used to manufacture the products belonging to 1 above Figure 1.3 shows the relationship between customers and the manufacturers of consumer products and the industrial machines used to produce these products. The behavior of customers as they “vote with their wallets” naturally influences the demand for certain products, which in turn affects product manufacturers and supporting industries. As retail sales increase, certain manufacturers flourish and business activity radiates to other manufacturers and business sectors according to the specifics required for the production of the given products. The need to design and develop increasingly useful, attractive, and sophisticated consumer products provides a fundamental stimulus for development and improvement in the manufacturing realm. 1. 2.

4


Customers

Manufacturers of consumer products

Manufacturers of industrial machines

Fig. 1.3 Relationship between customers and manufacturers of consumer products and industrial machines

Figure 1.4 shows a generalized manufacturing flow, which is usually the same for both consumer products and industrial machines. This flow begins with market research and proceeds through product development, product design, product manufacturing, and ends with the sale of the product. Market research

Product development

Product design

Manufacturing

Sales Fig. 1.4 Conventional product manufacturing flow

1.2 Historical Changes in Product Manufacturing Methodology Paradigms Modern methods for manufacturing machine products have been evolving in accord with industrial development. Historical changes in product manufacturing paradigms are shown in Fig.1.5. Figure 1.6 shows the changes in manufacturing methodologies.

1.2 Historical Changes in Product Manufacturing Methodology Paradigms

5

Taylor paradigm

Ford paradigm

JIT paradigm

Lean paradigm

CALS paradigm Fig. 1.5 Changes in manufacturing paradigms

Mass production

Job shop type production

Production to order

Production where customers participate in production designs Fig. 1.6 Changes in product manufacturing methodologies

At the end of the nineteenth century, Frederick Winslow Taylor (1856 ̶ 1915) devised management methods that paid special attention to the efficiency of manufacturing operations. These methods based on the Taylor paradigm were incorporated by US automobile companies of the period as the Ford paradigm so that large numbers of products could be produced at lower manufacturing costs,

6


which made them more accessible to the general populace and brought about the first step toward more affluent lifestyles. Henceforth, economies of scale were increasingly appreciated and taken advantage of by companies improving their manufacturing methodologies. Once people became familiar with mass-produced items, there was an increasing demand for more suitable and more highly developed products that would better fit evolving lifestyles. In order to meet the needs of increasingly prosperous customers, product makers were forced to change manufacturing methods from limited variety mass production to a job shop type of production, where a variety of products could be created to satisfy a range of preferences. The following requirements had to be met: 1. Prompt response to customer needs, development of products having high performance, high quality, and low cost, with as short as possible product development time, while avoiding dead-ends or oversights that waste time, energy, and materials 2. Development of new products that customers want to buy, not simply incremental improvements applied to existing products These requirements imply a change in the product manufacturing paradigm from one of “selling products that are produced,” common when mass production was the norm, to one of “producing products that will sell,” with a focus on fulfilling customer desires. A production method that could cope with such problems was the so-called Just In Time (JIT) method of production in which required quantities of parts are supplied to each process exactly when needed, a production method that can be effectively applied to job shop production. (This approach is explained in detail in Sec. 2.1.) The method was systematized in the United States of America in the latter half of the 1980s as a lean production method that could efficiently cope with demand fluctuations while maintaining production and fiscal efficiency. The JIT and lean production methods are now widely used around the world. In the latter half of the 1990s, a production method based on the CALS paradigm (Continuous Acquisition and Lifecycle Support) was developed, where all data pertaining to a product’s entire lifecycle are processed by computers so that real-time data exchange and decision-making becomes possible, facilitated by increasingly powerful information technology. The original meaning of CALS was replaced by the current meaning, Commerce at Light Speed. At present, the age of job shop manufacturing, in which customers select the most preferable products from a variety of products that makers prepare in advance, is giving way to a manufacturing paradigm that supports making products to order, where specific and detailed customer needs and desires can be economically dealt with at a relatively fine-grained level. Furthermore, we can envision the possibility that manufacturing in the future will eventually integrate some form of customer participation in the product design process, to maximize customer satisfaction, enabling scenarios of customer-maker collaborative manufacturing to some degree.

Exercises

7

In addition to the above two basic requirements concerning response to customer needs and development of novel products, the importance of the following additional points has been growing: 3. Increasing public awareness of adverse effects upon natural environments and depletion of natural resources has made mandatory the consideration of product lifecycle and recycling of parts or raw materials 4. The pursuit of mental as well as physical satisfaction requires design and manufacturing methodologies whose products ultimately suit the emotional and mental aspects of customer needs more closely Modern product manufacturing should, ideally, satisfy all of the foregoing factors, namely, 1 ̶ 4. To achieve these requirements, criteria for evaluating the satisfaction level for each factor are required. Figure 1.6 illustrates changes in product manufacturing paradigms during approximately a 100-year interval, where inter-company competition concerning product development was conducted according to particular criteria relevant during those times. Reflecting progress over time on a number of fronts, criteria for product manufacturing have evolved and become increasingly complex. Below are the main points of product design criteria trends over roughly the past half century: 1. Optimization of a single objective function, such as minimization of the operational time and minimization of the cost under the constraints of product performances and qualities 2. Recognition of the importance of conflicting relationships among characteristics, and use of more flexible optimizations from wider viewpoints, since improvement of a single objective function may cause another characteristic to be degraded 3. Inclusion in product criteria of influence levels of factors pertaining to natural environments, and factors pertaining to resources and recycling 4. Recognition of an increasing need in product design evaluation for inclusion of requirements pertaining to comfort and aesthetic factors when products are used.

Exercises 1.1 Why were mass production methodologies necessary during the early period of modern product manufacturing? How were these mass production methodologies carried out? 1.2 Discuss the problematic aspects of product manufacturing that is based on job shop type production, and methodologies that aim to resolve these problems.

2 Evaluative Criteria for Product Manufacturing and Optimization Fundamentals

This chapter explains key optimization technologies and the criteria for judging the aptness and quality of the factors being considered. First, fundamental criteria and characteristics are explained with reference to particular product manufacturing paradigms and how these have changed over time. The use of satisfaction functions is explained in terms of subjective attributes that can be integrated with sophisticated evaluative techniques to enable product manufacturing improvements now and in the future. The relationships between criteria and optimization processes are then discussed. Special attention is given to clarifying the process of design optimization. Next, developmental improvements in optimization techniques applied to industrial activities are described. The problem of related criteria that often have complex conflicting interrelationships is explored and the use of multiobjective Pareto optimum solutions to handle the multitude of product optimization details is explained. Last, the use of Pareto optimum solution sets in system optimization methodologies to maximize improvements in product design and manufacturing is discussed.

2.1 Evaluative Items and Criteria in Product Manufacturing To obtain optimum product design and manufacturing solutions, evaluative items and criteria for product manufacturing should first be defined, with the specifics depending on the particular nature of the product. The most fundamental items and criteria are described below.

10


2.1.1 Product Quality and Product Performance The aim of product manufacturing is to produce products that fulfill their functions, required performances, qualities, and characteristics. The criteria initially described below pertain to product qualities, which can be classified into two types: design qualities and manufacturing qualities. Design qualities correspond to values that customers require for the product, and, in the case of industrial machines such as machine-tools and industrial robots, these are the accuracies, efficiencies, operational energy requirements, and similar performance aspects. In the case of automobiles, drivability, acceleration and braking performance, fuel economy, crashworthiness, comfort, versatility, aesthetics, and so on, would be considered. On the other hand, manufacturing qualities pertain to the manufacturing processes used when producing products that incorporate desired design qualities. In the case of machine-tools, such qualities would correspond to dimensional variances, surface roughness in joint contact areas, processing accuracy, and so on. To ensure a satisfactory level of product quality, manufacturers must evaluate whether or not their products achieve designated design specifications. Here, variations pertaining to manufacturing processes are the principal evaluative factors. Qualities that customers require in the products they purchase are often regarded as being aspects of product performance. For example, accuracies, when considered as a product performance, correspond to certain levels of precision when the product is used for work or to accomplish its objective. Efficiencies are often evaluated by the time required to complete a task or sequence of operations and a product that can accomplish work more quickly is considered to have higher efficiency.

2.1.2 Manufacturing Cost The next important criterion in product design is the total manufacturing cost, the sum of the various costs required to actually manufacture the product. The material cost of structural members and components, machining costs, casting and forging costs, powder metallurgy costs, the cost of welding, assembly, and so on, are all included in the manufacturing cost. Examples of other costs that are included in the total product cost include labor expenses, overheads, advertising, and so on. Such costs can be clearly evaluated as quantitative magnitudes and, in some cases, optimization formulations can benefit from transforming certain other criteria, such as time, into costs. The most important criterion in the era of mass production shown in Fig. 1.6 in the previous chapter was to reduce the product manufacturing cost as much as

2.1 Evaluative Items and Criteria in Product Manufacturing

11

possible, while achieving the required product performances, and technologies for automating the manufacturing processes and reducing processing times were developed. In the era of job shop production, products having various combinations of product performances and product cost had to be manufactured to meet customer requirements. Production quantities for such tailored products were small, and the advantageous cost reductions enabled by mass production could not be realized. Presently, new technologies to reduce manufacturing costs are being developed and applied in job shop production scenarios, and innovative product manufacturing technologies are discussed in places throughout this book. The manufacturing cost should be minimized for each product performance and, to achieve this, the minimum manufacturing cost for each specific product performance should become a criterion in the product manufacturing optimization.

2.1.3 Product Demand, Lead Time, Inventory, and Delivery In the era of mass production, manufacturers simply produced their products continuously, and customers could usually obtain desired products without delay. In job shop production and production-to-order scenarios, product delivery times, i.e., the interval from when a product is ordered to when it can be delivered to a customer, become especially important. Whether or not customers can quickly obtain the objects of their desires can significantly affect their satisfaction levels, and manufacturers that can produce customized products more rapidly than their competition tend to thrive. Delivery times are often decided prior to manufacturing specific products, and maintaining promised scheduling is a practical requirement for successful product manufacturing. In connection with the importance of product delivery times, the lead time, i.e., the time required from when an order is placed to when the customer received the product, or, alternatively, the time required from when materials are prepared to when product manufacturing is completed, is important in present manufacturing environments. To accomplish reliably shortened delivery times, lead times must be reduced. In the present situation of increasing affluence in many parts of the world, not only are the requirements, needs, and preferences of customers diversified, but these also change over time, sometimes suddenly. To adapt promptly to such changes, manufacturers must bring their products to market as quickly as possible, by shortening product development times, while maintaining or improving required performances and qualities, and also reducing manufacturing costs. In product manufacturing scenarios that depend on a variety of manufacturing processes, reducing the processing time in each stage, processing certain jobs in parallel, and minimizing production mistakes are all required. Figure 2.1 shows the conceptual relationship between the demand and supply curves for a product where time and sales volume are indicated on the horizontal

12


and vertical axis, respectively. The demand for a product usually increases rapidly with time, reaches a peak, and then tapers off.

Fig. 2.1 Product lifecycle

When a maker competes in the marketplace with a product manufactured in response to demand for similar products, the product supply curve normally shows a period of growth, a period of maturity, and a period decline that is similar to the demand curve. In many cases, as shown in Fig. 2.1, the supply curve lags behind the demand curve, and the degree of this time lag can dramatically affect profits, since entering later into a competitive market for a given product will result in lower sales volumes. The region where a loss of potential profit could have been avoided by more rapidly responding to product demand is shown by the shaded area in the left portion of the figure. A related region, indicating a different relationship in the supply and demand curve, is illustrated by the shaded area in the right portion of the figure. Here, as sales volumes decrease, production quantities exceed demand, and costly unsold inventory is the unfortunate result. Year by year, the time spans of demand curves, from the period of growth to the period of decline, seem to be getting shorter, although this depends on the kind of products involved. In job shop production, such time spans can become even shorter as the number of product variations increases. For product manufacturing to be successful, the required volumes of parts and materials must be brought together at the right place and at the right time, so skillful inventory control is a practical necessity. Figure 2.2 shows the relationship between ordered parts or materials and the amount of inventory. When the amount of inventory decreases to a specific volume Q0 , a certain quantity of replacement parts or materials is ordered so that the inventory can be replenished in time without causing a delay in production.


13

Q0

Amount of useless inventory

Time Fig. 2.2 Parts or materials ordering sequence and inventory amounts

As inventory levels increase, the costs of storage and capital outlays increase, but, on the other hand, inventory shortages must be avoided, since interruptions in production can be extremely costly. Thus, minimizing financial loss due to running out of parts or materials, and minimizing the costs of maintaining adequate inventory levels have a conflicting relationship. Furthermore, excess inventory can be viewed as a kind of waste in product manufacturing, especially when this consists of parts or materials that are unlikely to be used.

2.1.4 Items Pertaining to Production Method There are two main production methods, production to stock and production to order. In production to stock, a specified quantity of the final product is maintained at each retail outlet as local inventory, and the makers replenish this inventory as the need arises. On the other hand, in production to order, the manufacturing begins only after product orders are received. To reduce lead times in production to order scenarios, some manufacturers maintain inventories of parts at certain stages of manufacturing, and these items are prepared according to in-house production to stock requirements. Most products require many steps to manufacture, and depend upon a variety of processes. Unless production lines run smoothly, however, shortages of certain parts required for a process may occur, resulting in an accumulation of parts that have been processed at an earlier stage but are unable to be transferred to the next process. To cope with such problems, a production method was devised where required quantities of parts are supplied to each process exactly when needed, the so-called Just in Time (JIT) method of production.

14


The JIT production method is also called the Toyota manufacturing method. Instruction sheets called “kanban” are used to delineate explicitly the routes parts must proceed along, from upstream processes to subsequent stages, and to provide manufacturing instructions applicable to processes along the way. Such production methods aim to minimize manufacturing losses, both in terms of material and time, by avoiding overruns or shortages of certain manufactured products, by preventing bottlenecks where certain parts are held up pending the arrival of other parts, and by minimizing transport and inventory losses. Minimization of these various losses and maximization of efficiency are important criteria in competitive product manufacturing. The Supply Chain Management (SCM) system shown in Fig. 2.3 illustrates an idealized efficient flow of goods and information between adjacent entities such as parts suppliers, product makers, wholesalers, retailers, and consumers, with communication and distribution links connecting the entire chain. This concept attempts to integrate the flow of both goods and information, so that the benefits of JIT manufacturing can be extended to include the critical distribution and sales aspects of the total process. Parts suppliers

Product maker

: Flow of goods

Wholesalers

Retailers

Consumers

: Flow of information

Fig. 2.3 Conceptual diagram of SCM

When production management systems are categorized from the standpoint of information and product flow, there are two main types: push manufacturing methods and pull manufacturing methods. Push manufacturing is a plan leading type of system, where production management functions are centralized and instructions are distributed to each manufacturing process. In such scenarios, production quantities are first determined and then an economical lot size is calculated prior to manufacturing. Acceptable order quantities, production procedures, and production management details are considered under the constraints of equipment availability and production capacity. This production system is mainly used in mass production. Pull manufacturing, in contrast, operates under a demand leading paradigm. Production procedures and production management details are tailored to the specific products for which the demands are known, and then manufactured as efficiently as possible. In such scenarios, problems such as inconsistencies between equipment production rates and in-process inventories, and productivity losses due to increased setup times must be avoided. To avoid these pitfalls and accomplish efficient manufacturing, new manufacturing technologies are required, such as manufacturing different kinds of products using the same equipment, as well as an approach that unifies the handling of materials and information. Here, the concurrent engineering concepts explained in Sec. 3.3 can be of great benefit. In pull


15

manufacturing, the flow of information concerning manufacturing starts from a fully downstream location, reflecting product orders or demand forecasts, and appropriate parts inventories required for each upstream process are sequentially secured. JIT production is an implementation of pull manufacturing. Pull manufacturing is often suitable in job shop manufacturing scenarios.

2.1.5 Flexibility in Manufacturing The requirement for manufacturing flexibility is an outcome of the paradigm change from mass production to job shop operations, where goods are manufactured essentially to order, rather than being manufactured in advance. Thus, the flexibility of manufacturing equipment is increasingly important. A production system where versatile machine tools are connected with conveyance machines is called a Flexible Manufacturing System (FMS). FMS components are flexible machine tools such as NC (Numerical Control) equipment, machining centers, automatic tool exchangers, conveyance machines, flexible work transporting systems, in-process parts storing areas, computer-control systems, and so on. The so-called Flexible Manufacturing Cell (FMC) is a smaller scale version of the FMS. Figure 2.4 shows an FMC example, where machine tools are grouped around an industrial transportation robot [1]. Fully equipping a FMS generally requires a large financial investment, and cannot necessarily cope with changes in manufacturing situations. In such cases, flexibly modified combinations of FMCs can often be used. In situations where flexible manufacturing is required, manufacturing equipment that can be easily configured for a wide range of manufactured quantities are required to cope with fluctuating levels of demand, and a broad range of product variety and volume. To fill the need for such equipment, compound machine tools have been developed in which several processes are integrated so that various jobs or operations can be accomplished by a single machine. Figure 2.5 shows a conceptual diagram of the relationship between manufacturing methods, required flexibility, and facility cost. In conveyor production systems, also called line production systems, workers are deployed along production lines and tasks such as assembly are conducted in sync with the transfer speed of the conveyor. Such systems are costly, but if arrangement of the workers and the sequence of tasks are suitable, good production quantity per unit time can be expected. However, production quantities may be limited by the efficiency of the slowest worker, and since proficiency varies among workers, conveyor production systems tend to have low flexibility.

16


Conveyance machine Machine tool 1

Parts storage area

M1 B2

B1

Industrial robot

Machine

M 2 tool 2

B4 M3

Conveyance machine

Parts storage area

Machine tool 3

Fig. 2.4 FMC composed of an industrial robot and machine tools

Conveyor production system Cell production system

Single - person standing - style production system

Flexibility Fig. 2.5 Conceptual diagram of manufacturing method, flexibility, and facility cost


17

In cell production systems, where the conveyors are excluded from the production line, the different proficiencies among workers can be assessed and the distance between adjacent workers adjusted, with some jobs even overlapped among a number of workers. Thus, with cell production systems, delays in the production flow can be minimized. In single-person standing-style production systems, one worker carries out tasks within a working space, such as assembling meals at a food stand. Such tasks are performed according to the personal ability of the worker, and job proficiency has a more direct relationship to the manufacturing efficiency. This style of manufacturing allows great flexibility with respect to the details of the required tasks, and the equipment cost is minimal. To cope with job shop production demands, cell production systems or single-person standing-style productions system are effective, depending on the size and nature of the products.

2.1.6 Processing Capability Process capability pertains to the maintenance of uniform qualities during the manufacturing process, and is evaluated by measuring variations in the attributes of manufactured workpieces. The characteristics of the products inevitably have variations, and such variations should be considered to be evaluative factors for the product that are as important as the actual product performance. These variations have many causes: such as variations in material properties, variations during manufacturing processes, changes in the humidity or temperature of the environment where the product is used, and so on. Deterioration in product functions and performances due to variations in its characteristics as the product is used must be avoided as far as possible. In quality engineering, one measure of quality is considered to be the minimization of variations in evaluative characteristics. In mass production scenarios, when continuous manufacturing of the same products is carried out, evaluation of variations in characteristics, and preventing variations from exceeding a certain value, are especially important. In job shop production scenarios, the magnitudes of these variations are important competitive factors that can have a major impact on product sales. Tolerance is often used as a criterion when evaluating whether or not the quality of a manufactured part or article is acceptable. Tolerance here is defined as the difference between the upper and lower limits of acceptable measured dimensions, and parts or goods that fail tolerance checks are rejected as unsuitable. Processing capability means the uniformity of the processes being regarded, and is generally measured by the variations of attributes such as the dimensions of parts after they are processed. The variance distribution of the characteristic being regarded is displayed using a histogram, such as that shown in Fig. 2.6. When this

18


distribution can be approximated by a normal distribution, it is expressed using the mean value of μ and the standard deviation of σ . The standard deviation shows the extent of the characteristic distribution, and process capability is usually expressed as ± 3 σ from the mean value μ , or a width of 6 σ , the total extent of ± 3 σ . The probability of the attribute quality lying beyond the 6 σ range is roughly 0.27%. When the tolerance is equivalent to 6 σ , roughly 0.27% of the production output will be considered unsatisfactory.

6σ

-6σ

-3σ -2σ -σ

σ

2σ 3σ

6σ

Mean value μ

Characteristic f Fig. 2.6 Variance distribution of a characteristic expressed as a normal distribution

In the semi-conductor industry, and certain other industries where extremely high quality is mandatory, ± 3 σ may be considered insufficient, so stricter tolerances are employed. In the United States, from about the1980s, Motorola, Inc. proposed their six sigma (6 σ ) campaign (actually ± 6 σ ) to reduce the number of products failing tolerance inspections. General Electric and other companies soon followed, and 6 σ became a slogan in many programs that aimed to improve the quality of goods. In practice, this tolerance was too strict, and ± 4.5 σ was employed instead, yielding a tolerance failure rate of roughly 3 parts per million. To increase processing capability, the use of equipment having higher operational accuracies, higher accuracy machining methods, and higher quality materials is necessary, which almost always increases the manufacturing cost. How to maintain or reduce manufacturing cost while ensuring the required quality is an important technological subject.


19

2.1.7 Safety and Reliability The safety and reliability of products are extremely important criteria in product designs. If an important product has a higher than expected rate of failure in service, much effort is usually devoted to determining the causes and how to prevent future occurrences of similar trouble. The need to consider adequately such issues when products are designed would seem to be common sense, but this is not always the case. Safety evaluations mainly focus on the prevention of harm or injury to people but reliability evaluations aim to ensure that the product functions normally for an expected period of time. Since people may be harmed or injured if the product does not function as planned, a system with lower reliability also tends to be less safe. Recently, the need to respond more quickly to customer needs and desires has made shortening product development time a high priority. Furthermore, computer systems are increasingly used in product design processes so that the number of trial experiments carried out can be minimized. In product design and manufacturing of advanced, large-scale products, a large number of expert engineers often operate within relatively narrow strata of the enterprise in which they work. In such situations, insufficient information sharing, or difficulties encountered during cooperative collaboration, may become significant obstacles to achieving company goals. Similarly, product safety and reliability issues are more likely to arise and will have to be given extra care and attention. The representative criteria concerning reliability are as follows: 1. Reliability: the probability that the product will provide the functions for which it is designed without any trouble over a certain period of time when the product is used as intended under specified conditions. The reliability R(t ) at time t of the product that is operating normally at time t = 0 is defined as follows: R(t ) = Pr {normal throughout the term [ 0, t ] normal at time t = 0} where Pr { A B } expresses the conditional probability of A under the condition B . Under the condition where the product is normal at t = 0, R (t ) is the probability that the product will be free of trouble during the term of use [0, t ]. When the reliability is high, the value of R(t ) approaches 1, and as the reliability becomes lower, R(t ) has smaller values. 2. Unreliability: obtained by subtracting the reliability value from 1, this is the probability that the product will fail to provide expected functions within a certain period of time. 3. Availability: the ratio of the actual working period over the required working period. 4. Failure rate: the ratio of the number of products or elements that develop trouble within a certain time period to the total number of items that has been operating without failing.

20


5. Mean time to failure (MTTF): the mean value of the time passed from the start of operation until failure occurs. 6. Maintainability: the probability for the recovery of the function, that is, the completion of a repair within a specified time limit after the occurrence of trouble. 7. Mean time between failures (MTBF): the mean time between the present trouble and the next trouble, where the time required for the detection of trouble and repairs are included. 8. Mean time to repair (MTTR): the mean time from the occurrence of trouble until the completion of a repair, where the time required for detection of the trouble, the time from when the trouble is detected until the start of the repair operation, the time from the start of the repair operation until the repair is completed, and the time from the repair completion until functionality is recovered are all included.

2.1.8 Natural Environment and Natural Resources Product manufacturing has a tremendous influence on natural environments and has led to a number of catastrophes as well as shortages or exhaustion of natural resources. In response to these concerns, consideration of product lifecycles and the recycling of products and materials have become indispensable aspects of responsible product designs [2]. In mass production scenarios where reduction of the manufacturing cost is the principal manufacturing paradigm, economical products that customers most strongly want to own are not always manufactured in sufficient quantities. In such cases, product makers usually plan to use large amounts of natural resources so that a large number of products can be consumed by as many consumers as possible. Recently, both customers and product manufacturers have recognized the pressing need to recycle natural resources and reuse product parts. In advanced product manufacturing, sufficient consideration of these factors must start from the conceptual design stages. An increasingly attractive trend is for mere product consumption to be replaced by a more complex, but ultimately more attractive, process in which products are designed so that they can be maintained rather than simply discarded, and can be more easily recycled when they are no longer of use. As this type of approach to design and manufacturing becomes more common, the long-term economic benefits will become more apparent. The goal of lifecycle designs is the sustainable development of economical societies, where product manufacturers recognize that minimization of environmental impact and careful use of natural resources make good business sense. Representative methodologies for evaluating lifecycle designs include LifeCycle Costing (LCC) and LifeCycle Assessment (LCA). LCC expresses the total cost of realizing and preserving a product’s required functions. For a simple case, LCC is expressed as follows:


21

LCC = acquisition cost + operation cost + recycling or disposal cost LCA is a methodology for quantitatively and objectively analyzing the impacts and influences upon natural environments during the total lifecycle of a product, starting with the acquisition of natural resources and continuing through manufacturing, transportations, sales, usage, recycling, and ultimate disposal. One of the criteria in product lifecycle designs is given as follows: Φ=

Satisfaction level for society as a whole Total damage to global environments

(2.1)

That is, the ratio of satisfaction levels due to the successful realization of product functions over the consequential impact and damage to natural environments should be maximized to preserve the long-term viability of economical societies and establish truly sustainable lifestyles. Well known terms pertaining to product lifecycle issues are Recycle, Reuse, and Reduce, and together these are often summarized as 3 Rs. Recycle here includes the recycling of caloric energy as well as the recycling of materials, so certain discarded materials can be used as fuel and other recycled materials can be reused as material ingredients for other products. Reuse calls for parts or elements of the original products to be reused as parts or elements in other products. And Reduce literally means reduction, i.e., reduction of product weights, reduction of energy used, elimination of unnecessary functions, prevention of excess production, and so on. Among the 3Rs, Reduce has the highest priority, and is the most directly related with product designs, since advanced designs aim to minimize the use of natural resources and reduce environmental impacts. The secondary highest priority is given to Reuse. Reuse of parts and machine elements brings about material and energy saving during manufacturing. Finally, Recycle is mainly concerned with reducing the need to use virgin natural resources, but it is important not to lose sight of the energy cost of the actual work of recycling products. While rising costs of materials such as metals and oils has made conservation an integral part of product manufacturing, there are still plenty of opportunities to reduce waste or inefficiencies that contribute to the exhaustion of natural resources and degradation of natural environments. An increasingly important goal of responsible manufacturers is to reduce use of mineral and fossil resources, and reduce the environmental impact of not only plant operations, but also product use.

2.1.9 Mental Satisfaction Level Currently, products offering high performances and qualities at reasonable costs are the norm rather than the exception. Given this situation, product qualities related to mental factors such as aesthetic characteristics, generally called Kansei

22


characteristics, are becoming distinguishing factors that both encourage and respond to customer discrimination. Kansei and aesthetic characteristics are explained in detail in Sec. 4.2.

2.2 Criteria Requirements For the product manufacturing paradigm trends shown in Fig. 1.6, product development competition was based on the optimization of criteria that were popular during certain historical time periods. Such criteria and criteria requirements have evolved significantly over time, and have become more complex during the history of product manufacturing. This progression is outlined below. 1. Optimization of a single objective, such as minimization of the time it takes to complete a process, and minimization of the manufacturing cost within a certain value range. 2. Consideration of conflicting relationships, where improvement in one objective may cause degradation in some other objectives. The need to exploit wider and more flexible viewpoints during optimization was recognized. 3. Recognition of the need to address product manufacturing and use impacts upon natural environments, and consider the depletion of natural resources when developing product design criteria. 4. The degree of comfort and mental satisfaction that product users experience is included in product design criteria. Maximization of customer satisfaction is required, and aesthetic factors must be considered along with conventional product criteria concerning functions, performances, qualities, and costs. The foregoing factors 3 and 4 include criteria for which quantitative evaluations are not easy, but advanced product design solutions should be determined by systematically considering all of the many factors. The evaluative factors in product manufacturing are defined as characteristics of criteria, which are classified into characteristics for which greater values are more preferable and characteristics for which smaller values are more preferable: 1. Characteristics for which greater values are more preferable: profits, satisfaction levels, utilities, effectiveness levels, accuracies (depending on the application, static displacements, roughness, deviations, and vibration (dynamic) displacements become characteristics for which smaller values are more preferable), efficiencies (when “time” is used for evaluating efficiencies, it corresponds to characteristics for which smaller values are more preferable. When the quantity of accomplished tasks is used during efficiency evaluations, “the amount of accomplished tasks” is categorized into characteristics for which greater values are more preferable). 2. Characteristics for which smaller values are more preferable: cost, time (time needed for completion of a job), noise, structural weight, dissatisfaction level, downtime, operating energy requirements.

2.2 Criteria Requirements

23

Product designs generally aim to provide greater satisfaction levels. The criteria of satisfaction levels can express not only objective attributes such as cost and weight but also subjective attributes such as human preferences. Dissatisfaction levels included in characteristics for which smaller values are preferable can be suitable for specific problems (not system-wide) where specific dissatisfactions must be minimized, but may not be suitable when aiming to design more advanced and more preferable products. So-called utility analysis, often used in the decision-making field, is a method for quantitatively expressing decision-makers’ subjective preferences or satisfaction levels for alternatives and attributes [3]. Here, utilities are values that express satisfaction levels for alternatives. In the following, for convenience, the symbols u and s are respectively used to represent “utilities” and “satisfaction levels.” Attributes are generally expressed as functions of design variables. When utilities and satisfaction levels are design variable functions, they are termed utility functions and satisfaction functions (or preference functions), respectively. There are various methods for numerically expressing decision-maker preference features for each attribute, and one example is shown in (2.2) below [4,5]. This function is also illustrated in Fig. 2.7. The horizontal axis shows ε ij which expresses how far the actual magnitude z j of attribute j is from the goal magnitude zij∗ , while the vertical axis expresses the satisfaction level sij of decision-maker i for attribute j that has a normalized value ranging from 0 to 1. sij =

1

π

tan −1{− a(ε ij − b)} + 0.5

(2.2)

where

⎧⎪ zij − zij∗ ⎫⎪ ⎬ when attribute z j is a characteristic for which smaller ⎪⎩ zij ∗ ⎪⎭

ε ij = ⎨

values are preferable. ⎧ z∗ − z ⎫ ⎪ ij ij ⎪ ⎬ when attribute ⎪⎩ zij ∗ ⎪⎭

ε ij = ⎨

values are preferable.

z j is a characteristic for which larger

24


1 0.8 0.6 0.5 0.4 0.2 -1

-0.5

b 0

εij

0.5

1

Fig. 2.7 Satisfaction function curve example

ε ij in the foregoing equation non-dimensionally expresses the distance between the actual value and the goal value, and has a value of 0 when the actual value is coincident with the goal value. By modifying the magnitudes of coefficients a and b , various satisfaction function shapes can be constructed. Smaller values of coefficient a yield more gently sloped function lines, while larger values of coefficient a increase the slope. Coefficient b corresponds to the value of ε where the satisfaction level is 0.5. The satisfaction level at the goal value is denoted as s0 , and expresses the satisfaction level of the decision-maker in the case where the goal value is obtained. Depending on the specific problem, the satisfaction level can be close to but somewhat below a value of 1 when the goal value is obtained. Since there are possible cases where values greater than the goal value are more preferable, the satisfaction level at the goal value is here set to a limit value smaller than 1. By adjusting the values of coefficients a and b , decisionmakers can formulate suitable satisfaction functions for the attributes being regarded. For example, decision-makers in product design and manufacturing divisions are likely to have conflicting requirements for maximizing product performances and minimizing product manufacturing cost. These requirements can be expressed using satisfaction functions and when the manufacturing cost is held constant, decision-makers in the product design division can safely ignore designs that do not satisfy a minimum level of a particular product performance where the satisfaction level is low. When the design begins to satisfy the required performance, the satis-

2.2 Criteria Requirements

25

faction level may suddenly approach a value of 1. Some satisfaction function features are shown in Fig. 2.8 a.

1 0.8 0.6 0.4 0.2 -1

0 0

-0.5

εij

0.5

1

(a) Steep slope case 1 0.8 0.6 0.4 0.2 -1

-0.5

0 0

εij

0.5

1

(b) Gentle slope case Fig. 2.8 Satisfaction function curve patterns

When decision-makers consider product manufacturing costs, lower values are preferable, but if a satisfactory product performance cannot be obtained at a given

26


manufacturing cost, product designs with higher product performance values may be more preferable despite increases in manufacturing cost. In such cases, the satisfaction function has a gentle slope as shown in Fig. 2.8 b. As already stated, many attributes should be systematically evaluated in current product manufacturing, and the multiobjective optimization methods explained in the following section can be effectively used to do so. There are two approaches, one of which uses inherent attribute magnitudes while the other uses normalized measures for all attributes. To normalize the measures, each attribute can be represented as a cost, satisfaction level, utility, or dissatisfaction level, and in some cases, reciprocals of satisfaction levels are used. Sometimes, to express the utility function for an integrated collection of attributes, multi-attribute utility functions are used [6].

2.3 Relationships Between Criteria and Optimization After evaluative criteria have been defined, the most preferable solutions based on the criteria must be found, and optimization methods provide the means. A basic optimization problem is formulated by including evaluation characteristics for product designs in an objective function f , and constraint functions g j ( j = 1,2,..., m ) and hk ( k = 1,2,..., p ) , as follows: f

→

minimize or maximize

g j ≤ 0,

j = 1,2,..., m

hk = 0,

k = 1,2,..., p

f , g j ( j = 1,2,..., m) , and hk ( k = 1,2,..., p ) are functions of design variables d i (i = 1,2,..., n ) . The design variables are determined by solving the foregoing optimization problem. The design variables are often expressed using vector symbols as follows: d = [ d1 ， d 2 ,..., d n ]T

For the objective function f , an evaluation factor is selected from among those pertaining to the maximization of profit, achievement of an important business strategy, gaining a competitive advantage, or the like. Constraints are set by using evaluation factors that must be satisfied without fail. Objective function f can be expressed as either a maximization or minimization problem, as desired, by expressing f as − f or 1 / f .

2.3 Relationships Between Criteria and Optimization

27

Problems aiming to obtain values of characteristics, performances, costs, etc., after setting design variable values, are called forward problems, while those seeking to obtain design variable values that satisfy the requirements of set characteristics, performances, costs, and the like, are called inverse problems. Design optimization problems are of the inverse type. The terms and symbols principally used in optimization problems and formulations are given below. Objective function： f Constraint function: g j , hk As with the foregoing formulation, for efficient numerical processing the constraint functions are typically placed at the left side of the constraint conditions, with zero to the right. g j is a constraint function for an inequality constraint condition, and hk is a constraint function for an equality constraint condition. Side-constraints: upper or lower constraints for design variables. Dimensional constraints are representative. Design variables: d i depending of the kind of problems, the term “decision variables” is also used, meaning the variables to be determined. State variables: s characteristics expressing the sates of the optimization subject such as displacements, stresses, and natural frequencies. Design variable space: coordinate space expressing the feasible region of the design variables. Objective function space: coordinate space expressing the feasible region of the objective functions. Figure 2.9 shows a design variable space for design variables d1 and d 2 where each design variable corresponds to a coordinate axis. Figure 2.9 a is an example for linear optimization problems (linear planning problems) in which the objective and constraint functions are both linear functions of the design variables. The shaded area corresponds to the feasible design region where all constraint conditions are satisfied. The contour lines of the objective function expressed by the dotted lines have smaller values when approaching the origin of the coordinate axes, and the optimum solution is obtained at point A. Optimization problems in which any of the objective and constraint functions is a nonlinear function of the design variables are termed nonlinear optimization problems (nonlinear planning problems). Figure 2.9 b shows a simple example in which all of the objective and constraint functions are nonlinear functions of design variables. The shaded area corresponds to the feasible design region defined by the constraint conditions. The contour lines of the objective function expressed by the dotted lines have smaller values when approaching the origin of the coordinate axes, and the optimum solution is obtained at point A. More complicated nonlinear optimization problems are explained in Sec. 7.2.

28

2 Evaluative Criteria for Product Manufacturing and Optimization Fundamentals g3=0 g2=0

g4=0

Feasible region g1=0

A

g5=0

Contour lines of objective function f

0

d1 (a) Linear optimization problem g2=0

g3=0


A

g4=0


0

d1 (b) Nonlinear optimization problem

Fig. 2.9 Simple examples of design variable space

Optimization problems such as shown in Fig. 2.9 are the simplest here, and optimum solutions can be easily obtained. The most important point when using optimization techniques is how to formulate the problems being regarded and to do this well requires great attention and deep consideration of the fundamental aspects of the problem at hand. Optimization techniques and procedures have been applied to many fields and jobs in product manufacturing, and application levels have advanced as shown in Fig. 2.10. Before optimization techniques were applied, improvement procedures


29

were generally attempted for each product design and manufacturing process separately. The accumulation of incremental improvements over time improved the quality and efficiency of product manufacturing but the limitations of this approach in terms of providing truly optimal solutions were eventually recognized. Optimization techniques became essential to the development of more preferable solutions, and were initially applied to certain expert working fields before spreading more broadly. When it was recognized that interrelationships between fields had to be addressed, optimization techniques were applied more broadly across various fields, according to the problem being considered, so-called multidisciplinary optimization, which is still a partial optimization. Currently, the importance of global optimization is gaining wider recognition, since the optimum solution for problems as a whole cannot be reliably obtained by simply accumulating partial optimizations.

Accumulation of partial improvements

Optimization applied to a specific field

Optimization applied to multidisciplinary fields

Global optimization

Breakthrough of optimum solutions Fig. 2.10 Developmental progress in optimization techniques applied to industrial activities

In any case, product designs always require minimization of product manufacturing cost, and, in most practical scenarios, methods that reduce this cost inevitably result in degradation of the product performances. There are cases where a specific product performance must be as high as possible, and the product manufacturing cost is consequently forced upward. Furthermore, when the upper or lower bounds of the constraints are set, their values determine the result of the op-

30


timum solution, but in practical scenarios, setting specific upper or lower bounds may be problematic, especially when certain factors are unclear. In these cases, formulating optimization problems with a number of objective functions that include such characteristics is effective, and design optimization problems of this type are generally called multiobjective optimization problems [7]. When there are two objective functions and smaller values of each objective function are more preferable, the objective of the multiobjective function is expressed as follows: f = [ f1, f 2 ] → minimize

(2.3)

As an example, consider a scenario where product designers are seeking a design solution that has a higher product performance while process designers engaged in practical manufacturing desire solutions that have lower product manufacturing costs. These two requirements naturally have conflicting interrelationships. Figure 2.11 shows the relationships between a product performance that must be maximized and the product manufacturing cost that always needs to be minimized [8]. When the product performance and the product manufacturing cost are respectively expressed by f1 and f 2 , the foregoing multiobjective formulation is changed as follows: f =[−f1, f2] → minimize

(2.4)

Fig. 2.11 Conflicting relationships between a product performance characteristic and manufacturing cost


31

Figure 2.11 is an objective function space where multiobjective functions form a coordinate space. The shaded area in Fig. 2.11 corresponds to the region that is feasible using presently available knowledge and technology. The line PQ corresponds to a Pareto optimum solution set for the two objective functions of the product performance and the product manufacturing cost. The Pareto optimum solution set is defined as a set consisting of feasible solutions in each of which there exist no other feasible solutions that will yield an improvement in one objective without causing degradation in at least one other objective. “Pareto” comes from the surname of V. Pareto (1848 ̶ 1923), an Italian economist. He presented a concept that in a situation where products are efficiently manufactured from existing resources and the manufactured products are efficiently distributed, when distribution of the product or the resources to a specific person is changed so that his or her satisfaction level becomes higher, the satisfaction level of some other persons is inevitably reduced [9]. A Pareto optimum solution set is also called a non-inferior solution set, but the Pareto optimum solution set term is generally used in engineering optimization fields. A Pareto optimum solution set such as shown in Fig. 2.11 is a set of candidate solutions from which the optimum solution is selected. The line (or, when there are three objective functions, the curved surface) is useful because it clearly shows the features of the solutions from a broad point of view. Designers usually seek solutions in the direction of the large arrow located in the feasible region. Looking at the design solutions at points A, B, and C on the Pareto optimum solution line PQ, we see that the design solution at point A provides excellent product performance, but at a very high manufacturing cost. The design solution at point C has a low manufacturing cost, but inferior product performance, and the design solution at point B offers rather good product performance and also a reasonable manufacturing cost. The solution actually used will be selected by the manufacturer after considering customer preferences and priorities. Designers generally look for practical design solutions on a Pareto optimum solution line. A global solution on the PQ line is difficult to obtain by the accumulation of partial optimizations that, for example, would yield solutions on the P''Q'' line located within the feasible region, but these are rather far from the Pareto optimum solution line where the best solutions are located. For example, solution point G inside the feasible design region is inferior to any solution on solution line DE, and thus should not be selected as a design solution. The foregoing discussion illustrates that searching for design solutions that lie on the global Pareto optimum solution line is an important part of practical product design and manufacturing. Given the competitive nature of the marketplace, it is obvious that companies making more preferable products that offer better value will usually gain market share. Obtaining Pareto optimum solutions that are superior in the global sense is therefore often of crucial importance in the development of successful product designs. Figure 2.11 expresses the relationship between the product performance and the product manufacturing cost as an optimization problem for product designs. In

32


value engineering (value analysis; VA), the “value” for a product is expressed by the ratio of “function” that customers desire over the “cost” required to realize the function, as follows: Value =

Function Cost

(2.5)

where “function” generally includes performance, quality, and operating factors other than cost. When conducting VA, products having greater magnitudes of the value criterion are judged superior. The display of a Pareto optimum solution set such as shown in Fig. 2.11 is useful not only because it displays specific solutions, but also because a range of candidate optimum solutions on the PQ line can be visually and quantitatively understood. By looking at the entire Pareto optimum line, the relationships between the conflicting objective functions can be clearly recognized and compared. While accurately judging the worth of a single solution in isolation is impossible, the quality of specific available solutions can be judged and verified by the relative comparison of a set of candidate solutions. In optimizations for product manufacturing, the initial focus is on obtaining solutions such those lying on the PQ line shown in Fig. 2.11, which are termed the global optimum solutions. After such solutions are obtained, it is usually necessary to search for even better solutions, such as those lying on the P'Q' line, which represent important breakthroughs, beyond the PQ line [10]. Given marketplace competition, there is significant pressure driving the evolution of product design solutions and product manufacturing techniques, and a currently successful product may rapidly lose its appeal due to the introduction of more sophisticated products that offer better customer value. The satisfaction levels of increasingly knowledgeable and sophisticated customers can only be met by continual improvements in product design and manufacturing. Currently, it is impossible to obtain the best product design and manufacturing solutions by using human abilities alone. Strategic and systematic concepts, environments that include powerful computational and information tools, as well as new technologies are necessary. The following chapters will explain these concepts and technologies in detail.

Exercises 2.1 Using a practical product such as an automobile, discuss the performances, qualities, or costs that are evaluated when selecting such a product for purchase. Putting yourself in the position of a product designer, discuss the most important performances, qualities, and/or manufacturing cost that must be considered when designing the product.

References

33

2.2 Giving costs pertaining to product manufacturing as examples, explore the relationship between each cost and its related characteristics, such as the structural material cost and the structural weight, the structural material cost and the structural rigidity, and the machining cost of the joint contact surfaces and the product’s operational accuracy or joint rigidity. 2.3 Discuss product manufacturing strategies that minimize excess inventory viewed as a kind of waste for products that have a short lifecycle. 2.4 Discuss the importance of processing capability when evaluating the quality of various processes. 2.5 Why should the consideration of product lifecycle factors be carried out at the earlier stages of product design? 2.6 Give examples of products for which not only product performances and qualities but also psychological satisfactions, such as pleasing aesthetics and feelings of comfort, are important when selecting products to buy. Describe the relationships between product performances/qualities and psychological satisfactions. 2.7 List ten characteristics for which greater values are more preferable and ten characteristics for which smaller values are more preferable. 2.8 List ten combinations of conflicting characteristics, and, for each combination, discuss design variable differences that bring about the best characteristic value for each characteristic. 2.9 Why does the number of criteria in product design and manufacturing increase as product manufacturing processes become more advanced and sophisticated? 2.10 Why is it necessary to use multiobjective optimization procedures for product design optimizations? 2.11 Why is flexibility in manufacturing equipment and industrial machinery a requirement for effective operation of job shop type production?

References 1. Hitomi K, Yoshimura M (1986) Operations scheduling for work transportation by industrial robots in automated manufacturing systems. Material flow, 3:131 ̶ 39 2. Brisssaud D, Tichkiewitch S, Zwolinski P (eds) (2006) Innovation in life cycle engineering and sustainable development, Springer, p111 3. Keeney RL, Raiffa H (1993) Decisions with multiple objectives. Cambridge University Press, p131 4. Yoshimura M, Kondo H (1997) Group decision making in product design and manufacturing. In: Proceedings of 1997 ASME Design Engineering Technical Conferences, September, Sacramento, California:1 ̶ 7 5. Yoshimura M, Kondo H (1996) Product design based on concurrent processing of design and manufacturing information by utility analysis. International Journal of Concurrent Engineering: Research and Applications, 4(4):379-388

34


6. Keeney RL, Raiffa H (1993) Decisions with multiple objectives. Cambridge University Press, p219 7. Cohon, LL (1978) Multiobjective programming and planning. Academic Press, p220 8. Yoshimura M (1993) Concurrent optimization of product design and manufacture. In: Parsaei HR, Sullivan WG (eds) Concurrent engineering - - contemporary issues and modern design tools, Chapman & Hall:159 ̶ 83 9. Samuelson, PA, Nordhaus WD (1989) Economics. McGraw-Hill 10.Yoshimura M, Kimura A (1994) Evolutional optimization of product design based on concurrent processing of design and manufacturing information. In: Proceedings of the Fifth /USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and Optimization (AIAA94-4299-CP), Part1, Sept.:434-442

3 Technologies for Product Manufacturing Innovation

Effective innovation in product design and manufacturing depends on a thorough understanding of leading concepts. This chapter describes fundamental concepts and strategies for innovating product design and manufacturing, namely (1) generation from the conceptual design stages, (2) concurrent engineering, and (3) collaboration. Concurrent engineering is a fundamental concept that aids obtaining global optimum design solutions using existing knowledge and technologies. When skillfully employed, concurrent engineering can both minimize manufacturing processing time and also assist in the development of useful alternative global optimum solutions that improve product performance while controlling product manufacturing cost. Collaboration is also a fundamental concept, increasingly used to break through existing global optimum solutions and to enable adroit deployment of a range of knowledge and skills in broader contexts. Successful collaboration, however, requires that certain necessary conditions be satisfied, and these are explained.

3.1 Generation of Better Products from Wider Feasibilities When searching for more preferable solutions using conventional methodologies and procedures, or within frameworks of readily available technologies and knowledge, truly global optimal solutions, or breakthrough designs, are elusive. New strategies that begin at the conceptual or outline design stages are required, and these must be based on essential product manufacturing considerations. Such strategies need to be constructed with the specific goal of obtaining preferable solutions. As described in Chap. 2, it is absolutely crucial that the widest possible feasible design regions be employed, by gathering as much knowledge and information concerning the product being developed, and ultimately by selecting the best design solution from the broadest possible set of potential solutions [1]. What

36


must be avoided is an automatic, rigid, or mechanical formulation of the optimization problem, where the outcome will inevitably be restricted and of limited potential in terms of providing a globally optimal solution. Figure 3.1 shows three key concepts for innovating product manufacturing: (1) conceptualization, i.e., generation from concepts, (2) concurrent engineering, and (3) collaboration. Each of these concepts is explained below.

Conceptualization

Key concepts for innovating product manufacturing

Concurrent engineering

Collaboration

Fig. 3.1 Key concepts for innovating product manufacturing

3.2 Generation from Conceptual Design Stages Product designers usually base their designs for new products on background experience and knowledge of existing products in the market. Unfortunately, this type of knowledge can be constraining if it prevents the development of new approaches that address the widest possible range of concerns, concerns that are not limited to product design, but include manufacturing and other factors affecting the potential success of the product being considered. Similarly, in the case of manufacturing methods, conventional scenarios often have limited flexibility. For example, in an attempt to improve manufacturing capability, automation and computer assisted methods may alone be applied to the production of certain parts, yielding incremental improvements rather a revolutionized production of truly superior products that could be achieved by following a more comprehensive approach. To achieve design circumstances that can take advantage of greater freedom in the design and production of high-performance products, the entire design optimization should be started from a condition where the subjects are simplified or abstracted [2,3]. Figure 3.2 shows a design scenario that begins with simplified or idealized conditions, prior to the development of optimization procedures.

3.3 Concurrent Engineering

37

To obtain superior design solutions, we must exclude preconceptions as far as possible, and create conditions that can include wider possibilities. The development of strategies and ideas in such circumstances means that a broad range of possibilities will be considered impartially. The methodologies used to develop design solutions, which start from conceptual product design stages and include product manufacturing concerns, employ principles that are fundamental to concurrent engineering and collaboration, and are explained below.

Simplification

Optimization

Implementation

(a) Optimization based on simplification

Idealization

Optimization

Implementation

(b) Optimization based on idealization Fig. 3.2 Optimization based on simplification and idealization

3.3 Concurrent Engineering Figure 3.3 illustrates the manufacturing sequence for products conventionally put on the market, namely (1) research and development, (2) product design, (3) manufacturing, and (4) marketing. Within a company, each of these operations usually corresponds to a single division, and within each division, particular decisions are made according to information received from upstream divisions. Various approaches to make the execution of tasks more effective and efficient are explored and executed in each division. Presently, the most common way to improve lackluster product design and manufacturing is to introduce computer-aided technologies into divisional operations. Such technologies include Computer-Aided Design (CAD), Computer-Aided Manufacturing (CAM), Computer-Aided Process Planning (CAPP), Computer-Aided Planning (CAP), Computer-Aided Testing (CAT), and so on [3, 4]. Furthermore, to enhance the smooth flow of information and materials throughout the product design and manufacturing operation, and to automate and integrate various tasks as much as possible, computer-aided technologies termed Computer-Integrated Manufacturing (CIM) are often used [5, 6]. The features and benefits of this kind of integration differ depending on the particular focus of the enterprise. The framework of system organization is generally the same, regardless of the product, but the details of how automatic connections

38


between the system components are deployed, and how information and material flow can be enhanced, differ according to how CIM technologies are integrated. Figure 3.4 shows a conceptual diagram for the sequential flow of information and materials between three hierarchical divisions such as product design, manufacturing, and marketing. The aim of CIM technologies is to make these various flows automatic and smooth.

Research and development (division)

Product design (division)

Manufacturing (division)

Marketing (division) Fig. 3.3 One-way product manufacturing flow

Decision maker A

Decision maker B

Decision maker C Fig. 3.4 One-way decision-making flow

In decision-making flows such as shown in Fig. 3.4, the decisions taken in upper divisions to implement various requirements and details become constraints with respect to decision-making in downstream divisions. For example, attempting


39

to reduce manufacturing costs after the details of a product design have already been decided will likely prove ineffective since it is the product design itself that largely determines the manufacturing cost. In a rigidly sequential manufacturing flow, cost reductions can seldom be implemented after the product design phase, such as at the process design stage when manufacturing methods and details are determined. Conflicting requirements may exist among divisions but these cannot be resolved due to the rigidly sequential manufacturing flow. Furthermore, a strictly chronological approach to product design and production is especially illsuited to current merchandising trends, where rapid product turnover and shortening the time to market are major concerns. The usual product manufacturing process, as explained above, consists of a one-way sequence of steps such as (1) product development, (2) product design, (3) manufacturing, and (4) marketing, with each step generally corresponding to a specific division that is hierarchically connected to other divisions. Within each division, the procedures for obtaining more preferable decision-making results are conducted according to the constraints of authority held by each particular division. However, for decision-making processes in a system defined as shown in Fig. 3.3, factors determined at a higher process level become decision-making constraints for the lower level processes, and the feasible decision-making spaces are thereby undesirably narrowed. Hence, one-way sequential decision-making processes seldom yield globally optimal design solutions. In contrast with the one-way sequential paradigm, Fig. 3.5 shows a concurrent engineering concept in which the major design factors pertaining to each division in Fig. 3.3 are concurrently and/or cooperatively decided at the highest level of product manufacturing. Research and development division

Product design division

Design proposals

Manufacturing division Fig. 3.5 Conceptual diagram of concurrent engineering

Marketing division

40


When Japanese industries develop new products, members of various divisions such as research and development, product design, manufacturing, marketing, and so on, traditionally gather, form a team, and discuss a range of topics concerning the product’s design and manufacturing, but concurrent activities are not the focus, and computer-aided systems are seldom used at this stage. In western countries, inter-divisional communication was historically rather poor, but around 1988, computer-aided systems began to be used and the term concurrent engineering and its concepts began to be popularized in the U.S. The technical term is also called simultaneous engineering. In concurrent engineering, decision-making concerning product manufacturing is conducted cooperatively, as shown in Fig. 3.5, where product design and manufacturing proposals are concurrently discussed and cooperatively determined by experts in each division, starting at the product’s conceptual design stages. The concepts and methodologies of concurrent engineering have been actively discussed in books and papers that clarify the practical benefits of concurrent engineering [7 ̶ 10]. To gain maximum benefit from concurrent engineering practices, the use of various optimization technologies is indispensable. Figure 3.6 shows the fundamental flow used when applying the concept of concurrent engineering to product designs. First, a wide range of evaluative factors and decision/design variables are impartially gathered, items which are conventionally decided according to experience. Next, the relationships between the evaluative factors are systematically analyzed and then suitable optimization procedures to obtain the global optimum solution are constructed. Optimization based on the concept of concurrent engineering is here called concurrent optimization.

Impartially gather evaluative factors and decision/ design variables, which are conventionally decided sequentially according to experience

Analyze relationships among evaluative factors

Construct optimization procedures to obtain global optimum solution Fig. 3.6 Fundamental preparation flow for executing concurrent optimization


41

Often, terms such as design for manufacturing [11], design for assembly, design for maintenance, design for distribution, design for quality, design for environment, and design for reliability are used, where a factor corresponding to an aspect of product manufacturing and product lifecycle replaces the X in Design for X. The X simply expresses factors that should be subject to concurrent and cooperative decision-making at the earliest product design stage. Design for Lifecycle and Lifecycle Design are both terms used when referring to product designs that must consider the entire scope of a product’s lifecycle. Concurrent engineering has philosophical similarities with CIM from the standpoint of integration but the former emphasizes simultaneous and concurrent decision-making in the early production stage.

Ts Process 1 Process 2 Process 3

Process N Time

(a) Sequential operation flow

Tc Process 1 Process 2 Process 3

Process N Time

(b) Concurrent operation flow Fig. 3.7 Comparison of sequential and concurrent operation sequences

42


Working processes that are based on concurrent engineering concepts can markedly reduce the time required to complete various jobs. Figure 3.7 a shows sequential operations where subsequent operations are performed after completing prior operations according to a sequence having N processes. Time Ts is required to complete N processes. Figure 3.7 b shows a parallel operation flow where the timing of certain operations is adapted so that some processes are cooperatively carried out simultaneously or in overlapping time segments. In this case, the time Tc needed to complete N processes can be dramatically shortened compared with Ts . Similarly, product design and development typically have many individual processes, but applying concurrent engineering concepts can greatly reduce product design and development times. The important point when applying concurrent engineering concepts to product manufacturing is not only to reduce the time needed to carry out various operations, but also to enable breakthroughs in the areas of product performance and manufacturing cost. Product designs that offer the best optimum performances are typically obtained under certain constraints, and then detailed manufacturing methods and procedures are determined so that the specified product design can be manufactured at minimal cost. However, since there are conflicting relationships between the requirements to reduce product manufacturing cost and increase product performances, it is impossible simply to obtain a unique optimal design solution. A set of candidate solutions is obtained, from which the best solution for the requirements of the particular product can be selected. X1

X1 X4

X5

X4

(a) N=20

X4

(b) N=16

X1

X1

X4

X3 (c) N=10

(d) N=24 Welded parts

Fig. 3.8 Examples of structural member cross-sectional shapes ( N : total number of welded parts, X : design variables)


43

Here, the importance of concurrently evaluating manufacturing cost and performance characteristics is explained using an example of a welded structure [12]. When a structure is constructed using welding operations, many structural segments must be welded together. To improve the performance of structural members (for example, reduce stress and increase rigidity), numerous partitions, ribs, or braces are often added, typically along the inside faces of the structural members. Figure 3.8 shows cross-sectional shape patterns where partitions are installed in the vertical direction, but the complex task of welding in the partitions increases the welding cost. When N is the total number of welded parts, the weldN

ing cost is expressed as ∑ Qi Li , where Qi and Li are the welding cost per unit i =1

length and the welding length at welded part i , respectively. The detailed design should be determined by simultaneous evaluations of structural member performances and manufacturing cost (here, welding cost). For example, as shown in the following equation, the detailed shapes of the structural member can be determined by minimizing the welding cost f so that the maximum value of the local deformations, Δ max , does not exceed the upper limit ΔU as follows: N

f = ∑ Qi Li → minimize i =1

(3.1)

Constraint condition: g = Δ max − ΔU ≤ 0

Figure 3.9 shows a conceptual diagram of the relationship between performance characteristics such as the maximum stress and the deformation (expressed as characteristics for which smaller magnitudes are more preferable), and the welding cost. After optimization, when the relation between the welding cost and the particular performance characteristic is obtained at Point P1 , the design solution can be simplified, reducing the welding cost while minimizing any deterioration in the performance characteristic, by plotting the line from point P1 to point P2 and beyond. In the region from point P2 through P3 , the welding cost decreases but the performance characteristic begins to deteriorate, and in the region from point P3 through P4 the performance characteristic deteriorates greatly while the welding cost decreases only a little. Line P1P2 P3P4 corresponds to the Pareto optimum solution set line of a multiobjective optimization problem to maximize the performance characteristic and minimize the welding cost. The features of the line shown in Fig. 3.9, obtained by solving (3.1), indicate that a design solution in the P2 P3 region of the line is most suitable.

44


P1

P2

P4

P3

Performance characteristic Fig. 3.9 Relationship between performance characteristics and welding cost for welded structure

When concurrent engineering principles are applied, the decision-making required for product design and manufacturing factors is cooperatively carried out, simultaneously and concurrently. Concurrent engineering therefore means that all divisions work together cooperatively and at the same time, to make decisions concerning a range of factors before determining product details, a task that is facilitated by the use of computer networks. Skillfully employing “concurrent engineering” concepts is advantageous not only in terms of time, since concurrently and cooperatively processing related tasks is more efficient, but it also increases the probability of achieving more preferable results concerning the product’s manufacturing costs, performances, and qualities. Concurrent engineering leads to wider feasible design regions that include more preferable solutions, and makes it easier to obtain the best possible solutions from among these more versatile feasible regions. Competitive requirements, conflicting factors pertaining to different divisions, and tradeoff relationships among product characteristics can all be appropriately resolved, and an enterprise atmosphere of mutual understanding and improved cooperation can be achieved. To accomplish the range of decisionmaking required, many factors must be considered and large scale decisionmaking problems must be solved. Products manufactured by various makers are bought by consumers who then use them, and maintain them when necessary, until they cease to be useful. At that time, certain product parts and materials can, in certain cases, be reused or recycled, while the remainder is disposed of. This flow of products, from creation, through use, repair, reuse, recycling, and disposal, forms what is called a product’s


45

lifecycle. To achieve optimal product designs, all factors and items pertaining to a product’s lifecycle should be fully considered at the earliest possible product design stage. That is, as the concept is shown in Fig. 3.10, the full range of factors concerning a product’s lifecycle, such as the manufacturing and purchase of machine components, the assembly, use, maintenance, disassembly, disposal, material recycling, and reuse of parts and materials, should all be concurrently considered and optimized from the initial design proposal stage [13].

Product design

Reuse of elements and pieces Material recycling

Manufacturing and purchase of machine components

Disassembly

Design proposals

Disposal

Maintenance

Assembly Use

Fig. 3.10 Conceptual diagram of lifecycle design

During a product’s lifecycle, various inconveniences may arise. Some of these undesirable outcomes affect the consumer’s ability to use the product or derive the expected degree of satisfaction from it, while others may affect the environment in which the product is used, or the environment at large. If steps to mitigate such unwelcome circumstances are considered only when they occur, the potential for implementing the best possible solution or improvement will obviously be less than if such scenarios were considered at the product’s early design stages. Concerning manufacturing plant maintenance, product designers need to consider practical production planning and requirements that guarantee ease of maintenance, based on communication with production management and maintenance experts. In practical plant maintenance scenarios, maintenance experts, production management experts, and plant designers all have access to a common networked database, where past procedures and know-how can be reviewed and new knowl-

46


edge stored. Other maintenance experts can then use this know-how and knowledge in their maintenance operations. Over time, the accumulated knowledge of plant designs and operations can become a useful resource that increases the speed and effectiveness of decision-making, not just for the topic of focus, but also for broader enterprise operations, such as future plant designs, due to the range of knowledge obtained by inter-divisional networked communications. Conventional sequential manufacturing processes operate by accumulating partial optimizations, whereas manufacturing processes based on concurrent engineering aim to employ globally optimum results, as shown by the PQ line in Fig. 2.11. Similarly, methodologies and technologies based on concurrent engineering alter conventional methods used for decision-making in product design and manufacturing so that more preferable product design solutions can be globally obtained. The decision-making methods for concurrent design and engineering are re-defined, and conventionally developed manufacturing systems that mainly focus on the automation of design and manufacturing processes and the streamlining of information flow are reconstructed. Concurrent engineering can offer an additional benefit and effectiveness in the form of synergy effects generated by concurrently considering items that originally were separately considered. This matter has some relationship with collaboration concepts that are described in the following section. Concurrent engineering, one of the most important optimization strategies, is an engineering concept and set of procedures often used in the fields of product design and manufacturing, but the concept and fundamental procedures can also be used to obtain better solutions for general optimization and decision-making problems.

3.4 Collaboration The concept of collaboration as popularly used in various fields is that different experts, each of whom possesses specific knowledge, skills, technologies, and/or know-how, cooperatively share a variety of information as they develop a new project. A group of experts here may correspond to experts in the same field who are working at different companies, or they may be experts in completely different fields. The potential advantages offered by this kind of multidisciplinary collaboration are wider reaching than with conventional collaborative endeavors that are limited to a single group, corporate division, or local society. Figure 3.11 shows the concept of collaboration where three different companies cooperatively develop a new product. Individual persons, a single division, factory, or company, have limited technological and material resources. In product manufacturing, new technologies and knowledge are always required to achieve products with higher functions and higher performances at lower manufacturing costs, but it is seldom possible for a single enterprise or company to have at its disposal the entire range of knowledge

3.4 Collaboration

47

and technology required. To cope with such situations, collaboration among different enterprises or companies is a potentially useful product manufacturing paradigm. Such collaboration, where companies as a group maximize their available management and technological resources while using external resources in areas where they are weaker, is becoming popular, put into practice by so-called virtual enterprises or virtual corporations.

Group A

Generation of new design solution

Group B

Group C

Fig. 3.11 Concept of collaboration among companies having different expert technologies

Before networking technologies became ubiquitous, distance was an obstacle to effective collaboration. Now, the use of effective information networking technologies is routine. The ability to gather widely available knowledge and experience, and concurrently evaluate a broader range of factors concerning products, can lead to more preferable design solutions. Here, new production methodologies are constructed, in which diverse experts cooperatively approach unified goals, increasing the potential for valuable breakthrough design solutions. However, achieving the benefits of collaboration does not absolutely require the use of networked information systems. Such benefits can be readily expected if group members can simply conduct cooperative meetings face to face. It is expected that the results of collaboration will benefit all group members, so that satisfaction values are far above what they might have been if collaboration was not employed. However, with absent proper conditions for successful collaboration, not only will such benefits be lost, but actual harm may be done, due to unwanted dissemination of proprietary technologies, knowledge, expertise, and the like. The bold line PQ in Fig. 2.11 expresses the Pareto optimum solution set of feasible results that use existing technologies and knowledge based on the concurrent engineering concept described in the former section. But actually employing col-

48


laboration concepts corresponds to searching for solutions on the P'Q' line that break through the PQ line.

β

α

Knowledge of designer α

Knowledge of designer β

α

β

Common knowledge

Combined knowledge sets Fig. 3.12 Concept of knowledge sharing

Two designers, α and β , who are representatives of respective companies, collaborate and cooperatively develop a product. Figure 3.12 shows the concept of knowledge sharing among designers α and β . Knowledge that one designer lacks may be supplied by the other. How to use effectively such external knowledge and information is often of key importance when developing new products [14]. Here, the cooperative development of designs for an industrial robot by designers α and β is considered as an example. The areas of knowledge required for designing the product are materials (item I1 ), cross-sectional shapes of arms (item I 2 ), and motors at joints (item I 3 ). The characteristics to be evaluated are the total weight W of the structure, the maximum dynamic displacement δ at the endeffector point, and the operation time T .

3.4 Collaboration

49

For the cooperative project, the designers α and β have the following kinds of knowledge to begin with. Designer α is very experienced in designing industrial robots that have high operational efficiency, is experienced in the use of high speed motors, knows how to develop designs for lightweight moving arms that exploit their advantages, but lacks detailed knowledge concerning materials. Designer β is somewhat experienced in developing industrial robots that have high operational accuracy, has knowledge and use experience concerning materials that positively influence designs for operational accuracy, but lacks detailed knowledge of motors used in arms capable of high speed movement. In the above engineering knowledge situation, a project to develop a product whose performance achieves both high operational efficiency and high operational accuracy is considered in which each of the two designers offers knowledge to the other and knowledge is mutually shared.

Pareto optimum solution set line of designer α before collaboration

Pareto optimum solution set line after collaboration

Pareto optimum solution set line of designer β before collaboration

Ideal point before collaboration

0

Operation time T

Fig. 3.13 Change in Pareto optimum frontier due to collaboration

The results obtained by each of the designers separately and the results obtained via collaboration are compared as shown in Fig. 3.13. The design solutions obtained by designer α have smaller values of T but larger values of δ compared with the results obtained by designer β . Thus, when product designs are produced by designers working alone, design solutions with preferable values for both T and δ cannot be obtained. In contrast, collaboration based on knowledge

50


sharing yielded a Pareto optimum solution line that included design solutions nearer the ideal solution, defined as achieving designer α ’s best value for T and designer β ’s best value for δ . The most preferable design solution can be selected from the collaborative solution set. Here, knowledge sharing among only two designers was discussed. The term “designers” can be replaced by “groups” or “divisions,” and, perhaps, “enterprises.” Knowledge sharing can even be expanded to assist problem solving in virtual enterprises, where different enterprises are virtually combined for a specific project, using networked systems. Since collaboration offers considerable potential for breaking through present design levels and finding more preferable solutions to design problems, it should be considered an important concept in manufacturing methodologies. However, successful collaboration depends on supporting systems and practical implementation of tools. For example, during product design activities, information in various forms must be communicated among the collaborating group members. To manage the flow, storage, and dissemination of digital information, computer networks and database technology is a practical requirement. One framework for cooperative support is called CSCW (Computer Support Cooperative Work), and practical systems are termed groupware. One of the most important points for successful collaboration concerns the suitable selection of collaboration partners. In highly networked operations, the number of candidate partners is likely to be large. In such cases, a supporting system that assists the selection of suitable partners is preferable, given a reliable method for quantitatively estimating the effectiveness of the collaboration with different sets of partners. Practical procedures for determining the viability of cooperative work among designers α and β are as follows: Step 1. Items to be evaluated for the product design are selected. The items are denoted I i (i = 1,2,..., N ) , where N is the total number of items Step 2. Whether or not each of the two designers α and β can acquire new knowledge when the knowledge of item i is shared between them is set up as follows: ωαi = 1 when designer α acquires new knowledge ωαi = 0 when designer α does not acquire new knowledge ωβi = 1 when designer β acquires new knowledge ωβi = 0 when designer β does not acquire new knowledge

Step 3. When new knowledge is obtained, acquisition of the knowledge is meaningless for the project if the knowledge is not important for the given designer. Designer α and β therefore respectively define importance levels sαi and sβi for item i (for example, the pair comparison method explained in Sec. 7.2.2 can be used).

3.4 Collaboration

51

Step 4. The benefit levels of knowledge sharing are defined. Designer α ’s benefit level Sα obtained by sharing knowledge concerning all items is calculated as Sα = ∑ ωiα siα

(3.2)

i

Similarly, designer β ’s benefit level S β obtained by sharing knowledge concerning all items is calculated as

S β = ∑ ωiβ siβ

(3.3)

i

Step 5. Truly cooperative work or project development is only feasible when each partner can mutually benefit from sharing his or her knowledge. If the effectiveness of the collaboration particularly favors one designer, or a subset of a larger group of designers, the collaboration is infeasible. The viability of the cooperative project is determined. First, if either designer α or β can acquire new knowledge they consider important, the cooperative project is considered viable. Next, when both designers’ benefit levels under knowledge sharing is high, the cooperative effort is judged viable according to the following evaluation. The product of designer α ’s benefit level Sα and designer β ’s benefit level S β is denoted Ψ : Ψ = Sα S β

(3.4)

With Ψ L defined as the lower bound of Ψ , the cooperative project is judged unviable when Ψ < Ψ L . A default value of 0.25 is given for Ψ L , equivalent to each designer having a benefit level of 0.5. The value of Ψ L can be adjusted, depending on the needs of both designers. When Ψ ≥ Ψ L , a further evaluation is conducted. If neither designer can acquire useful knowledge, or if their benefit levels differ considerably, the cooperative project is unviable. The ratio of the two designers’ benefit levels is evaluated as follows: Φ=

min{Sα , S β } max{Sα , S β }

(3.5)

where Φ is 0 ≤ Φ ≤ 1 . Cooperative work is judged viable when Φ ≥ Φ L , and unviable when Φ < Φ L . The default value of Φ L here is 0.5, i.e., one designer having a benefit level twice

52


as large as the benefit level of the other designer. Φ L can also be altered at the request of either designer. Next, a problem situation is considered in which a primary company P0 selects a collaboration partner from among candidate partner companies P1, P2 , ..., PN when starting a project to develop a new product, where N is the total number of candidate partners [15]. The procedures have the following two steps: Step 1. Unsuitable candidate partners are excluded based on the levels of their existing technologies. Step 2. The most suitable partner is selected from among the candidate partners remaining after step 1 from the standpoint of technologies that must be newly developed, with lower development requirements being preferable. When selecting a collaboration partner, most weight should be given to potential partners that have the most expertise in technologies different from those held by the primary company initiating the cooperative project. In Step 1, first, the absolute value θ jP0 Pi of the difference between the evaluation level for technology j offered by candidate partner Pi and the evaluation level for technology j

owned by primary company P0 is evaluated: θ jP0 Pi =| T jP0 − T jPi |

(3.6)

where T jP0 is the evaluation level of primary company P0 ’s technology j and T jPi is the evaluation level of candidate partner Pi ’s technology j .

The sum Si of the absolute values of the evaluation level differences in technologies where primary company P0 ’s level of technology is lower than candidate partner Pi ’s technology level is obtained under the condition of θ jP0 Pi = 0 when T jP0 < T jPi as follows: Si = ∑ ω jθ jP0 Pi

(3.7)

The sum I i of the absolute values of the evaluation level differences in technologies where candidate partner Pi ’s technology level is higher than primary company P0 ’s technology level is obtained under the condition of θ jP0 Pi = 0 when T jP0 > T jPi as follows: I i = ∑ ω jθ jP0 Pi

(3.8)

3.4 Collaboration

53

Then, the value Ei of the existing technologies that the primary and partner company pair can expect to use without further development during collaboration is expressed as Ei = S i × I i

(3.9)

The value of Ei also represents the magnitude of the differences in the specialties of the primary and partner company technologies. Greater differences yield larger values of Ei . When Si and I i are, respectively, expressed on the abscissa and ordinate of rectangular coordinates as shown in Fig. 3.14, Ei corresponds to the area of the rectangular region. The difference levels in the specialties within the primary and partner company technologies can thus be visually understood, and larger areas represent correspondingly greater differences in the levels of technology held by the two companies. Although the larger the graphed area, the greater the benefits of collaboration, the shape of the area is critical. Square shapes indicate a mutually beneficial relationship, where elongated rectangular shapes reveal a fundamental asymmetry in the potential collaboration, where one company benefits more than the other. Such unbalanced collaborations are likely to have unsatisfactory outcomes.

Area Ei Ii

0

Si

Fig. 3.14 Relationship Si and I i (absolute values of technological level differences for superior and inferior evaluation vis-à-vis the primary company)

Unpromising candidate partners are excluded using the following procedures: 1. The differences between the horizontal and vertical lengths of the rectangular shapes Si and I i shown in Fig. 3.14 are evaluated. Candidate partners showing large differences between Si and I i are excluded from the list of candidate partners, due to the basic asymmetry in the expected benefits.

54


2. Then the magnitudes of Ei are compared among the candidate partners. Candidate partners having small magnitudes of Ei are excluded from the list of candidate partners, since minimal benefit would accrue from collaboration. In Step 2, candidate partners are selected from the remaining group of candidate partners considering technologies to be newly developed. First, the differences between the goal level of technologies to be developed and the present level of technology are obtained. The required amount of technology γ k that must be newly developed, Lγ P , is defined for the primary company P0 and candidate k i

partner Pi as follows:

Lγ Pi = Ekg − Eka k

(3.10)

where Ekg is the goal level of technology γ k to be developed, and Eka is the present level of technology γ k . The total amount of development required for the development of a new product without collaboration, ϕ Pi , is obtained as follows: m

ϕ Pi = ∑ Lγ k Pi k =1

(3.11)

where ϕ Pi also expresses the level of developmental difficulty without collaboration for the product under consideration. The quantitative sum of technologies to be collaboratively developed is denoted ϕ P0 Pi and is obtained as follows: m

ϕ P0 Pi = ∑ min( Lγ k P0 , Lγ k Pi ) k =1

(3.12)

For collaboration, candidate partners having smaller values of ϕ P0 Pi are preferable. The partner whose new technology requires less additional development is selected. The ratio representing the reduction in the amount of technologies that must be newly developed when collaboration is conducted, D0i , is defined with respect to the primary company as follows:

D0i =

ϕ P0 − ϕ P0 Pi ϕ P0

× 100

(3.13)

References

55

The reduction ratio in the amount of technologies that must be newly developed when collaboration is conducted, Di , is defined with respect to a candidate partner company Pi as follows: Di =

ϕ Pi − ϕ P0Pi ϕ Pi

× 100

(3.14)

When each of the reduction ratios D0i and Di for the amount of technologies to be newly developed, from the point of view of the primary company and the best candidate partner chosen above, are considered to be high enough, and the difference in the required amounts of technologies to be newly developed is considered to be small for both the primary company and the candidate partner, Pi is selected as the collaboration partner. Cutthroat competition among companies too often leads to thoughtless destruction of nature and poor utilization of natural resources, and technological escalation between companies, each seeking to outdo the other when competing for customers, does not guarantee socially useful benefits. Hence, collaboration among companies in a global marketplace is a promising production paradigm, now, and for the future. However, collaborative efforts among companies do include significant risk of failure, and the number of corporations experiencing difficulties after mergers or cooperative projects has been increasing. The selection of the most suitable product development partner, using methods such as the one discussed above, is thus an increasingly important technological subject.

Exercises 3.1 Why does the implementation of concurrent engineering concepts during product design lead to more preferable design solutions, not only in terms of product development time, but also in terms of product performance, product quality, and manufacturing cost? 3.2 Discuss the potential for breakthrough design solutions that can occur when experts who have different knowledge or technical expertise collaborate, and explain the particular conditions that facilitate obtaining successful results.

References 1. Yoshimura M, Izui K (1998) Machine system design optimization strategies based on expansion and contraction of design spaces. In: Proceedings of the 7th

56


AIAA/USAF/NASA/ISSMO Symposium on Multidisciplinary Analysis and Optimization. AIAA-98-4749, September, St. Louis, USA, pp320 ̶ 330 2. Yoshimura M, Nose K (1994) Generation of conceptual design for structural shapes and functional elements of machine systems having no preconceptions concerning the design. In: Proceedings of the 1994 ASME Design Automation Conference, Vol.2, Sept.:83 ̶ 89 3. Benhabib B (2003) Manufacturing—design, production, automation, and integration. Marcel Dekker 4. Hitomi K (Supervisor), Nakajima K, Yoshimura M, Yoshida T (eds) (1984) CAD, CAM, and CAP by computers (in Japanese). Kyoritsu-Shuppan 5. Harrington J (1973) Computer integrated manufacturing. Industrial Press 6. Hitomi K (Supervisor), Nakajima K, Yoshimura M, Yoshida T (eds) (1993) Fundamentals of CIM – Design, manufacturing, and management by computers (in Japanese). KyoritsuShuppan 7. Yoshimura M (1993) Concurrent optimization of product design and manufacture, In: Parsaei HR, Sullivan WG (eds) Concurrent engineering ̶ contemporary issues and modern design tools, Chapman & Hall, London:159 ̶ 183 8. Yoshimura M (1994) Integrated optimization of produt design and manufacturing. In: Leondes CT (ed) Control and dynamic systems ̶ concurrent engineering techniques and applications, Volume 62, Academic Press, San Diego:167 ̶ 219 9. Yoshimura M, Takeuchi Y, Hitomi K (1984) Design optimization of machine-tool structures considering manufacturing cost, accuracy and productivity. Transactions of the ASME, Journal of mechanisms, transmissions, and automation in design, 106(4):531 ̶ 537 10. Yoshimura M, Itani K, Hitomi K (1989) Integrated optimization of machine product design and process design. International journal of production research, 27(8):1241 ̶ 1256 11. Yoshimura M (1996) Design optimization for product life cycle. In: Huang GQ (ed) Design for X: concurrent engineering imperatives, Chapman & Hall, London, pp424 ̶ 440 12. Yoshimura M, Takeuchi A (1994) Concurrent optimization of product design and manufacturing based on information of users' needs. International journal of concurrent engineering: Research and applications, 2(2):33 ̶ 44 13. Doi K, Chujo Y, Yoshimura M, Nishiwaki S, Izui K (2009) Construction of an optimum system design method considering product lifecycle. International journal of sustainable engineering, 2(3):171 ̶ 183 14. Yoshimura M, Yoshikawa K (1998) Synergy effects of sharing knowledge during cooperative product design. Concurrent engineering: Research and applications, 6(1):7 ̶ 14 15. Yoshimura M, Izui, K, Kida S (2005), Decision support system for selecting collaborative product development partners. International journal of concurrent engineering: Research and applications, 13(1):5 ̶ 11

4 Involvement of People in Product Manufacturing

The roles that people play are especially important when product manufacturing innovation is a goal. People naturally tend to make aesthetic and emotional judgments about things and, given the proper environment, enjoy creating new things and working collaboratively with colleagues. Aesthetic and ergonomic factors have increasing importance when customer evaluations are allowed directly to affect product design improvements, since the creation of more preferable products greatly depends on human judgment and sensibility. To utilize best the power of collaborative groups in design and manufacturing, the selection of group members should also be optimized. This chapter describes the relationships between customers and manufacturers, and between product manufacturing and the particular product characteristics that are most important to certain groups of people. Kansei engineering and ergonomics are explained and methods for incorporating these important concepts are elucidated through product design examples. Optimal collaboration environments that enable consideration, integration, and balancing of collaborating group members’ abilities and personal preferences are then described.

4.1 Roles of Individuals in Product Manufacturing 4.1.1 Human Abilities Human abilities can be roughly categorized into physical and mental abilities, and practically all circumstances in which people live and work have both physical and mental aspects. These mental aspects can be split into logical and aesthetic categories, the combination of which is distinctly human. In contrast, most physical ability that people have can be imitated by machines or automated systems, and certain aspects of human logic can be replicated by computers, so that theo-

58


retical or analytical evaluations can yield quantitative results. However, quantitatively evaluating subjective and aesthetic factors is generally a much more difficult proposition. Customer satisfaction levels for popular products greatly depend on mental factors, especially subjective and aesthetic factors, but codifying and analyzing these important factors is elusive. Much effort has been spent in the past to reduce human labor during product design and, particularly, manufacturing, and many operations are now routinely automated and/or computerized. However, enthusiasm for mechanical solutions can sometimes obscure the fact that it is people who are actually responsible for generating more preferable products. In the end, the creation of more preferable products greatly depends on human judgment and sensibility. The inherent abilities of people to make judgments, to create new things, and work collaboratively with colleagues, are summarized and shown in Fig. 4.1. Working as designers, people display a range of abilities in making aesthetic/emotional judgments, evaluating the beauty, desirability, fit and finish, functionality, and degree of comfort that various products provide. People also reveal a range of abilities for creating products or concepts that embody the above practical and aesthetic characteristics. And people have various personal traits that either enhance or inhibit the realization of higher than expected levels of achievement when working collaboratively, in sympathy with their colleagues and coworkers. The abilities such as those shown in Fig. 4.1 can seldom be acquired by a dispassionate accumulation of knowledge or training, and are highly individual.

Capability to make “aesthetic/emotional judgments” about things

People’s inherent capabilities Capability to work Capability to “collaboratively in “generate” new sympathy” products or concepts with coworkers

Fig. 4.1 Inherent human abilities from the Kansei point of view

4.1 Roles of Individuals in Product Manufacturing

59

4.1.2 Relationships Between Customers and Manufacturers A major goal of product manufacturing is to provide customers with as much satisfaction as possible during the ownership and use of their chosen products, and discriminating customers are keenly aware of the degree to which a given product does, or does not, provide appropriate satisfaction levels. The quantitative attributes of product performances, qualities, functions, and manufacturing costs are comparatively easily evaluated by product manufacturers, but the active participation of customer evaluations concerning product designs is indispensable when seeking to implement aesthetic and ergonomic factors in product designs. In addition to factors influencing customer purchasing decisions, customers are usually intimately involved in the use and maintenance of their chosen products, and also the disposal of products when they are no longer of use. An example scenario of advanced product design would have the product manufacturer display the results of practical operational and maintenance simulations to potential customers, so that personal observation and evaluation could beneficially influence customer purchasing decisions concerning the offered products. Recent trends indicate that customers are becoming increasingly sophisticated and sensitive concerning the impacts that the manufacture and use of products have upon natural environments, as well as the need to plan for recycling of product components when they are discarded. Product designs that ignore these trends and pay little attention to customer desires are increasingly likely to fail in the marketplace. For example, products that require large quantities of natural resources, or products that soon fail in service, are unlikely to obtain high levels of customer satisfaction, even if the cost of such products is low. On the other hand, energyefficient products that are environmentally friendly are increasingly attractive and often obtain high levels of customer satisfaction. In short, the harmonies that products achieve with respect to the environments in which they are used, and their environmental impact, are becoming especially important factors in raising customer satisfaction levels. Thus, manufacturing paradigms where customers can not only select the best products from among those offered, but also participate effectively in the design of products that are responsive to their specific needs and requirements, as shown in Fig. 1.6, are desirable. Many product manufacturers currently integrate lifecycle concerns in their product designs as a matter of course, but there is still a need, and potential benefit, for customers to participate more actively in the design of better products. That is, when manufacturers and customers can truly collaborate in the design of products, goods that enable efficiencies during both manufacturing and use become possible, and maximal customer satisfaction and affluence can be attained with minimal negative impact.

60


4.2 Kansei Engineering As people’s lifestyle becomes increasingly affluent, the demand for greater varieties of products that offer superior functions and qualities also increases. Aesthetic factors that influence the mental and emotional satisfaction of the people using new products are becoming increasingly important in purchasing decisions. Kansei engineering is a technology for incorporating mental and emotional subjective factors into conventional product designs, by evaluating human aesthetic and emotional requirements for products along with objective factors such as product performances and qualities at the product design stages [1]. People have five physical senses, namely sight, hearing, taste, smell, and touch, as shown in Fig. 4.2, and each of them relates to a person’s Kansei. The English phrase closest in meaning to Kansei would be subjective attributes or aesthetic factors, but it is practically impossible to translate the delicate nuance of the Japanese word. Nevertheless, the phrase Kansei engineering is well known and now has global currency [2].

Sight Touch

The five human physical senses

Smell

Hearing

Taste

Fig. 4.2 The five human physical senses

Kansei engineering is, in a broad sense, included in the field of ergonomics. In ergonomics, factors that mitigate human fatigue while ensuring a durable degree of comfort when products are used, and the physical “fit” of products with people are mainly evaluated. On the other hand, in Kansei engineering, emotional and mental satisfactions, as well as comfort, are considered. The evaluation of Kansei is necessary not only for products used by customers at large, but also for industrial machines such as machine tools and industrial robots that are used in product manufacturing plants, so that worker comfort will be maximized and factors that decrease worker motivation will be minimized. Furthermore, the evaluation of Kansei is also important in developing and evolving the product manufacturing support systems explained in Chap. 5, so that the Kansei abilities of the people

4.2 Kansei Engineering

61

undertaking product development, design, and manufacturing will be most effectively utilized. The evaluation of Kansei features is most widely applied for coloration and outward appearance of products, but user preferences and Kansei for how products move during use are also important for machine products. An example of this type of Kansei evaluation is shown by the flow depicted in Fig. 4.3, where the movements of manipulators used to deliver a cup of juice into the hand of a human subject are to be evaluated [3].

Preparation of n design alternatives

Principal component analysis evaluation of analysis results

Kansei evaluation using the SD method

Draw evaluation map

Grasp Kansei features for design alternatives Fig. 4.3 Flow of Kansei evaluation for products using the SD method

The manipulator holds a cup and juice is poured in at point A, as shown in Fig. 4.4. The manipulator then delivers the cup to a customer at point B, and the Kansei factor for the movement is evaluated. A variety of mechanisms and movement features of the manipulators, generated by the combinations of various design specifications, are shown in Fig. 4.5. The mechanisms having three or more degrees-of-freedom result in increased product cost and control complexity, and have functional redundancies, but if such mechanisms generate superior Kansei impressions, such designs may have merits that outweigh the additional costs. Concerning the movement of the end-effector (robot hand), curvilinear movement may be inferior to straight line movement in terms of efficiency, but may be superior in Kansei aspects. Different acceleration characteristics before and after the constant velocity portion of the hand-effector movement may also generate different Kansei impressions. From a combination of various design specifications, the following ten product design alternative models are obtained:

62


B Manipulator Cup

A

Fig. 4.4 Manipulator transporting a cup

1. Two degrees-of-freedom, straight movement, symmetrical movement pattern 2. Two degrees-of-freedom, straight movement, unsymmetrical movement pattern 3. Two degrees-of-freedom, curvilinear movement, symmetrical movement pattern 4. Two degrees-of-freedom, curvilinear movement, unsymmetrical movement pattern 5. Three degrees-of-freedom, straight movement, symmetrical movement pattern 6. Three degrees-of-freedom, straight movement, unsymmetrical movement pattern 7. Three degrees-of-freedom, curvilinear movement, symmetrical movement pattern 8. Three degrees-of-freedom, curvilinear movement, unsymmetrical movement pattern 9. Four degrees-of-freedom, straight movement, symmetrical movement pattern 10. Four degrees-of-freedom, straight movement, unsymmetrical movement pattern The design objectives are (1) sufficient operability, (2) minimization of the product cost, and (3) satisfaction of Kansei requirements that the manipulator movement gives a good impression when cold beverages are delivered during summer or hot beverages during winter. Examinees, customers who are considering purchasing the product, evaluate the Kansei features and express their preference levels for the machine’s movement on individual computer displays. An ex-


63

ample case where a cup of cold liquid is delivered during summer is explained below.

(a) 2 degrees-offreedom

(b) 3 degrees-offreedom

(c) 4 degrees-offreedom

(i) Degrees-of-freedom in mechanisms

(a) Straight-line path

(b) Curved path

(ⅱ) End-effecter movement

0

Time t

(a) Symmetrical pattern

0

Time t (b) Asymmetrical pattern

(ⅲ) End-effecter velocity patterns Fig. 4.5 Alternative model specifications

Quantitative evaluation of Kansei feelings for product designs is generally difficult, but the semantic differential (SD) method is often used to measure people’s feelings for a given product by using a set of adjective word pairs [4]. The SD method is carried out as follows: Step 1. A set of adjective word pairs is prepared that expresses the Kansei features of the product being considered. Step 2. The examinees evaluate each product model, and provide numerical measures for each of adjective word pairs.

64


Step 3. Principal component analysis is conducted using the obtained numerical data, and the word pairs for the most effective Kansei impressions are selected. An evaluation map is constructed to express quantitatively the Kansei features for each product model. Table 4.1 shows the six adjective pairs used to evaluate the Kansei impressions created by manipulator movements. The range of the left and right column word evaluations in Table 4.1 has five segments, corresponding to magnitudes of ̶ 2, ̶ 1, 0, 1, and 2, with 2 indicating strong agreement with the meaning expressed by the right column adjective for the movement, while ̶ 2 indicates strong agreement with the left column adjective. A zero expresses an intermediate (neutral) impression. Each examinee enters magnitudes for each word pair for the movement of each manipulator model displayed on the screen, and selects the most preferable model according to their Kansei feelings. Table 4.1 Complementary adjective pairs for SD measures

X1 Ponderous

Nimble

X2

Simple

Sophisticated

X3

Common

Unique

X4 Mechanical

Human-like

X5

Deliberate

Speedy

X6 Cold (Uninviting)

Warm (Inviting)

The numerical data obtained for each pair of adjectives are analyzed using principal component analysis. The six principal components obtained are denoted z1 , z 2 , ..., z6 . The eigenvalue and eigenvector for each principal component are shown in Table 4.2. The eigenvalues express the variances, and increasing magnitudes of the variance corresponds to increased importance of the principal component. Each element of the eigenvector corresponds to each Kansei adjective, represented as X1, X2, …,X６．The sum of the eigenvalues equals the total number of elements, X1, X2, …, and X６, six in the case here. The contribution ratio of the first principal component Z１is 60.6％ ((3.63633/6)×100), while the contribution ratio of the second principal component Z2 is 31.0％ ((1.86112/6)×100). The sum of the contributions at the first and the second contribution ratios is 91.6%. Since the first and second principal components express Kansei impressions with a value in excess of 90%, it can be considered that the other principal components can be


65

disregarded. Thus, the principal components Z１ and Z2 are selected for the horizontal and vertical coordinate axis, respectively, as shown in Fig. 4.6. Table 4.2 Eigenvalues and eigenvectors for each principal component

Principal Eigenvalue component

Z1 Z2 Z3 Z4 Z5 Z6

Eigenvector

3.636 1.861 0.3163 0.1261 0.04587 0.01427

(0.362, –0.492, –0.355, –0.213, –0.494, –0.460) (–0.456, –0.126, 0.505, –0.641, 0.00921, –0.329) (–0.579, –0.433, 0.0831, 0.428, –0.473, 0.253) (–0.411, –0.288, –0.673, –0.207, 0.497, 0.0674) (0.306, –0.661, 0.396, 0.236, 0.507, –0.0116) (0.247, –0.187, 0.0404 ,–0.512, –0.169, 0.782) Human-like 0

Z2

Model 10

0

2 degrees-of-freedom manipulator Model 9

3

4 degrees-of-freedom manipulator

Ponderous

2

0 3

Model 7

Model 5

Model 4

1 Model 3

Model 1

5

Z1 Nimble

Model 2

2 8 Model 8

Model 6 3 degrees-of-freedom manipulator

Mechanical Fig. 4.6 Example of evaluation map for the manipulator designs

It can be seen that in the eigenvector of Z１ in Table 4.1, only X１ has a positive value, and X１ represents the adjective that expresses a nimble feeling, while in the eigenvector of Z2 in Table 4.1, only X4 has a large negative value, and the Z2 axis mainly expresses a range in feeling from human-like to mechanical. Each product model is then plotted in the coordinate space, and an evaluation map is constructed. Each circle on the map corresponds to a product model. The number of votes obtained via the questionnaire indicates which model is most preferable from the standpoint of Kansei feeling and is denoted Ei for model i . The number of votes

66


can be considered to be proportional to the popularity of the product model. In Fig. 4.6, the lightness or darkness of a circle expresses the number of votes obtained, with darker shades corresponding to a larger number of votes. The horizontal axis shows the level of nimbleness of the arm movement, where greater positive values correspond to feelings of increased nimbleness, while greater negative values indicate that the subject felt that the movement was heavier or more ponderous. Increasingly positive values on the vertical axis indicate the perception of increasingly human-like movement, while increasingly negative values are mapped to increasingly mechanical movement. In these maps, models having the same degrees-of-freedom, i.e., models having the same number of arm segments, are enclosed in a large circle, since such models tend to form groups on the map. Movements of models having two-degrees-of freedom are generally nimble, movements of models having three-degrees-of freedom are generally mechanical, and movements of models having four degrees-of-freedom are humanlike. In this way, the degree and quality of Kansei feelings generated by each product model can be appreciated by examining the evaluation map. Model 6 in Fig. 4.6 received the maximum number of votes indicating favorable Kansei impressions. This product model is a three-degrees-of-freedom manipulator which has the greatest initial acceleration and straight movement of the end-effector. The model having the next highest number of votes is model 2. To select the best solution, the product cost must be considered, in addition to the number of votes. On the evaluation map, the product cost increases along the direction indicated by the arrows. The thickness of the arrow between two models is proportional to the magnitude of the cost difference. The cost of model i is denoted Ci . The most popular model having the greatest number of votes is named as model S . The number of votes and the cost for model S are denoted E s and C s . Since the models indicated by the arrows beyond model S cost more than model S , these models can be excluded from further consideration. The models prior to Model S have smaller product costs, and have conflicting relationships with model S . Hence, based on the concurrent evaluations of the Kansei preference levels and the product costs, the best product model is selected from among the product models. Figure 4.7 shows the relationships between the number of votes and the product costs for product models 6, 5, 2, and 1 which exist prior to model 6. The dashed line that fits the points of models 1, 2, and 6 can be considered as a Pareto optimum solution line for the multiobjective optimization problem having the two objectives of maximizing votes and minimizing product cost. Since model 5 is located on a point inferior to the Pareto optimum solution line, it is excluded from further consideration. Then, by considering the tradeoff relationships among the product models on the dashed line, the most preferable product model is selected. The foregoing discussions correspond to an example where customers and makers collaboratively evaluate product design alternatives.


67

8

Model 6

7 6 Model 5

5 4 3

Model 2

2

Model 1

1 0

1

2

3

4

5

6

7

8

9

10

Number of votes Fig. 4.7 Relationship between number of votes and cost

Product designs that take Kansei impressions into account along with other conventional attributes are increasingly popular, and manufacturers cannot afford to ignore this. In the struggle to design especially attractive products for customers, objective attributes such as static displacement, product weight, and dynamic performance, for which quantitative evaluations are technologically possible, should be integrated with Kansei attributes and product cost constraints. The objective attributes and subjective Kansei attributes are usually evaluated independently. Such independent evaluations are effective when there are no interrelationships between the objective and subjective attributes, or when designs that have improved objective attributes also have higher preferences for subjective attributes. However, when the Kansei and objective attributes have conflicting relationships, for example in a sleek design for an automobile body that looks attractive but sacrifices interior spaciousness, concurrent evaluations of the Kansei and objective attributes are required [5]. Here, industrial robots are considered as a design model. As objective attributes, static displacement, the maximum vibrational displacement for a step force, product weight, and load capacity are evaluated, while the outward appearance is considered as a subjective attribute. The design variables are the dimensions of the product and its component parts, and the appearance of the overall shape and painted surface. Satisfaction functions expressing the satisfaction level for each of the objective attributes and the subjective attribute are constructed. To obtain the relationship between the five design variables and the outward appearance, each design variable is divided into three categories A, B, and C as shown in Table 4.3. The shapes of the product models are built from combinations of these categories, and are shown on a computer display. Examinees evaluate each product model, and give it a grade on a scale of 1 to 10. The data shown in Table 4.4 are obtained by calculating the average value of the evaluated magnitudes. These data are then

68


processed using Type I quantification theory multi-attribute analysis [6], and the satisfaction function u5 for the Kansei feeling is obtained. Table 4.3 Category variables for design variable values

Design variable

Category

x(1)

A 1.0

B 1.4

C 1.8

x(2) x(3) x(4) x(5)

0.1 1.0 1.0 0.0

0.25 1.5 1.4 0.5

0.4 2.0 1.8 1.0

Table 4.4 Design alternatives expressed via category variables and evaluated magnitudes

Design alternative

Evaluated magnitude

Design variable

1 2 3

x(1) C B A

30

A

x(2) x(3) A C B B C A

B

B

x(4) x(5) C A A A B A

B

C

1 4 6

9

To define the weighting coefficient for each attribute, examinees conduct pair comparisons for the attributes. In the example, the following weighting coefficients are obtained: ω1 =0.3437 for static displacement, ω2 =0.0681 for the maximum vibrational displacement, ω3 =0.1339 for product weight, ω4 =0.1369 for load capacity, and ω5 =0.3173 for outward appearance. Then the integrated preference function of the objective attributes and the subjective attribute, U 0 , is obtained: 5

U 0 = ∑ ωi ui i =1

(4.1)

To obtain product designs that provide the highest satisfaction levels for customers, as explained above, it is necessary to consider concurrently both objective and subjective attributes of the designs. It is sometimes said that since Kansei im-

4.3 Ergonomics

69

pressions are subjective, it is therefore impossible to create designs reliably that appropriately integrate Kansei impressions. However, even if the data representing people’s subjective evaluations appears scattered, it is often possible to extract common points or generalities that are particularly relevant to the product design being considered. When the target of a product design is focused on increasingly specific groups of customers, Kansei evaluations can provide increasingly useful support in product design decision-making.

4.3 Ergonomics Ergonomics is a field of study that focuses on considerations of usability, comfort, safety, and how to minimize fatigue or discomfort as people make use of various products, and the level of satisfaction that people derive when using products is an important aspect of product designs [7,8]. This implies that human physical and mental functions and features should be thoroughly considered when designing products that people will use. The aspects of ergonomics considered here have two fields of application: (1) ergonomics during product manufacturing processes, where the comfort of workers at manufacturing plants is evaluated, and (2) ergonomic factors that will be valued by the intended customers who will use the product. Figure 4.8 shows a procedural flow for designing products that incorporate ergonomic considerations. The scale arrangements of goods where human body scales are standard are shown in Fig.1.2. Determination of the appropriate scale is a fundamental requirement when designing products that people use, since it greatly influences product usability. A number of measurement units have their origin in the scale of the human body, or its parts, such as the foot and, correspondingly, the Japanese “shaku. ” In terms of a rational basis for ergonomics, it may therefore be useful to determine ergonomic product dimensions based on the scale of the human body. Virtual reality evaluation

Input of initial design

Design improvement considering ergonomics

Design proposal considering ergonomics

Ergonomics database Fig. 4.8 A design flow for designing products incorporating ergonomics

70


Figure 4.9 shows a design drawing of a forklift operated by a seated human driver. In designing products that workers ride and operate, the positions, shapes, and angles of pedals, handles, and levers used for controlling and carrying out various loading and unloading operations, as well as maintenance of clear sightlines, must all be determined with great attention paid to worker comfort levels, and maximal reduction of worker discomfort or exhaustion.

Region accessible by upper limbs

Handle knob Lever Region accessible by lower limbs

Sight line

Pedal Fig. 4.9 Operations in a forklift

Important standard criteria for designing products that implement good ergonomics include principles of motion economy, which is composed of the ten following aims: 1. Minimization of movement distance 2. Basing the movement directions of adjustments on natural movements 3. Elimination of sudden directional changes, and promotion of smooth or fluid movements 4. Concurrent use of both hands wherever possible 5. Use of symmetrical or aligned motions when using both hands 6. Reduction in the number of basic movement types 7. Election of reasonable movement sequences 8. Maintenance of appropriate movement speeds 9. Elimination of unnecessary or needlessly repetitive motions 10. Combination of separate motions into a single motion wherever possible

4.3 Ergonomics

71

When operating forklift machines, the muscular functions and accompanying motions of the limbs, arms, feet, and hands of the operator manipulate the pedals and levers so that mechanical tasks can be accomplished. In the human body, several muscles generally contribute to the motion and control of each joint. The various muscles contribute their forces and cooperatively generate the required complex and smooth joint movements and motion control. Various muscle forces are combined and regulated in an orderly fashion, and these can be determined according to certain criteria [9]. When simulating the application of various muscle forces with the aim of minimizing total physical effort, the muscular forces can be estimated by solving an optimization problem that minimizes the energy consumed [10]. To determine motion paths that will minimize worker fatigue when operating machines, analyses based on an optimization procedure can be effective [11]. The energy consumed by muscles differs according to the degree of force applied along various paths, as legs, arms, and the rest of the body move to accomplish the required tasks. The location, resistance, and travel range of the machine’s pedals and levers are important design factors that affect the energy that an operator must expend when operating the machine. The physiological energy consumed can be expressed as a function of muscle cross-sectional areas, lengths, contraction speeds, forces applied, and so on. The optimal movement paths and velocities of legs, arms, hands, and so on can be obtained by minimizing the sum of the energies consumed by muscles that satisfy equilibrium relationships of muscular torques and the forces required to rotate body joints. A schematic illustration of the musculature of a human right arm, based on anatomical data, is shown in Fig. 4.10. The points where each muscle is attached to the represented bones are based on human anatomy but, in actuality, muscle attachment points are somewhat spread out. To simplify the problem here, these attachment areas are modeled as points and the muscles that control the movement of fingers are omitted. The upper portion of the limb is modeled using 30 numbered muscles. The human arm can perform various kinds of movement such as grasping, and handling a part after moving it to a desired goal point. As an example of a simple task using such a movement, a conducted simulation is shown in the left side of Fig. 4.11, where a grasped part is pressed to the surface of a wall. The grasped part is cubic in shape, has edge dimensions of 6 cm, a mass of 0.1kg, and the operating force applied to the wall is 50 [N] . The results of the muscle force contributions obtained by solving an optimization problem are shown in Table 4.5a. Another movement example, shown in the right side of Fig. 4.11, illustrates two kinds of conducted simulation, torques of + and ̶ 50 N･cm that are applied around the center axis of the hand, corresponding to the tightening or loosening of a screw. The computed results of the simulation are shown in Table 4.5b. The movement that results in an optimum posture when the movement is finished, which corresponds to a posture where the least amount of energy was consumed by the muscles, is considered optimal. The elbow joint angle is then

72


changed by a certain amount and the optimization problem is again solved. The condition that returns a minimum value of the objective function is considered to be the optimum posture. For the movement shown in the left side of Fig. 4.11, the total energy consumed by the muscles was calculated and plotted as shown in Fig. 4.12, and the optimum posture was determined from the results. That is, the minimum objective function value of the consumed energy was obtained when the hinge joint angle was about 120˚, enabling optimal posture. 28 29

23 24 12

25

30

16 15

27

11

18

17

19 20 22 26

21

14 9 10

13 8 6

7 1 2

4

: Muscles (The numerical values show the number given for each muscle)

3

(a) View from left front

5

(b) View from right rear

Fig. 4.10 A musculoskeletal model for a human upper limb (muscles indicated by lines; numerical values are muscle numbers)

θ4

Motion A Fig. 4.11 Motion patterns

Operating force

θ4

Operating torque

Motion B

4.3 Ergonomics

73

Table 4.5 Estimated results of muscular force sharing for the static motions

(a) Motion A Muscle 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Muscle force (N) 86.25 85.86 89.35 0.0 0.0 35.60 0.0 0.0 45.88 74.54 216.36 206.66 0.0 0.0 0.0 0.0 0.0 173.18 153.89 60.64 260.98 129.56 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

(b) Motion B Muscle

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Muscle force (N) Torque Torque 50(N·cm) –50(N·cm) 0.0 0.0 0.0 0.0 0.0 9.16 0.0 0.0 0.0 2.60 8.69 15.40 0.0 0.0 0.0 0.0 0.0 2.13 9.24 8.02 10.27 11.13 0.0 0.0 0.0 2.76 0.0 0.0 1.66 1.06

0.0 0.0 0.0 0.0 0.0 0.0 22.45 19.90 10.27 0.0 23.21 0.0 0.0 0.0 0.0 0.0 0.0 13.02 14.55 1.75 28.40 0.48 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

74


0.0

40.0

80.0

120.0

140.0

Elbow joint angle θ4 (degrees) Fig.4.12 Relationship between elbow joint angle θ 4 and objective function ψ of the total energy consumed by muscles

4.4 Collaboration Circumstances The concept of collaboration explained in Chap. 3 is an important strategy for breaking through solutions obtained by conventional product manufacturing approaches, but the members of the collaborating group must be suitable. If the collaborating members are not well-suited in terms of knowledge, expertise, and other factors, the possibility of successful collaboration will be low despite great enthusiasm for the project. This chapter discusses collaboration circumstances, particularly with respect to the optimal allocation of human resources (i.e., worker, hereafter HR or HRs) to a specific project for product development. Figure 4.13 is a conceptual diagram of a company organization structure. Each HR usually operates within a single section in their area of competency. These sections are generally called divisions, and they contain vertically organized structures where designated HRs operate in various hierarchies. When a product development project requires a specific collection of HRs, they can be selected across divisions and gathered together according to the specific requirements, to accomplish best the project goals. The manager in each section usually considers the career paths of the HRs in his section, and can appropriately select individuals according to their expert abilities, gathering them as needed from various enterprise divisions. In this way, horizontal organizational structures that include various expert areas can be constructed.

4.4 Collaboration Circumstances

75

Company

Division A

Division B

Human resource a1

Human resource b1

Human resource a2

Human resource b2

Human resource a3

Human resource b3

Project i

Fig. 4.13 Conceptual diagram of company organization

The optimum allocation of HRs in a project should be logically decided based not only on workers’ technological knowledge and abilities, but also on their motivation, their career path, and the human relationships between the people in the group that is formed. How to handle these factors and develop an effective optimization procedure is an important technological subject, especially since descriptions based on qualitative evaluations must be translated into quantitative expressions [12]. The following explains an example. In certain companies, some HRs are assigned to more than one project. The allocation ratio of HR j to project i is denoted x ij , which expresses the participation ratio per unit working hour. Here, for simplicity, the allocation ratio is set equal to the skill ability allocation ratio of HR j for project i . The summation of x ij for all projects for HR j must be 1. For each project, the ratio of the sum of the HR skill abilities allocated to the project to the amount of skills needed to complete the project within the prescribed development period is expressed as the skill satisfaction level objective function f1 , which is to be maximized. Since a lack of available skills has the greatest influence on the feasibility of the project, f1 is the most important objective function in the optimization of HR allocation. The ability level of HR j for skill k is classified into four levels, {0, 0.5, 1, 2}, where “2” indicates that the person has sufficient knowledge and experience to teach others concerning the tasks at hand, “1” indicates a certain amount of knowledge and experience, “0.5” indicates fundamental knowledge but no experience,

76


and “0” indicates no knowledge or experience. Table 4.6 shows an example classification of HRs based on their skills. Table 4.6 HR skill ability matrix

HR1 HR2 HR3 HR4 HR5 HR6

Skill 1 2 1 0.5 1 2 0.5

Skill 2 0 1 0 1 2 0.5

Skill 3 0.5 0.5 2 1 0 1

Skill 4 0.5 0 1 2 2 0

The section manager usually has skills in mind for each HR, according to the person’s career path, and such skills are here called skills to be promoted. When skills to be promoted are included in the HR allocation problem for a project, the satisfaction level of the section manager concerning subordinate HR career paths is defined as the career path satisfaction level objective function f 2 , which is to be maximized. These satisfaction levels have numerical values from 0 to 1 as shown in Table 4.7, where increasing magnitude expresses increased career satisfaction among the HRs. Table 4.7 Satisfaction levels for HR career path


Project 1 1 0.25 0.5 0.75 0.25 1

Project 2 0.25 1 0.25 0.5 0.5 1

Project 3 0.25 0.5 1 1 0.5 0.75

Project 4 1 0.25 0.75 0.5 1 0.25

Next, principal motivating factors for workers themselves when participating in projects concern their own requirements for acquiring new skills or improving current skills, and the human relationships that affect the working environment. The satisfaction levels of project participants for skill acquisition are expressed as values from 0 to 1 where larger values are more preferable, as shown in Table 4.8. The satisfaction levels pertaining to human relationships in the working environment are similarly expressed, and shown in Table 4.9. If a given HR has experience working with a partner on the same project, the satisfaction level for collaboration with that partner is judged according to their compatibility. When collaboration experience is lacking, the satisfaction level is set as an intermediate value of 0.5. The summation of the above two satisfaction levels is defined as the motivational objective function f 3 , which is to be maximized.

Exercises

77

Table 4.8 Satisfaction levels of project participants for skill acquisition

Project 1 0.75 0.25 0.25 1 0.25 0.75


Project 2 0.25 1 0.25 0.75 1 1

Project 3 0.5 0.5 1 1 0.25 0.5

Project 4 1 0.5 1 0.5 0.75 0.25

Table 4.9 Satisfaction levels pertaining to human relationships in the working environment


HR1 0.75 0.5 0.5 1 1

HR2 0.75 1 1 0.5 0.5

HR3 0.5 1 0.25 1 0.75

HR4 1 0.5 1 0.5 0.75

HR5 1 0.5 1 0.5 0.75

HR6 1 0.5 0.75 0.25 0.75 -

The above mentioned three objective functions { f1, f 2 , f 3} are formulated as a multiobjective optimization problem and the obtained Pareto optimum solution sets are presented to a decision-maker so that the most suitable solution can be selected. Even if the numerical data shown in Tables 4.6 ̶ 4.9 and the optimization formulation are initially insufficient, improved modeling can be conducted on the basis of the evaluations and reconsideration of the obtained results, so that significant support for HR allocation decision-making can be expected, particularly after some trials are carried out.

Exercises 4.1 Discuss the importance of inherent human capabilities in product design and manufacturing. 4.2 Explain practical examples where Kansei requirements and product performance and functional requirements have conflicting interrelationships. 4.3 Clarify the reasons why effective collaboration is important in product manufacturing. 4.4 What evaluative factors are best to use when selecting members of a project from horizontal organizational divisions? Discuss the importance of these factors.

78


4.5 Discuss the influence that positive vs. passive collaborative attitudes held by people collaborating on project may have upon the collaboration results, in terms of differences in the optimum solutions. 4.6 Discuss the relationships between human body scales for product designs and the degree of comfort or ergonomics obtained when using such products.

References 1. Nagamachi M (1995) Kansei engineering: A new ergonomic computer-oriented technology for product development. International journal of industrial ergonomics 15:3 ̶ 11 2. Yoshimura M, Papalambros PY (2004) Kansei engineering in concurrent product design: A progress review. In: Proceedings of the TMCE2004, p.177 3. Yoshimura M, Horie S (2001) Aesthetic design of machine systems using evaluation maps. In: Proceedings of DETC’ 2001: ASME Design Engineering Technical Conferences and Computers and Information in Engineering Conferences, Pittsburgh, DETC2001/DAC21154:1 ̶ 9 4. Osgood CE, Suci GJ, Tannenbaum PH (1957) The measurement of meaning. University of Illinois Press 5. Yoshimura M, Yanagi H (2001) Strategies for implementing aesthetic factors in product design. International journal of production research, 39(5):1031 ̶ 1049 6. Hayashi C (1950) On the quantification of qualitative data from the mathematico-statical point of view. Ann.Inst.Statist.Math.:35 ̶ 47 7. Galer I (ed) (1987) Applied ergonomics handbook. Butterworths, London 8. Chapanis A (1996) Human factors in systems engineering. Wiley, New York 9. Penrod DD, Davy DT, Singh DP (1974) An optimization approach to tendon force analysis. J. Biomech. 7:123 ̶ 29 10. Crowninshield RD (1978) Use of optimization techniques to predict muscle forces. Trans. ASME , J. Biomech. Eng., 100:88 ̶ 92 11. Yoshimura M, Masui H (1992) Prediction of muscular force sharing in a human upper limb and determination of optimal limb motion. JSME international journal, 35(330) C:574 ̶ 581 12. Yoshimura M, Fujimi Y, Izui K, Nishiwaki S (2006) Decision-making support system for human resource allocation in product development projects. International journal of production research, 44(5):831 ̶ 848

5 Product Manufacturing Support Technologies

Despite the availability of useful Kansei information and tailor-made manufacturing capability, optimal decision-making will still be elusive in circumstances where there are numerous evaluation factors, characteristics, and design variables. Systems that support a broad range of human abilities, and facilitate their full expression and utilization, are required. This chapter explains the characteristics of such representative supporting systems: product shape description technologies, technologies for analysis of performance characteristics, technologies that support generation of product ideas, database technologies, manufacturing support technologies, and other technologies. The importance of information network systems is also discussed, since these systems and technologies play crucial roles in enabling better decision-making during product design and manufacturing.

5.1 Representative Supporting Systems The major technologies that support decision-making in product manufacturing are shown in Fig. 5.1. The focus of the decision-making process is at the center of the figure, and represents the objective of each stage of the manufacturing process. Items located in the ring nearest the center represent optimization methods and decision-making support methods. Proceeding outward, items in the next ring represent network-information systems, a shared database, and virtual reality (VR) technologies that the inner processes rely upon. Finally, a range of technologies that support complex product manufacturing are arrayed in the represented segments beyond the core, with a number of different classes noted in the outermost ring. Over time, technologies are developed so that better judgments and decisions can be made more quickly and easily, and implementations carried out will more truly reflect personal abilities, personalities, and Kansei information. Each of these technologies is explained below.

80


Quality design Scheduling, production planning

Digital mock-ups

Network-information systems,

Modular design

Rapid prototyping

Decisionmaking

QFD

CAE Structural analysis

Kansei information processing system

ERP

Simulation

PDM TRIZ

System supporting group generation of ideas

Database Fig. 5.1 Technologies that support decision-making in product manufacturing

5.1.1 Product Shape Description Technologies

5.1.1.1 Three-Dimensional CAD Technology Figure 5.2 shows the flow of product manufacturing which begins with the use of CAD three-dimensional (3-D) product designs. The representation and interpretation of the shape of workpieces or product parts in the form of data that can be

5.1 Representative Supporting Systems

81

processed by computers is one of the most important technologies in modern product manufacturing. Product design using three-dimensional CAD

Trial manufacturing CAD/CAM

Manufacturing CAM

Testing CAT

Structural analysis Simulation CAE Fig. 5.2 Computer-aided technologies in product manufacturing flow

Solid models (three-dimensional models) in which the shapes of parts and products are digitally displayed have revolutionized various aspects of the product design and manufacturing process, as these enable personnel to understand quickly the details of shapes to be manufactured, evaluate physical characteristics such as center of gravity and weight, and facilitate the output of CAM information, automatic testing of manufactured parts, and so on. Three-dimensional displays of product and part shapes are easier to understand than flat drawings, which reduces the possibility of misunderstanding when transmitting shape information. CAD data acquired or produced at the design stage can be directly linked with manufacturing processes, which reduces the risk of manufacturing inferior or erroneous items, increases the potential for higher quality production in shorter timeframes, and so on. Furthermore, comparison of manufacturing data with CAD data makes automatic CAT (Computer-Aided Testing) for quality possible. By incorporating this process at the early stage of manufacturing, poorly manufactured or substandard parts can be exposed and information concerning troublesome aspects that are interfering with smooth operations can be fed back to preceding processes. In short, various examination results can be front loaded, which facilitates the practical use of the concurrent engineering concepts explained in Sec. 3.3. Finally, perhaps the strongest merit gained when using three-dimensional CAD processes is to reduce product development periods as well as manufacturing costs, since various design problems and obstacles can be uncovered and resolved at the early stages through the sharing of CAD data across the enterprise. Ideally, product manufacturing should rely on unified data sets that include the data for product design, analyses, manufacturing, testing, and the like, with product and parts shapes and other data all digitalized. This technology is called digital engineering and is an increasingly vital requirement for gaining maximum benefit from concurrent engineering and collaboration, where design and manufacturing information is made visible and easy communication among a wide range of people engaged in product manufacturing is the norm.

82


5.1.1.2 Digital Mockup Technologies Digital-mockups enable the display on a computer screen of the conditions in which parts are to be assembled. This technology of model mockups assembled according to practical production details enables rapid decision-making. Based on the examination of displayed product models, trial manufacturing processes can be streamlined, and the practical production manufacturing processes can be started more quickly after the product design work is completed. 5.1.1.3 Rapid Prototyping Technologies Rapid prototyping is a technology through which prototype models having the same shapes as potential products can be quickly made [1]. These are usually crafted in a different material than will be used for the final, practical parts, but are based on the same CAD data. This is especially useful because it allows personnel to check physically the suitability of part designs, and facilitates the discovery of design and manufacturing points that need improvement. The trial manufacturing of actual practical parts is usually very costly in terms of both time and money, so rapid prototyping is becoming more popular since it simplifies this process. Figure 5.3 shows the flow for the making of prototyping parts from three-dimensional (3D) CAD models using what is essentially a three-dimensional laser printer type of device. The resulting prototype parts represent the shapes and features of their designs with reasonable accuracy, and are superior to three-dimensional screen displays for directly giving both visual and tactile feedback to designers. That is, a prime advantage of rapid prototyping is that it permits evaluations that depend on tactile as well as visual senses. 3-D CAD input model of a part

Generate 3-D solid data

Generate layered slice cross-sectional data

Sequentially harden resin layers using laser

Prototyped part Fig. 5.3 Flow of rapid prototyping using stereolithography


83

5.1.2 Technologies for Analysis of Performance Characteristics Structural analysis methods such as the finite element analysis method, which can analyze the dynamic characteristics of a product in its operating state using computer software, can accurately evaluate stresses, vibrations, thermal deformations, hydrodynamic and electromagnetic behavior, and so on, at the design stage, enabling the construction of better and more practical complex structural shapes. The analysis of product performance characteristics and various allied simulation technologies are generally called CAE (Computer-Aided Engineering). CAE concepts were first proposed around 1980 by J.R. Lemon, an associate professor at Cincinnati University, USA [2]. At that time, vibrational analyses of machine tool structures were being done and he founded the Structural Dynamics Research Corporation, a pioneering engineering consulting company in its time. The finite element method (FEM) is the most widely employed technology for obtaining various dynamic characteristics at the design stage, using computer simulation. With this method, the characteristics for parts and elements at the design stage can usually be accurately obtained. Product simulations based on the FEM are particularly effective in the quest to obtain better and sometimes optimum design solutions while minimizing the need for trial manufacturing. However, there are significant limitations with the FEM for machine products when dealing with design areas that currently lack sufficient theoretical elucidation, such as contact joints, frictional dynamics, nonlinear characteristics, and abrasion. Thus, many problematic subjects remain to be solved in the field of CAE. Fig. 5.4b shows an example of a CAD display for a die-plate for a mold clamping unit used in the injection-molding machine pictured in Fig. 5.4a. A toggle mechanism makes the plate open and close. The CAD model was transformed into a CAE model so that the static and dynamic characteristics could be analyzed, and shape optimization of the plate was conducted to minimize the structural weight while keeping the rigidity constant, to improve the dynamic performance, and reduce the material cost. An example of the optimized results is shown in Fig. 5.5. The traction method [3] was used in this shape optimization. An advantage of this method is that the phenomenon of wavelike deformations in the shape surface that often appear during shape optimizations are avoided, and the optimized results can easily be transferred to practical manufacturing processes. During the optimization process, external forces are applied vertically to receiving area of the die plate as shown in Fig. 5.6a. As the boundary condition of the shape optimization, a shape restriction was applied so that the shape changes will not occur in locations where other parts are present during assembly. Figure 5.6b shows the appearance of the part, particularly the shape-restricted holes, where the toggle links will be installed, that are preserved from shape deformation. Using this kind of shape optimization, the weight of the plate was reduced by 16.7% from the origi-

84


nal design. After shape optimization, the CAE model can be modified further to take specific manufacturing factors into consideration.

Fig. 5.4 Toggle mechanism and die-plate for mold clamping in injection-molding machine

Fig. 5.5 Shape optimization result for a die plate


85

Fig. 5.6 Conditions of applied load and shape restriction

5.1.3 Technologies that Support Generation of Product Ideas The main feature of collaboration is that group members cooperatively achieve a goal by making use of each other’s knowledge, information and ideas, as explained in Sect. 3.4. The advantages that effective collaborations provide are not confined to the increased scale of knowledge and information that can be accessed within the group of participants, but also extend to mutual influence and stimulation, which can lead to fresh problem-solving approaches, so-called synergy effects. Brainstorming [4] is perhaps the most famous method that supports a creative group process. The KJ-method developed by Kawakita supports creative activity both for an individual and a group [5]. In the KJ-method, participants write their ideas on cards and arrange them according to a rule that aims to simulate the participants’ creativity. People are continuously recalling various bits of information and acquiring fresh stimuli from their immediate surroundings and environment. When struggling to create fresh ideas or improve existing ones, people utilize a wide range of information, such as past knowledge, recent experiences, unexpected insights, and so on. The collective sum of such information presents a wealth of potential connections and suggestions that can trigger a sudden inspiration, which creative thinkers can turn into fresh ideas. Here, this phenomenon is called the chain reaction process. Having a sudden inspiration or creating a new idea is only the first step. The initial inspiration or idea is still like a rough stone and needs additional work, refinement, and polishing in order to bring it to a useful level of sophistication. The process that takes an initial, half-formed idea and develops it into something useful is here called the refinement process.

86


Using the above mentioned two sub-processes, the creative process can be represented as a sequence of cycles consisting of these processes. In a typical cycle, a person creates a new idea in a chain reaction process, and then develops the created germ idea into a more sophisticated entity via the refinement process. The communication process in a creative group activity is shown as in Fig. 5.7. The chain reaction and refinement processes for creative group activity are repeated. During creative group activity, various kinds of information flow back and forth in one-to-one and one-to-many forms of intercommunication among group members. This information flow stimulates members’ creativity and increases the chance that group members will have new inspirations and create fresh ideas. When one group member presents a new idea, all members discuss it and, through further development and refinement, raise it to a more sophisticated level. Based on a basic model of creative group processes, a procedure for visualizing this process during collaborative design is constructed [6,7]. The model shown in Fig. 5.7 represents group activity as a collection of cycles consisting of chain reaction and refinement processes. To construct an intercommunication procedure and enable it to be visualized and presented during collaborative design work based on this model, a basic unit of creative activity is defined, called a session. Each session consists of a single initiating action and several subsequent actions, related to or triggered by the initial action. Here, the former is called a trigger action, the latter a response action. When a member of the design team presents a trigger action, such as presenting a new idea, a new session is started and then developed as members present response actions, such as when expressing opinions.

Fig. 5.7 Creative group process


87

By using trigger and response actions, the procedure for a typical session consists of the following three steps. Step 1. Trigger action: when a designer creates a new idea or encounters a problem, he presents it as a trigger action and a new session starts. Step 2. Response action: team members present a response action when they have opinions, new ideas and any kind of problem that arises during a session. Designers can freely present a new response action related to a trigger action or to any previously presented response action during the session. Therefore, the relationship between a trigger action and subsequent response actions can be displayed as a tree structure such as shown in Fig. 5.8. It is difficult for designers to manage and find their way along the path of such non-sequential communications, but a support system including such displays can aid designers by recording the history of trigger and response actions and displaying this history in a graphically useful manner. Step 3. Check if there are many new response actions: if there are any new response actions, return to step 2. If there are no new response actions, the particular session is finished.

Trigger action Response action (1) Response action (2) Response action (4) Response action (5) Response action (3) Response action (n) Fig. 5.8 Trigger and response actions and associated tree structure

During a session, several ideas can be presented. However, from the viewpoint of the chain reaction process, a new idea is not always related to the stimulus that triggers it, as when a designer has an inspiration that seems unrelated to current stimuli. Furthermore, a topic handled in a session may change during the session, so that a given topic at the outset may become completely different after evolution within a single session. In these cases, such ideas or topics are best treated independently from the present session, because this makes it easier for designers to understand the intercommunication flow. Here, this type of operation is called a session fork. When people attempt to create a new idea, this is rarely done in a single step. Rather, to try and solve the problem creatively, a general direction of thought and

88


exploration is first considered. Subsequently, details are added and a viable idea may sprout from a germinal seed concept. Here, the development of concepts between the initially set target goal and the achieved creative idea is called an approach. During the design process, as various approaches are considered, designers’ thought processes become more coherent, the design goals come into clearer focus, and productive ideas can be created more effectively. During design activities, designers create new ideas in two steps, the exploration of approaches and the gradual realization of ideas. However, designers do not necessarily distinguish these steps and in some cases they may unconsciously be intermingled. Here, these steps are clearly separated and the designers cooperate so that the two processes can be handled discretely. The advantage of clearly separating these steps is that this increases the potential yield of useful ideas from various approaches. When approaches and ideas are considered separately, the process where designers create finished ideas from target goals can be represented as a hierarchical structure consisting of three layers: target goals, approaches, and ideas. There may be several approaches that achieve a given target goal and several approaches that result in a useful idea. Furthermore, most simple hierarchical structures of the design process have one goal, a single approach and one idea, but in many cases, more alternatives exist in the lower portion of the hierarchical structure. In addition, one idea may result from several concurrent approaches, or, in some cases, a single approach may achieve several target goals. If these cases are considered, the overall hierarchical structure of a design project can be represented as a large and intricate structure consisting of numerous target goals, approaches, and ideas, as shown in Fig. 5.9. Here, the hierarchical structure is called a link-structure. The session method provides a procedure for cooperation when creating ideas during the collaborative design process, but it does not suggest the direction of exploration as the designers attempt to create ideas. Using the link-structure allows designers to focus more easily on the specific ideas that are most likely to result in a product that best achieves the design goals. Designers cooperate to create approaches and ideas as they conduct numerous sessions, while the development of the link-structure that records and organizes the creative process allows a larger number of ideas to be tracked and developed than would be the case without such a procedure. The link structure also clarifies which ideas are superior and have the most potential to result in a successful product design. Figure 5.10 shows the overall flow of the design process, in particular, the creation of a conceptual design proposal. This procedure consists of following two phases: the generation of ideas and the discussion of a design proposal. First phase: generation of ideas. In the first phase, designers attempt to create as many ideas as possible, using the session method to create ideas and approaches, and the link-structure as a tool for exploring the direction of new ideas and approaches. Note that in this process, ideas are generated, but the relationships among these ideas are not considered.


89

Second phase: discussion of the design proposal. In the second phase, designers select and combine ideas from the link-structure developed in the first phase, to create a feasible design proposal. All of the designers discuss and evaluate the created design proposal and decide whether or not to adopt it. If the design proposal is adopted, further design processes cease. If the design proposal is not adopted, the designers analyze the obstructive problems in the design proposals and they then either return to the first phase or create a new design proposal for further second phase discussion. Both the first and second phases proceed and are managed using the session method. A new session starts whenever a designer presents a new design proposal as a trigger action. A session finishes when the discussed design proposal is adopted as the final design proposal, or when designers have finished analyzing problems with the discussed design proposal after deciding to reject it. Goal (1)

Approach (1)

Idea (1)

Goal (2)

Approach (2)

Idea (2)

Approach (3)

Idea (3)

Idea (4)

Fig. 5.9 Link structure of goals, approaches, and ideas

Start

Generation of approaches Generation of ideas Generation of ideas

Generation of design proposal Discussion of design proposal Discussion

Rejected

Adopted End

Fig. 5.10 Overall flow of design process

Idea (5)

90


5.1.4 Database Technologies If data concerning product manufacturing are separately managed in each division or section, the beneficial implementation of concurrent engineering and collaboration concepts is impossible. Figure 5.11 shows a conceptual illustration of the Product Data Management (PDM) method in which product and part data pertaining to product planning, development, design, manufacturing, marketing, maintenance, etc. are collectively managed. The data include part configuration tables, design drawings, specifications, design manuals, part management data, material ordering (procurement) data, cost information, analytical results of part strength, and so on. Product Planning

Development

Product design

Product Data Part configuration tables, Design drawings, Specifications, Design manuals, Part management data, Material ordering data, Cost information, Analytical results for parts strength, and so on

Collective management

Maintenance

Manufacturing Marketing Fig. 5.11 Conceptual diagram of PDM

TRIZ, a database system for storing and accessing ideas and knowledge concerning past product manufacturing, may be used to support product development processes, and may also be used for development of new products [8]. The acronym TRIZ is an English derivation based on the initial letters of Russian words meaning Theory of Innovative Problem Solving. The theoretical construction and practical use of this approach started in Russia in 1946. Genrich Altshuller, the originator of TRIZ, believed that the mental processes carried out when people invent and solve problems obey certain laws. He surveyed approximately 400,000 patents to gather supporting data, and then developed formulas or rules that could be applied during the work of producing new in-


91

ventions. TRIZ is fundamentally based on the concept that technological systems evolve according to various laws or rules. In this method, problem descriptions are first defined, the problem is modeled, crucial issues are determined and examined, and practical solutions are ultimately deduced from ideal solutions using the so-called material-field analysis modeling method and an algorithm for solving core problems in the invention process. Essentially, guiding principles for solving the problem in hand are given based on the knowledge of past technological examples, and new design solutions are obtained from combinations of existing technologies. In general, the scope of knowledge and thinking of designers is limited, but TRIZ aims to transcend these obstacles.

5.1.5 Manufacturing Support Technologies

5.1.5.1 Bottlenecks in Supply Chain Management Theory of Constraints (TOC) is a known technological methodology for globally optimizing the entire flow of specific manufacturing processes, from acquisition of raw materials all the way to marketing, which is a supply chain management (SCM) goal. In the TOC method as presented by the Israeli physicist Goldratt [9], the bottleneck portion of goal-oriented processes becomes the active constraint in the optimization problem, i.e., the constraint that most critically influences the optimum solution. Thus, when improvements that resolve one or more bottlenecks are serially found, the flow of the manufacturing processes will likely approach a globally optimum solution. In SCM, each process or activity is considered to be connected in series, as links in a chain. The maximum throughput of the entire collection of processes will be limited by the throughput of the process or chain joint that becomes the bottleneck. In other words, when the production output for a given unit of time is a criterion, the process having the lowest productivity constrains the production output as a whole and acts as a bottleneck. Therefore, production output is critically dependent on the productivity of the bottleneck process, and even if each process is independently optimized for productivity, optimal output of the entire manufacturing process will not be obtained. The foregoing is an outline of the TOC, but with the TOC, heuristic partial optimization of the system is iterated, and a globally optimum solution from a theoretical standpoint cannot be achieved.

92


5.1.5.2 Design Incorporating Quality Engineering When producing parts whose detailed designs have already been determined, the most important technological subject in product manufacturing is finding ways to reduce manufacturing cost without spoiling the quality of the results. In designs incorporating quality engineering, the manufacturing cost can be minimized while preserving the required quality of produced parts by concurrently considering the dimensional accuracy of the parts and their surface quality, together with the manufacturing cost. Factors that cause unwanted variations in product qualities and functions are called quality loss factors, and result from variations in environmental conditions, corrosion, lack of uniformity in material characteristics, manufacturing process variations, and so on. In the Taguchi method [10], a loss function is used to express the magnitude of such losses due to variations, which customers are especially sensitive to: Loss due to product performance variance= constant × (characteristic value y – goal value m )2 where the square of the difference of the characteristics for the goal value expresses the loss. The loss here corresponds to the idea of a dissatisfaction criterion as explained in Sect. 2.2. In conventional product manufacturing, design variables (parameters) are determined so that specific values for functions, performances, and qualities can be achieved, and subsequently, or after manufacturing, the scattering of these factors, that is, the degree to which actual product values depart from the design values, is evaluated. Usually, attempting to reduce scattering increases manufacturing costs significantly, as this involves improving machining accuracies, using higher quality materials, and so on. When concurrently considering the relationships between manufacturing cost and the scattering of characteristics, influences differ according to parameter settings or conditions for each characteristic. In the Taguchi method, under the condition where many parameter combinations can achieve the objective function goal or performance values, the combination of parameters that yields the smallest scattering of characteristics is first searched for, and then specific parameters, termed adjustment factors, are adjusted so that the performance or function values approach the goal values as closely as possible. Adjustment values that have little influence on the scattering of the characteristics but greatly affect the values of the characteristics are used when adjusting the characteristics to the goal values. In the Taguchi method, the variances of characteristics or functions are evaluated by using SN ratios, which are measures for estimating function robustness,


93

i.e., for evaluating stability levels. These SN ratios are analogous to ratios of signal vs. noise in communication engineering, and are obtained as follows: SN ratio =

Magnitude of function efficiency β 2 = Magnitude of function error α2

(5.1)

where β expresses the sensitivity and α is the standard deviation as the error measure. The squares of β and α are used for the SN ratio because the loss is proportional to the square of the deviations. The foregoing design method for reducing the deviation of the characteristics is called the parameter design method, in which robust designs for a product composed of many parts is first created and then reductions in the product manufacturing cost are sought. The parameter method does not aim to obtain optimum designs, but aims to reduce variations in the characteristics. Thus, design solutions obtained by this method cannot be considered optimal. Designs in which the variations of characteristics are evaluated are generally called robust designs, and to achieve these, the following two types of methods are used: 1. Design methods assuring that design constraints will be satisfied even if there are variations in the characteristics 2. Design methods that aim to reduce the sensitivities of the designs for variations in the characteristics The Taguchi robust design method corresponds to type 2 above. When attaining a particular level for the performance characteristics is important, a type 1 method is used, and the robust design optimization method is based on the optimization procedures used [11 ̶ 13]. Concerning how the scattering of design variables or parameters causes variations in the performances, qualities, and functions of a manufactured product design, the following two situations exist: 1. Cases where probability analysis is used: a tolerance of ±3σ for variables or parameters is defined 2. Cases where worst case analysis is used: a tolerance of ± Δ expressing the maximum variance from the mean value, that is, the maximum allowable tolerance, is defined For cases following number 1 above, the variances of design variables are incorporated in probabilistic expressions in the formulation of the design optimization, while for number 2, the variance ranges of the objective function and the constraint functions are set as the worst case conditions, considering variations resulting from the absolute value of 3σ . When the variables incorporating variations are denoted d j (j=1,2,…,n) the transmitted variations in constraint function gi and objective function f can be approximated using first order estimates of the Taylor series. For example, the

94


standard deviation σ g i of constraint function g i for the standard deviation σ d j of d j is obtained as follows: n

⎛ ∂g

⎞

⎝

⎠

2

σ g2 = ∑ ⎜ i σ d j ⎟ i ⎜ ⎟ j =1 ∂d j

(5.2)

For cases following number 2 above, when the worst value of the variation for variable d j is Δd j , the value Δgi transmitted to constraint function gi is obtained by linear estimation as follows: ∂g i Δd j ∂ j =1 d j n

Δg i = ∑

(5.3)

For cases following number 1 above, the constraint in the optimization problem is transformed as follows: g i + kσ gi ≤ 0

(5.4)

where value of k is 3 when the tolerance is ±3σ , while for number 2, g i + Δg i ≤ 0

Figure 5.12 shows the design variable space for design variables d1 and d 2 that have variances. When variations are not considered, the constraint functions g1 and g 2 are plotted as shown; however when variations are considered, these lines move to g1′ and g 2′ . The feasible region becomes smaller, and the optimum design solution shown by the intersection of the objective function contour lines moves from A to A'. g′ =0 1

g2′=0


A′

g2=0

A

0


d1

Fig. 5.12 Feasible design region changes due to transmission of variations


95

When optimizing cases corresponding to number 2 above, the transmitted variations should be minimized, and (5.2) or (5.3) is included in the objective function of the optimization. 5.1.5.3 Evaluation for Ease of Assembly Most manufactured products are composed of a number of parts, and the time and financial cost required for assembly is often a very considerable percentage of the product manufacturing cost. The assembly time depends on the product design. Methods that can quantitatively evaluate assembly difficulties for particular product designs can effectively reduce assembly costs, and one such method has been proposed by Boothroyd [14, 15]. Figure 5.13 shows how part shape influences the difficulty of assembly for simple solid shapes that must be inserted into corresponding holes by an industrial robot. α is the maximum rotational angle of the part around the axis perpendicular to the direction of insertion, while β is the maximum rotational angle of the part around the part’s insertion direction. As the magnitudes of α and β increase, the time required to insert the part also increases, due to the increased difficulty of assembly. α β

(a)

α =180°,β = 0°

(b)

α =360°,β = 0°

(c)

α =180°,β = 90°

(d)

α =180°,β = 180°

α　:　Maximum rotational angle of part around axis perpendicular to direction of insertion

β　:　Maximum rotational angle of part around part's insertion direction

Fig. 5.13 Part insertion pattern during assembly

96


The ease of maintenance and replacement of parts during the time products are in use, and the reuse and recycling of parts and materials after the product ceases to be useful, depend on the ease with which the product can be disassembled. These considerations are increasingly important due to the need to address environmental concerns, and should be addressed at an early stage during the concurrent design of products, as explained in Sect. 3.3. 5.1.5.4 Group Technology When the manufacturing paradigm evolved to job shop production, and later to production to order, the need to manufacture a wide variety of products became increasingly important. The pressure for variety tends to increase manufacturing cost, and how to reduce these costs successfully while achieving production variety is now one of the most important subjects in the field of product manufacturing. Reductions in product manufacturing cost depend on the use of design and manufacturing strategies that offer the greatest variety of products when assembled from as small a number of individual parts as possible. To do this effectively, the concept of group technology (GT) from the product design stage is often considered, although GT is usually applied at the manufacturing stage. GT was first proposed in 1958 by Mitrofanov, who made grouping of parts for manufacturing the fundamental practice [16]. The grouping of parts is conducted by coding the design and manufacturing attributes of the parts. Many part classification methods and coding systems have been developed, and most use number digits. Figure 5.14 shows a conceptual example of simplified part shapes and their corresponding codes. The first group of digits expresses the fundamental outside shape of the part, the second group of digits deals with secondary inside shapes, part dimensions and principal manufacturing processes, while the third group of digits encodes additional shape data such as holes, threads, and grooves, and other manufacturing processes. GT is a concept for increasing productivity by taking a collection of parts and grouping them according to their similarity in terms of shape, dimensions, and manufacturing processing, which leads to reductions in product manufacturing costs. GT has been used to increase the efficiency of manufacturing processes, applied to an existing product design, but for maximum utility it must be applied from the product design stage, based on the concept of concurrent engineering [17]. Using GT from the product design stage brings about the following advantages pertaining to product manufacturing processes:


1st digit Fundamental outside shape

2nd digit Secondary inside shape and fundamental dimensions

97

3rd digit Additional shapes such as holes, threads, and grooves

Fig. 5.14 Conceptual diagram of GT

1. During product design, the number of parts that must be newly designed can be reduced, and the automation of design processes and creation of NC programs for machining become easier. 2. During production planning and management, process planning and scheduling become easier. 3. During manufacturing, manufacturing paradigms can be changed from a job shop paradigm where work contents and processing are different for each job, to a lot or mass production paradigm. Since various jobs can be grouped and applied to the same group of parts, processed with the same jigs and tools, set-up times can be reduced and effective reductions in manufacturing costs can be obtained. Manufacturing costs can also be reduced by using identical or nearly identical parts as much as possible when manufacturing products or families of products. However, as shown in Fig. 5.15, when manufacturing cost is reduced by grouping parts, the product performance and quality may be degraded. If this occurs, the conflicting relationships between the positive and negative consequences of grouping parts should be concurrently evaluated. For example, when the manufacturing cost is not considered at the product design stage, the design variables of parts will have different values. In other words, when the requirement for the product performance is selected as the single objective function, a product design having the minimum objective function value (in cases where minimization is preferable) usually has design solutions where all parts of the product have differ-

98


ent design variable values. In such cases, the manufacturing cost is likely to be very high. As shown in Fig. 5.15, the region between P1 and P2 corresponds to the area where the greatest reduction in manufacturing cost can be achieved with the least degradation in product performance.

P3

P4

P2

P1 0

Magnitude of product performance degradation　ΔP

Fig. 5.15 Relationship between manufacturing cost reduction and product performance degradation for parts groups

In the region between P3 and P4, the manufacturing cost reduction as a consequence of parts grouping is small, while the product performance is unacceptably degraded. To take best advantage of GT, the incremental value of the objective function value of the product performance from the minimum objective function value where all parts have different design variable values should be evaluated. The design solution that has the minimum manufacturing cost from among the design solutions where the incremental values of the objective functions are less than the permissible value is selected as the optimum design solution. 5.1.5.5 Process Planning The purpose of process planning is to select the optimum manufacturing method and procedures from among an existing set of feasible manufacturing methods and procedures whenever a variety of alternatives are available [18,19]. The usual criterion in the optimization is to minimize the manufacturing cost under the constraint of satisfying the required characteristic (e.g., accuracy) of the part. The im-


99

portance and significance of process planning in product design optimization is explained below. (1) Relationship between the accuracy required for a part and manufacturing cost Precise machining of contact surfaces of parts that will be joined together is a major requirement. The relationship between the machining cost for the contact surface and the required machining accuracy is conceptually expressed as shown in Fig. 5.16.

Machining accuracy Fig. 5.16 Relationship between machining accuracy and machining cost

The higher the required machining accuracy, the greater the machining cost, as shown in above plot which displays a quadratic relationship. (2) Relationship between machining accuracy and product performance Figure 5.17 shows the conceptual relationship between the machining accuracy used when manufacturing a part and the product performance of the product for which the part is a component.

A

Machining accuracy Fig. 5.17 Relationship between machining accuracy and product performance

100


Generally, higher machining accuracy when manufacturing machine parts will bring about higher product performance; however, as shown in Fig. 5.17, there is a point beyond which additional increases in machining accuracy (at additional manufacturing cost) fail to increase the product performance. To achieve optimal manufacturing parameters for a given product, only design solutions that lie on the Pareto optimum solution line for the relationship between the product performance and the product manufacturing cost should be selected, as shown in Fig. 2.11. As can be understood from Figs. 5.16 and 5.17, following a requirement for unnecessarily high machining accuracy of parts will unnecessarily increase the machining cost while providing no practical benefit. In such cases, the solution will lie inside the feasible region shown in Fig. 2.11 rather than on the Pareto optimum solution line. (3) Alternatives in part manufacturing When manufacturing machine parts to required shapes and accuracies, there may be a number of alternative design and manufacturing solutions. The following are representative manufacturing methods for machine parts: 1. Machining 2. Forging 3. Casting 4. Welding Which of the above alternatives should be selected depends on the factors explained below (5). Often, however, choices are made based on data derived from experience. When the parts to be machined have complex shapes, there are many machining alternatives for removing material and finishing surfaces. Parts that are initially forged, cast, or welded will probably have to undergo some combination of machining procedures such as milling, grinding, super polishing, and others before the manufacturing process is completed. When the highest machining accuracy is required, super polishing may be used at the last step, but when the required accuracy is not so high, only simple milling may be sufficient. (4) Determination of the best alternative for part manufacturing Figure 5.18 expresses the conceptual relationship between the accuracy required for a machine part and the minimum manufacturing cost that will achieve the required accuracy. The curve corresponds to the Pareto optimum solution line for the multi-objective optimization, maximizing the machining accuracy and minimizing the manufacturing cost. In Fig. 5.18, for example, machining methods a and b are used to achieve the desired accuracy at point 2. The required cost for each machining method is shown on the vertical axis and the sum of these costs is plotted on the Pareto optimum solution line. In this example, “a” corresponds to milling, “b” to grinding, “c” to super finishing, and “d” to laborious hand honing.


101

d c

a 1

2

b

b

a

a 3

Accuracy required for a machine part Fig. 5.18 Relationship between accuracy required for a machine part and minimum manufacturing cost to achieve it

In general, a larger number of manufacturing processes is needed to achieve higher accuracy of machine parts, machining processes that incur increasingly large cost per unit area as the degree of precision increases. The goal of process planning is to obtain and adhere to Pareto optimum solutions for manufacturing processes such as shown in Fig. 5.18. (5) Factors related with selection from a number of manufacturing alternatives The following factors influence the selection of optimum manufacturing alternatives: 1. The number of parts to be manufactured 2. Part material 3. Part dimensions and shapes When a large number of parts must be made, forging, die casting, mold pressing and other techniques can be used despite potentially high initial setup costs because the total cost is supportable in view of the expected sales. On the other hand, when a single part or a small number of parts are to be manufactured, machining and/or welding are often used due to their relatively low cost. Concerning part materials, when weight must be minimized, aluminum or other sophisticated metal alloys can be used. Similarly, when high strength is a primary requirement, parts can be forged rather than cast. Small parts can be machined, large complex curved parts can be cast, large but simple structures can be welded, and parts that require superior hardness can be manufactured using powder metallurgy tech-

102


niques. For all these cases, the most appropriate manufacturing process or processes must be selected only after careful consideration of all influential factors, including time constraints and the logistics of plant capacity. (6) Methods for obtaining the optimum solution There are two basic kinds of method to obtain desirable process planning results: (1) the variant approach and (2) the generative approach [19]. The generative approach is an ideal method in which manufacturing methods and manufacturing procedures are automatically generated using artificial intelligence and other techniques. Construction of generative approaches applicable to complex practical machine elements is a daunting task at present, but holds promise for the future. The variant approach is a method based on the use of databases containing technological knowledge obtained from experience and GT-based coding methods explained in the foregoing section are used. Variant approaches generally yield several alternatives and optimization techniques are then applied to obtain the best alternative and detailed process planning solutions. 5.1.5.6 Module Technology Module technology is another important technology that aims to accommodate a variety of customer needs and requirements while reducing both product design and manufacturing costs and product development time [20, 21]. When a variety of products must be manufactured, product manufacturing costs often increase dramatically. Module technology seeks to reduce manufacturing costs by assisting the development of product design ideas and efficient manufacturing methods and machinery. That is, the conflicting relationships between the creation of the desired product variety and resulting undesirable increases in manufacturing costs are focused on and resolved as far as possible. Module technology is similar in concept to the GT explained in Sect. 5.1.5.4 above. The following two requirements must be met to benefit from module technology: 1. Commonality of product parts 2. Commonality of manufacturing plant processes and equipment used to make both the parts and the product The first of the above requirements implies that, as far as possible, the same parts will be used in different products, while the second implies that different products and parts will be made in the same manufacturing plant, using the same equipment. In 1 above, common parts are used for different products. The parts or units are designed so as to be useful in a variety of different products, and are called modules or modular parts. Large modules, modules having higher levels of usage commonality, or modules where common parts are assembled into a larger module, are called a platform. Figure 5.19 shows a conceptual diagram of modular design relying on a platform, in which common parts are used in different products. The


103

product parts that customers can directly evaluate, for example the appearance of the exterior and interior of automobiles, differ to accommodate personal preferences. Thus, the purpose of module technology is to make maximum product diversification compatible with reductions in product manufacturing cost.

Fig. 5.19 Conceptual diagram of modular designs based on a single platform

To the conditions of item 2 above, a single manufacturing plant must be employed for both parts production and product assembly. For example, when a manufacturing plant is designed, modular equipment for manufacturing the products is installed on a platform so that a specific but highly variable range of products or parts can be manufactured. When a different range of products or parts must be manufactured, the modular equipment on the platform will be changed accordingly. The important point in modular designs is the decision-making required to implement an appropriate level of flexibility so that modular parts or equipment can be used across a range of products or plants. Although Fig. 5.15 is a conceptual diagram pertaining to GT, this diagram also expresses the conceptual relationship between reductions in manufacturing cost due to modular designs and the possible

104


degradation of product performances. When designing a variety of related products, these relationships must be carefully considered. 5.1.5.7 Scheduling Technology Manufacturing machine products generally requires many processes, and determining how to reduce the time required to complete each process is of fundamental importance when the aim is to reduce product lead time and reduce manufacturing cost. Planning for job shop type manufacturing is complex due to dependencies on factors such as the existence of common parts, the required number of parts and materials and the time taken to acquire and manipulate these, and comprehensive inventory management. If such management and processing is not precisely carried out, the overall operation will likely suffer from excess inventories, supply delays and a lack of required parts, or overlapped orders of common parts and inventory confusion. In order to prevent these troubles, production management information and production technology information must be continuously available, and timely and precisely detailed instructions concerning required items and quantities must be developed and executed. It is practically impossible to conduct such sophisticated activities by merely relying on experience and intuition. Material Requirements Planning (MRP) uses a computer programming system to oversee job operations [22]. After the material requirements plan is finished, production scheduling is determined so that the production time will be minimized. Figure 5.20 shows the flow from the time orders are received from customers, proceeding to planning how to meet material requirements, production scheduling, executing manufacturing, and finally delivering the completed products to customers.

Order Customers

Production scheduling

Accumulation of orders

Execution of manufacturing

Assessment of MRP

Customers

Fig. 5.20 MRP and scheduling

As an example of production scheduling, consider the problem of minimizing the total processing time (makespan) of parts machining operations in an FMC (Flexible Manufacturing Cell) shown in Fig. 2.4, where various machine tools are deployed around an industrial robot used for transporting the parts [23, 24]. The


105

industrial robot transports parts from a conveyor and loads them onto the cell’s machine tools, transporting parts serially from each machine tool to an adjacent buffer station or from a buffer station to the next adjacent machine tool as required, and, finally, to the conveyer outputting finished parts from the cell. The optimum part transporting sequence that minimizes the makespan, from the starting point when the first part on the input conveyer is grasped to the ending time when the last part is placed on the output conveyer, is obtained. Data for the machining and transportation times for each part are shown in Table 5.1. The industrial robot not only performs transportation tasks but also additional operations such as measuring so that parts will be properly positioned when installing them on the machine tools. The times for these additional operations are added to the part transportation times. Here, a flow-shop type of manufacturing system is used, in which each part is machined using the same sequence of operations. Table 5.1 Data for four-part, three-machine tool, industrial-robot-transported flow-shop scheduling problem

Part no. 1 2 3 4

Part no. 1 2 3 4

B1→M1 60 30 45 20

Processing time for machine tool (s) M1 M2 M3 50 55 30 40 60 55 65 40 60 30 70 30 Transportation time for robot (s) M1→B2 B2→M2 M2→B3 B3→M 3 10 50 10 20 10 20 10 40 10 35 10 70 10 45 10 20

M3→B4 10 10 10 10

The best parts transportation sequence was obtained by using the branch-andbound method (explained in Sect. 6.1) as a discrete design variable optimization problem. Figure 5.21 shows the obtained Gantt chart, the time management chart for the sequence of operations. The total processing time is obtained as 685 (s), where the optimum sequence of 24 (4 (the number of parts) × 3 (the number of machines) × 2 (the number of transportation types)=24) transportation operations is determined. The Gantt chart is an effective method for visually displaying such work processes, and was developed by Henry Laurence Gantt around 1903. Such charts are popularly used as a means of sharing information among production workers.

106

5 Product Manufacturing Support Technologies Transportation from conveyor or part storage area to machine Transportation from machine to part storage station or conveyor 1

2 1 1

2 1

2

1

1 22

1

2 1

1

22

1

2 1

221

2

2

22 3

2 2 3 4

3 42

1

3 4

3

14 1

3 4

14 1

1

Robot Part No. 2

3

4

1

Machine 1

2

3

4

1

Machine 2

2

3

4

500

600

1

Machine 3

0

100

200

300 Time (s)

400

700 685

Fig. 5.21 Gantt chart for the optimum solution

Generally, scheduling problems [25 ̶ 27] are mainly managed and solved by process designers, who focus on how to obtain optimum solutions for the kind of problems shown in Table 5.1. However, the data in Table 5.1 could change radically if the designs of parts or products were altered, and even if the processing time has been minimized for a certain process under consideration, there may be unwanted delays in the overall flow of operations, as certain parts wait before the next process can be started. In such cases, if other related processes are concurrently considered, more preferable solutions may be obtained. Thus, in these problems as well, the application of concurrent engineering concepts is vital. Another method for displaying processing schedules is PERT (Program (Project) Evaluation and Review Technique), in which each work flow of the project is drawn using arrow diagrams, and the time required for each operation and work order is described.


107

5.1.6 Technologies to Acquire Information Concerning Customer Needs

5.1.6.1 Collection of Customer Needs Information and Customer Grouping Usually, the primary imperative in product design optimization is to maximize the satisfaction levels of potential customers. Satisfaction levels are generally obtained by integrating evaluations for many product attributes, but an alternative method is to base such evaluations on the customer needs for the product to be designed. Customer needs are usually diverse, so customers are often grouped and product designs most suitable for each particular group are created. Figure 5.22 shows a product optimization method incorporating customer needs. Information concerning customer needs for the products at hand is first gathered by means of questionnaires. Next, the customers are divided into groups based on cluster analysis of the range of customer preferences concerning product attributes. Then, a specific customer group is selected as the target for the product manufacturing. Finally, the details for the product to be designed and manufactured are determined using an optimization method that maximizes the satisfaction level of customers belonging to the selected group.

Survey of customer needs

Grouping of products to be designed based on cluster analysis for desired values of product attributes

Execution of design optimization to maximize satisfaction level of the customers belonging to the selected group Fig. 5.22 Flow of design for maximizing customer satisfaction levels based on market surveys

Using an example of circumstances in which manufacturing enterprises purchase industrial machines, gathering and analyses of data concerning the demand are explained [28]. Here, the following two assessments are conducted to understand the needs of the intended market:

108


1. Assess which product attributes are especially import to customers making purchasing decisions 2. Estimate how many products will be purchased, at what cost, and assess customer importance levels for each attribute selected in 1 Using the collected data, the potential users are divided into groups using cluster analysis in which similarities of user needs are evaluated by measuring the range in desired values for the product attributes used. When n product attributes, the most important factors affecting customer purchasing decisions, are specified, the distance d j, j +1 between j ’s desired value zi, j and user j + 1 ’s desired value zi , j +1 is defined for a product having product attributes i (i = 1, 2,..., n ) as follows: na

d 2j , j +1 = ∑ ai {( zi∗, j − zi∗, j +1 ) / zis }2 i =1

(5.5)

where zis is the standard value of product attribute i , introduced so that product attributes having different units can be evaluated using the same scale. ai is a weighting coefficient for product attribute i where larger values are given for more important product attributes. After group analysis is carried out, a particular product can be targeted for development, since, for example, it has become evident that satisfaction levels for similar products available on the market are low. 5.1.6.2 Technologies to Acquire Customer Needs in Product Designs

The most important evaluative factor for optimization is the satisfaction level of potential customers for the product. A representative procedure for incorporating customer needs in product designs is quality function deployment (QFD), in which the relationships between customer needs and technological factors in product design and development are clarified. The important strategic point of QFD is the concurrent evaluation of customer needs and technological factors. QFD was based on the quality guarantee itemized list proposed in Bridgestone’s Kurume factory and the quality list in Mitsubishi Heavy Industry’s Kobe shipbuilding yard. In 1972, Yoji Akao presented QFD as a quality development system [29]. Over the following years, procedures were further developed, made more sophisticated, and introduced in various industries [30]. The use of QFD spread, especially in Europe and America. Figure 5.23 illustrates a portion of a QFD matrix table for a personal computer product. On the vertical axis, requirements concerning product quality from the point of view of potential customers are listed, while the horizontal axis lists functional quality characteristics for items evaluated by the product maker. The relationships between requirement and characteristic items are visually indicated by


109

◎，○，and △ symbols, where ◎ expresses a very strong relationship, ○ expresses a moderately strong relationship, and △ expresses a weak relationship. Element quality Speed

Quick drawing Quick starting Quick processing

Operability

◎ ○ ◎ △ ◎ ○ △ ◎ ○

◎

4

5.2

5

8.0

3

4.4

Easy alphanumeric input

◎

2

3.2

Easy drawing

◎ ○ △

2

2.6

3

2.8

Ergonomics Importance level of element quality

3.2 8.4 6.6 4.2 4.4

Fig. 5.23 Example of QFD

QFD is a procedure for securing product qualities that help lead to customer satisfaction, carried out by analyzing the relationships between customer needs and product characteristics. When applied to the development of machine products, step-wise development is performed by taking into account product requirement qualities, quality elements, functions, mechanisms, and, finally, parts. At each stage, the relationships between a variety of technological elements and different aspects of the product being developed, such as product requirement qualities and quality elements, or quality elements and functions, are expressed using matrices. These matrix tables make it easier for different divisions of an enterprise to understand and evaluate product knowledge and information. Therefore, QFD is not only effective for integrated decision-making in product development, but also useful for collaborative decision-making among a variety of decision-makers. Figure 5.24 is a conceptual diagram of methodologies for finding the optimum product designs using the hierarchical deployment structures of product requirement qualities, quality elements, functions, and mechanisms.

110


Element quality

Product quality requirement Mechanism

Function

Product quality requirement

Element quality

Function

Mechanism

Design alternatives Fig. 5.24 Conceptual diagram of design methods considering QFD hierarchical structure

5.1.7 Technologies Supporting Enterprise Management Decisions made concerning product manufacturing are also directly related with the management of the enterprise producing the product. Supporting systems that can integrate enterprise activities into the product manufacturing decision-making, such as Enterprise Resource Planning (ERP), are starting to be used. Figure 5.25 shows a diagram of ERP concepts. ERP supports the totality of enterprise opera-

5.2 Utilization of Information Technology for Product Manufacturing

111

tions, from initial material procurement, to product manufacturing that fully utilizes enterprise capabilities and technological and human resources, to finally supplying customers. Integrated enterprise management Assets, Money, Investments Material suppliers

Material procurement

Manufacturing

Providing customers with value Marketing

Customers

Human resources Fig. 5.25 Conceptual diagram of ERP

5.2 Utilization of Information Technology for Product Manufacturing The important technologies for effectively conducting system designs in complicated and large-scale product manufacturing include networked information technology, common databases, and VR technologies that are indicated in the third layer from the center in Fig. 5.1. The use of electronic computers that began in enterprise operations in the 1950s, where data processing was the main task, spread into areas of engineering and production in the 1960s. In the latter half of the 1960s, the concept of a management information system (MIS) was presented, whereby managers could acquire necessary information for enterprise management decision-making in the required form and at the required time. The importance of the concept of information then started to be recognized, where data becomes information when the receiving person finds it useful and of value. At the beginning of the 1970s, the concept of a decision support system (DSS) was presented, in which decisionmaking is supported by a DSS made available to executive officers and managers facing enterprise management problems in the course of operations. The notion that information systems could be useful assets for enterprise management decision-making became popular. And in the latter half of the 1980s and beyond, the concept of a Strategic Information System (SIS) spread, where information systems are utilized for increasingly advanced decision-making tasks. Figure 5.26 shows the flow from data collection, extraction or discovery of information and knowledge, to the final decision-making. The phrase data mining appeared around 1995, and meant that, from a large quantity of accumulated data,

112


particularly useful knowledge and information buried therein could be extracted, much as valuable items are from bulk material in mining operations.

Data collection Extraction or discovery of information and knowledge

Decisionmaking

Fig. 5.26 Decision-making based on data collection

The most important technology required when implementing the concept of concurrent engineering shown in Fig.3.5 is a comprehensive database where data are collected, duplicates eliminated, and adjustments made for increased applicability to the purpose in hand. Especially important is the notion that such databases need to be common, that is, commonly accessible and used by a broad collection of people working cooperatively. Now, and increasingly as time goes on, competitive designers must recognize the needs of a variety of customers more and more quickly, and develop more efficient and effective design activities. In such product development scenarios, product evaluations from different standpoints and angles, and the production of superior products that stand out in the marketplace, are required. Such design problems require the participation of any number of experts in different fields, all united and participating in well-orchestrated design activities. When several experts having different technological fields participate in product designs, the design problems become more complicated and diversified. In such situations, it is expected that related experts will concurrently solve their problems using collaborative tools and methods. Information technology, such as networked information systems, is an indispensable enabler of mutual intercommunication. Collaborations that use networked information systems can more easily unite people having different technologies and knowledge, even if they are geographically scattered. Thus, more preferable decision-making can increasingly be expected. Figure 5.27 shows a conceptual diagram of mutual intercommunication where designers having different expertise, such as a structural designer and a control de-

5.2 Utilization of Information Technology for Product Manufacturing

113

signer, cooperatively conduct product design optimizations that aim to satisfy a customer’s requirements. Here, it is effective to use supporting agent systems so that personnel can be spared the need to conduct time-consuming and complex jobs that require minimal intelligence to carry out.

Fig. 5.27 Conceptual diagram of intercommunication between two expert designers using supportive agent systems

In Figure 5.27, the activities of experts are supported by computerized agent hardware as well as software agents that are transferred across the network to which the experts’ computers are connected. These mobile agents supply a kind of dispersed operation technology in which software programs operate as they move over the network connecting the computers. Collaboration among suitable experts using networked systems is increasingly attractive and efficient, and more preferable design activities employing broader and more precise knowledge can be expected. In such circumstances, further development of advanced supporting systems using techniques such as agents is necessary [31 ̶ 33]. To collect data throughout product lifecycles and extract useful information and knowledge from the collected data, the concept of ubiquitous data and information collection, display, and utilization is now being put into practice. The foregoing descriptions were for technologies pertaining to information networking and common database information systems. An additional type of information system technology is that of VR, where artificially synthesized worlds provide people with sensory inputs that stimulate their imagination and enable a dynamic communication exchange with representations of the designed objects. VR technology is defined as a technology that presents synthesized computer information to human visual, auditory, and sensory organs so that a convincing mental construct recognized as an exterior or artificial world is created [34 ̶ 37]. VR technology can enable people to experience entering into graphic spaces and manipulating constructs that appear realistic, and it also permits them to act upon such virtual worlds or objects within while receiving appropriate sensory feedback,

114


thanks to sophisticated software and hardware systems. Using VR technology offers the following advantages in product manufacturing: 1. Product designing can be conducted from a systematic viewpoint that advantageously uses a range of human senses such as sight, sound, and touch 2. Designs more suitable to the environments in which products will be used can be effectively achieved 3. Manufacturing processes for the product can be understood from systematic viewpoints 4. Designers and customers can experience virtual situations in which the product is used 5. Customer needs and requirements for the product can be effectively investigated 6. Collaboration during product design and development can be facilitated 7. Remote control and management of geographically diversified manufacturing plants can be accomplished VR technology is expected to be used in decision-making scenarios for practical designs, manufacturing, marketing, and product operation by displaying objects for examination and manipulation in virtual spaces as synthesized data and information.

Exercises 5.1 Compare the effectiveness of applying CAE using simplified models and detailed models. Then, discuss how the use of simplified models facilitates obtaining more preferable design solutions. 5.2 Discuss the detailed requirements for the display and arrangement of product ideas in support systems that will be used for group collaboration. 5.3 Discuss the usefulness and value of using a database that contains past knowledge and know-how when developing new product designs. 5.4 Discuss the relationships between sequentially resolving bottleneck processes in product manufacturing processes and obtaining global optimum solutions. 5.5 Discuss methodologies that can achieve decreases in product manufacturing cost without degrading the level of product qualities. 5.6 Describe why, in lifecycle designs considering the reuse and recycling of parts, evaluations for ease of assembly and disassembly are important. 5.7 Discuss the relationship between the requirement to minimize the total processing time of a group of processes and the requirement to minimize just one process of a group of processes. 5.8 Why is it necessary to collect and classify a range of product manufacturing requirements when the aim is to maximize customer satisfaction levels? 5.9 Discuss the potential effectiveness of a system that fully supports enterprise management, such as the one shown in Fig. 5.25.

References

115

5.10 Discuss how the evolution of product shape description technologies is affected by detailed understanding of human physical senses. 5.11 Explain the roles of GT and module design for reducing product manufacturing cost while allowing a wide variety of products to be produced. 5.12 Summarize the similarities and differences between GT and module technology. 5.13 Explain the roles of networked information systems in the operation of concurrent engineering and collaboration for product manufacturing activities.

References 1. Narahara H, Saito K (1996) Shape analysis of solidified photopolymer for the three dimensional photofabrication. International journal of JSPE, 30(4):311 ̶ 316 2. Lemon JR, Tolani SK, Klosterman AL (1980) Integration and implementation of computeraided engineering and related manufacturing capabilities into mechanical product development Process. Gi-Jahrestagung, Saarbrucken, Federal Republic of Germany, October 1 3. Azegami H, Takeuchi K (2006) A smoothing method for shape optimization: traction method using the Robin condition. International journal of computational methods, 3(1):21 ̶ 33 4. Osborn AF (1979) Applied imagination. Charles Scribribner’s Sons, New York 5. Kawakita J (1967) The way of thinking (in Japanese). Chu-Kou Shinsho 6. Kobayashi M, Yoshimura M, Nishiwaki S, Izui K (2003) A method for supporting creative interaction during collaborative design. In: Proceedings of DETC'03 ASME 2003 Design Engineering Technical Conferences and Computers and Information in Engineering Conference, Chicago, Illinois, USA, DETC2003/DAC-48223, September:1 ̶ 10 7. Kobayashi M, Yoshimura M, Nishiwaki S, Izui K (2004) Collaboration support system based on visualization of communication processes. In: Proceedings of DETC’04 ASME 2004 Design Engineering Technical Conferences and Computers and Information in Engineering Conference, DET C2004-57785:1 ̶ 7 8. Terninko J, Zusman A, Zlotin B (1998) Systematic innovation: an introduction to TRIZ (Theory of inventive problem solving). St. Lucie Press 9. Goldratt EM, Cox J (1992) The goal – a process of ongoing improvement. North River Press 10. Taguchi G (1993) Taguchi on robust technology development. ASME Press, New York 11. Parkinson A, Sorensen C, Pourhassan N (1993) A general approach for robust optimal design. Transactions of the ASME, Journal of mechanical design, 115: 74 12. Parkinson A (1995) Robust mechanical design using engineering models. Transactions of the ASME, Journal of mechanical design, 117: 48 13. Doltsinis I, Kang Z (2004) Robust design of structures using optimization methods. Computer methods in applied mechanics and engineering, 193:2221 14. Boothroyd G (1979) Design for economic manufacture. Annals of the CIRP, 28(1):345 ̶ 350 15. Boothroyd G, Dewhurst P, Knight P (1994) Product design for manufacture and assembly. Marcel Dekker 16. Mitrofanov SP (1966) Scientific principles of group technology, part I. National Lending Library of Science and Technology, Boston 17. Yoshimura M, Hitomi K (1986) Application of group technology to design optimization of machine structural systems. Transactions of the ASME, Journal of mechanisms, transmissions, and automation in design, 108(1):3 ̶ 9 18. Yoshimura M, Itani K, Hitomi K (1989) Integrated optimization of machine product design and process design. International journal of production research, 27(8):1241 ̶ 1256

116


19. Chryssolouris G (2006) Manufacturing systems – theory and practice. Springer-Verlag, New York:183 ̶ 217 20. Otto KN, Wood KL (2001) Product design – techniques in reverse engineering and new product development. Prentice Hall, p.370 21. Fujita K, Yoshida H (2004) Product variety optimization simultaneously designing module combination and module attributes. Concurrent engineering – research and applications, 12(2):105 ̶ 118 22. Hitomi, K (supervisor), Nakajima K, Yoshimura M, Yoshida T (eds) (1993), Fundamentals of CIM – Design, Manufacturing, and Management by Computers (in Japanese), KyoritsuShuppan 23. Hitomi K, Yoshimura M (1986) Operations scheduling for work transportation by industrial robots in automated manufacturing systems. Material flow, 3:131 ̶ 139 24. Hitomi K, Yoshimura M, Morimoto (1984) Optimal scheduling for manufacturing systems with industrial robots. In: Computer-integrated manufacturing and robotics published in a bound volume at the ASME Winter Annual Meeting, 11pp., December 25. Papadimitriou CH (1982) Combinatorial optimization: algorithms and complexity. PrenticeHall 26. Baker KR (1974) Introduction to sequencing and scheduling. Wiley, New York 27. French S (1982) Sequencing and Scheduling: an Introduction to the Mathematics of the JobShop. Ellis Horwood 28. Yoshimura M, Takeuchi A (1994) Concurrent optimization of product design and manufacturing based on information of users' needs. International journal of concurrent engineering: research and applications, 2(2):33 ̶ 44 29. Akao Y (1990) QFD – integrating customer requirements into product design. Productivity Press 30. Otto K, Wood K (2001) Product design. Prentice Hall:289 ̶ 300 31. Danesh MR, Jin Y (2001) An agent-based decision network for concurrent engineering design. Concurrent engineering: Research and applications, 9(1):37 ̶ 47 32. Eynard B, Liénard S, Charles S, Odinot A (2005) Web-based collaborative engineering support system: applications in mechanical design and structural analysis. Concurrent engineering 13(2):145 ̶ 153 33. Yoshimura M, Takahashi K (2001) Collaborative design among different fields in mobileagent environments. International journal of concurrent engineering: Research and applications, 9(2):146 ̶ 154 34. Okulicz K (2004) Virtual reality-based approach to manufacturing process planning. International journal of production research, 42(17):3493 ̶ 3504 35. Benhabib B (2003) Manufacturing – design, production, automation, and integration. Marcel Dekker, p.128 36. Ong SK, Yuan ML, Nee AYC (2008) Augmented reality applications in manufacturing: a Survey. International journal of production research, 46(10):2707 ̶ 2742 37. Westkämper E (2007) Strategic development of factories under the influence of emergent technologies. Annals of the CIRP, 56(1):419 ̶ 422

6 Optimization Technologies for Product Manufacturing

The technologies that support product design and manufacturing as explained in Chap. 5 are currently enjoying rapid and vigorous evolution and development but the implementation of certain supporting technologies and the establishment of ideal decision-making under the criteria described in Chap. 2 remain problematic. In future product manufacturing scenarios, rational and efficient decision-making will increasingly depend on the optimization technologies. Suitable optimization methodologies are essential to carry out the best possible decision-making when using a supporting system. This chapter explains the present state of optimization technologies for product designs, and presents fundamental methodologies and strategies that are applicable in simple as well as complex scenarios, including system design optimizations. In the explication of practical product design optimizations, special emphasis is placed upon the need to avoid automatic or formulaic application of optimization techniques, and methods for constructing and formulating successful optimization problems based on careful and thorough understanding of the subjects being regarded are provided.

6.1 Fundamental Optimization Technologies and Difficulties in their Application In machine product designs, characteristics such as manufacturing costs, operational accuracies, operational efficiencies, and energy required for operation are evaluated. As explained in Sect. 2.2, these characteristics can be roughly categorized into two groups: characteristics whose values need to be improved, and characteristics for which specific constraints are defined. To obtain rationally design solutions that satisfy such requirements, optimization problems are formulated as explained in Sect. 2.3. Optimization techniques such as mathematical programming methods and genetic algorithms (GAs) are being vigorously developed

118


and obtaining solutions for a wide variety of optimization problems is becoming easier, but confidence that the obtained solution is a truly global one often remains elusive. Furthermore, to increase the efficiency of optimization processes for complex optimization problems, methods for obtaining practical solutions by using the response surface method based on the design of experiments are being actively studied and employed, methods in which the characteristic response surface for design variable vectors can be obtained using a minimum number of sampling points. Many instructive textbooks for design optimization methods have been published by Johnson [1], Wilde [2], Haug and Arora [3], Siddall [4], Vanderplaats [5], Haftka and Kamat [6], Papalambros and Wilde [7], Arora [8], and Mastinu, Gobbi, and Miano [9]. In this chapter, optimization technologies pertaining to product manufacturing are discussed.

6.1.1 Linear Programming Problems Optimization problems in which all of the objective and constraint functions are expressed as linear functions of design variables are called linear programming problems. When a feasible region exists in the design variable space in this type of problem, there is only one local optimum solution, which can be easily obtained by the simplex method, or other means. In economics, profit and cost are often expressed as linear functions, and linear programming methods are often used, but the optimization problems used in product design and manufacturing can seldom be expressed as linear programming problems. However, when solving nonlinear programming problems by using linear approximations, linear programming methods are often employed.

6.1.2 Nonlinear Programming Problems and Local Optimum Solutions Optimization problems in which one or more of the objective and constraint functions are expressed as nonlinear functions of design variables are called nonlinear programming problems. In nonlinear programming problems, numerous local optimum solutions may exist in the feasible design variable space. Nonlinear programming methods usually first set initial design variables and then, using these, search for design solutions that minimize or maximize the objective function. The solutions obtained by such processes are locally optimum solutions, and it is often very difficult to be sure that a particular solution is in fact the globally optimum solution.

6.1 Fundamental Optimization Technologies and Difficulties in their Application

119

Figure 6.1 shows a design variable space of two design variables d1 and d 2 where the contour lines of the objective function to be minimized and the boundary lines of the constraint conditions are drawn. When there are no constraints, it can be seen from the contour lines of the objective function that local optimum solutions exist at points A, B, C, and D. The feasible region defined by constraint g1 ≤ 0 is located in the lower right area from the line where g1 = 0 , while the feasible region defined by constraint g 2 ≤ 0 is located in the upper right area from the line where g 2 = 0 . The feasible design variable space is shown as a shaded area. The local optimum solution near point A is located at point A' on the boundary line of constraint g1 ≤ 0 , while the local optimum solution near point C is located at point C' on the boundary line of constraint g 2 ≤ 0 . Hence, four local optimum solutions exist, at points A', B, C', and D in the feasible design variable space. The solution having the smallest objective function value among the four local optimum solutions corresponds to the globally optimum solution.

g1=0 A A′ S1

S3

B

C′ S2

D

C g2=0 Design variable　d1 Fig. 6.1 Design variable space in a nonlinear optimization problem, and local optimum solutions

Various methods can be used as searching methods in nonlinear programming problems. In the Sequential Linear Programming (SLP) method, nonlinear optimization problems are sequentially approximated to linear optimization problems at each solution searching step, while in the Sequential Quadratic Programming (SQP) method, nonlinear optimization problems are sequentially approximated to quadratic optimization problems at each solution searching step. In the penalty

120


function method, optimization problems having constraints are transformed into optimization problems free of constraints by adding the constraint terms to the objective function as a penalty term. The solution is then sequentially searched for by altering the response parameter. One sequential optimization algorithm based on the penalty function method is called SUMT (Sequential Unconstrained Minimization Technique). In the foregoing optimization methods, first, an initial design point is set in the design variable space. When the initial design point is point S1 as shown in Fig. 6.1, a local optimum solution at point A' is obtained. When the initial design points are points S2 and S3, local optimum solutions at points C' and B are respectively obtained. Since in complicated practical optimization problems the features of the design variable space are unknown during optimization, it is difficult to judge whether or not the solution obtained by the optimization method is the globally optimum solution. The nonlinear optimization problem having objective function f and inequality constraint functions g j ≤ 0 ( j = 1,2,..., m) are formulated as explained in Chap. 2 as follows: f (d ) →

minimize or maximize

g j (d ) ≤ 0,

j = 1,2,..., m

(6.1)

(6.2)

Lagrangian function L is expressed using Lagrangian multipliers μ j as follows: m

L = f + ∑ μj gj

(6.3)

j =1

To evaluate whether or not a given design point d* is a local optimum solution in the optimization problem, the following Karush-Kuhn-Tucker condition (KKT condition) [10, 11], a necessary condition for local optimality, must be satisfied: ∂L ∂d

*

=

∂f ∂d

*

m

∂g j

j =1

∂d*

+ ∑ μj

=0

(6.4)

∂L ≤0, ∂μ j

j = 1,2,..., m

(6.5)

μj ≥ 0,

j = 1,2,..., m

(6.6)


μ Tj g j ,

j = 1,2,..., m

121

(6.7)

The first term in the right-hand side of (6.4) is the gradient of the objective function f at design point d* , while the second term is the summation of the constraint function gradients at design point d* , each of which has a Lagrangian multiplier coefficient. Each Lagrangian multiplier μ j is non-negative, and the values are classified into two categories: μ j > 0 and μ j = 0 . When μ j > 0 , constraint function g j must satisfy g j = 0 from (6.7). At design point d* , the constraint condition becomes active. Various situations exist depending on which constraints are active, but when there are no active constraints and KKT conditions are satisfied, design point d* is considered a local optimum solution inside the feasible region. When there is only one constraint condition, Lagrangian function L is expressed as follows: L= f +μ g

(6.8)

As a structural optimization example, when the static rigidity k s of a machine structure is the objective function, and the constraint function concerning the structural weight W is the constraint function g , the optimization problem is formulated as follows: k s →maximize

(6.9)

g = W −WU ≤ 0

(6.10)

The foregoing objective can be reformulated as a minimization problem as follows: −k s →minimize

(6.11)

Lagrangian function L is then expressed as L = − k s + μW

(6.12)

From (6.4) for the KKT condition, the following condition must be satisfied for design point d* to be considered a local optimum solution:

122


−

∂k s

*

∂d

+μ

∂W ∂d*

=0

(6.13)

Since the static rigidity k s behaves monotonously with respect to the design variables,

∂k s

∂d*

always has a positive value. That is, the value of k s increases as

design variable values increase, and decreasing design variable values cause the value of k s to decrease. The structural weight W enjoys a similar monotonous re∂W

lationship with the variables, so

∂d*

always has a positive value.

As explained above, the value of Lagrangian multiplier μ must be greater than or equal to zero for design point d* to be a local optimum solution. From the foregoing considerations for k s and W , when Lagrangian multiplier μ is greater than zero ( μ > 0 ), g = W − W U = 0 , from (6.7). A local optimum solution can then be obtained using the product’s weight as an active constraint. When μ = 0 , a local optimum solution does not exist since (6.13) is not satisfied. In this example of an optimization problem where static rigidity is included in the objective function and the structural weight is included as a constraint function, a single local optimum solution exists, and thus it can be considered to be the global optimum solution. This would also hold if the structural weight were the objective function and the static rigidity the constraint function, since both characteristics behave monotonously with respect to design variables. As an additional structural optimization example, when the dynamic rigidity k d of a machine structure is the objective function, and the constraint function including the structural weight W is g , the optimization problem is formulated as follows: kd → maximize

(6.14)

g = W −WU ≤ 0

(6.15)

From (6.4) for the KKT condition, the following condition must be satisfied for design point d* to be considered a local optimum solution: −

∂kd

∂d*

+μ

∂W ∂d*

=0

(6.16)

When the value of Lagrangian multiplier is greater than zero ( μ > 0 ), a local optimum solution exists, analogous to the case for the static rigidity problem ex-


plained above. Even if the value of Lagrangian multiplier μ is zero ( μ = 0 ),

123

∂kd

∂d*

can be equal to zero. Therefore, local optimum solutions can exist inside the feasible region and local optimum solutions may exist inside the feasible region besides the local optimum solution pertaining to the active constraint condition [12]. This is due to the non-monotonous relationship of the dynamic rigidity k d with respect to the design variables.

6.1.3 Multiobjective Optimization Problems As explained in Sect. 2.3, designers must usually deal with numerous requirements when formulating practical problems as optimization problems. In such cases, the optimization problems are expressed as multiobjective optimization problems having a number of objective functions [13]. Pareto optimum solution set

A

Feasible region

C D

Contour lines of the weighted sum f0 of objective functions

0

E F

I

B Ideal point

ε

f1

Fig. 6.2 Multiobjective function space and Pareto optimum solution set

Figure 6.2 shows a multiobjective function space having two objective functions, f1 and f 2 , each of which includes the same design variables. When design solution point A, where objective function f1 is minimized, does not coincide with solution point B, where objective function f 2 is minimized, the set of opti-

124


mum solutions for the multiobjective optimization problem exist on the thick solid line connecting point A and point B. When at each solution on line AB an improvement in one objective function brings about a degradation in the other objective function, the solution set on line AB is called a Pareto optimum solution. If points A and B are coincident, the optimization of one objective function will bring about the optimization of the other objective function and formulation, as a multiobjective optimization problem is not necessary in such cases. Point I in Fig. 6.2, composed of the minimum value for each objective function, is called the ideal or Utopia point, but this particular solution is infeasible, since it lies outside the feasible design region. The most fundamental procedure in dealing with multiobjective optimization problems is first to obtain the Pareto optimum solution set and next to evaluate quantitatively the conflicting relationships among the objective functions. There are various methods for obtaining Pareto optimum solution sets, as follows. 6.1.3.1 Weighted Sum Method (Weighting Method) In this method, a weighting coefficient ω ( 0 ≤ ω ≤ 1 , ω ≥ 0 ) is applied to objective functions f1 and f 2 as in the following equation, and the result is expressed as f0 : f 0 = ωf1 + (1 − ω ) f 2

(6.17)

In the multiobjective function space shown in Fig. 6.2, f 0 is illustrated by a dashed straight line for a specific value of ω . The solution having the minimum value of f 0 within the feasible region corresponds to one solution of the Pareto optimum solution set. By gradually and incrementally changing ω from 0 to 1 and obtaining the solution that minimizes f 0 , the Pareto optimum solution set (line) between points A and B can be obtained. In this method, particular attention must be given to the units of each objective function, since these may differ, such as when one unit expresses weight while another expresses cost, the result being that a summation of such objective functions would be illogical. In such cases, it is necessary to normalize each objective function so that the output values are non-dimensional, and various methods can accomplish this. In a representative method, objective function f i is reformulated as f i ′ using the objective function value at the ideal point, f i ′ , and the maximum value of the objective function, f imax , as follows:


f i' =

f i − f iI f imax − f iI

125

(6.18)

When the Pareto optimum solution line has a concavity in the inner direction of the feasible region, as shown in Fig. 6.2, Pareto optimum solutions corresponding to the concave area cannot be obtained by changing the value of weighting coefficients, but such situations are rare in practical product design problems. When there are regions where Pareto optimum solutions cannot be obtained during the practical application of the weighted sum method, such regions can be given special attention and solutions can be obtained by using, for example, the ε-constraint method, as explained below. 6.1.3.2 Epsilon( ε )-Constraint Method / Bounded Objective Function Method In this method, among the objective functions, one objective function is specified as the operating objective function, and the other objective functions are then included in the constraints of the optimization formulation. The upper limit of each constraint function is set as ε . By incrementally changing the magnitude of ε during serial optimizations, useful Pareto optimum solutions can be obtained in a stepwise manner. As an example, for a two objective function optimization problem, the optimization formulation expressed when f1 is included in the constraints is as follows: f 2 → minimize

(6.19)

f1 ≤ ε

(6.20)

also subject to the constraints included in the original optimization problem. In Fig. 6.2, the Pareto optimum solution at point E is obtained by the foregoing optimization for the value of ε on the f1 axis. The optimization is then iterated by incrementing the magnitude of ε by a small value Δε , starting with the f1 value at point A and continuing through the f1 value at point B, to obtain a Pareto optimum solution set from point A through point B. Even if the Pareto optimum solution line has a concave shape between points D and F, as shown in Fig. 6.2, Pareto optimum solutions can also be obtained for the extent of the concave segment. In multiobjective optimization problems having three or more objective functions, the evaluative functions other than the evaluative function selected as the operating objective function are included in the constraints, as above, and dif-

126


ferent values of ε are defined for each constraint. The Pareto optimum solution set can then be obtained by incrementing the various defined values of ε .

6.1.4 Optimization Problems Including Discrete Variables Since the outlines of certain designs are created prior to the detailed design stage of the optimization process, the dimensions of shapes are often determined at the detailed design stage. In such cases, design variables are usually expressed as continuous design variables, and conventional linear or non-linear programming methods can be used. However, when discrete design variables such as kinds of materials, machining methods, and manufacturing methods are included in the design variables, the optimization problems then become discrete design variable optimization problems (also called combinatorial optimization problems) [14 ̶ 17]. The so-called enumeration method operates by obtaining optimum solutions for each possible combination of the included discrete variables, using a nonlinear mathematical programming method, and then the best solution is selected from among the obtained solutions, a process that has high computational requirements. Figure 6.3 shows a conceptual feature where feasible regions for eight kinds of materials are discretely distributed in a two-dimensional function space representing the structural rigidity and the structural weight. To obtain optimum solutions for such discrete problems as efficiently as possible, the branch-and-bound method is often used.

0 Fig. 6.3 Discrete feasible regions in a two-characteristic function space


127

The branch-and-bound method aims to eliminate unnecessary calculations as much as possible during branching operations by decomposing the feasible solution set into sub-sets and using bounding operations to calculate upper or lower bound values of the criterion for each sub-set. Bounding operations are important. Consider the following conditions: 1. If the optimum solution for a sub-set is obtained, no further branching operations for the sub-set is required 2. If a given sub-set and all lower sub-sets generated from this sub-set are determined to be incapable of providing the optimum solution for the original optimization problem, no further branching from this sub-set is required When either of these conditions is satisfied at a branching node, the node is considered to be terminated. Thus, evaluation criteria must be set so that the status of these conditions can be determined. Furthermore, a sub-set that is not yet terminated is said to be active. When no active sub-set problems remain, the solution searching processes end and the best solution among the candidate optimum solutions that have been obtained thus far becomes the optimum solution for the original problem. In the explanation concerning scheduling technology presented in Sect. 5.1.5.7, an example of operation scheduling in the FMC system was used. The problem is to determine the optimum workpiece transporting sequence that an industrial robot will perform. In the example, an outline of the branch-and-bound procedures will be explained using Fig. 6.4 [18]. In the manufacturing cell shown in Fig. 2.4, three machine tools are installed where machining operations are conducted for four kinds of parts, and an industrial robot located at the center of the cell is used to transport the parts. At the top level shown in Fig. 6.4, there is the set S0 that includes all feasible solutions. This set is hierarchically broken down into sub-sets. First, since there are four types of part that are transported to machine tool 1, four sub-sets containing these parts are arranged as S1 , S 2 , S 3 , and S4 . Parts are transported serially to machine tool 1 from the sub-sets S1 , S 2 , S3 , and S4 . In Fig. 6.4, the location of each branch is a node and at each node i , the lower bound LBi is obtained, where a makespan time below the lower bound value cannot be achieved for any selection of the remaining transportation sequences. Since the sub-set having the minimum value among LB1, LB2 , LB3, and LB4 is considered to be most promising for further searches, the next branching is conducted at this sub-set. Now, when the lower bound value of sub-set S 3 is the lowest among S1, S2 , S3 , and S4 , further branching sub-sets are expressed as S5, S6 , and S7 . Such branching procedures are repeated where the priority for branching order is given to the node having the minimum lower bound among all active nodes at the present searching stage.

128


Set of all feasible solutions S0

S2

S1 LB1

LB3

LB2 LB5

S5

Sm+1

S3 S6

S4

LB4

S7

LB6

LB7

Sm+2

S8

S9

LBm+2 LB8

LBm+1

LB9

Sn

Sm

LBn

LBm

Fig. 6.4 Conceptual diagram of the branch-and-bound method

When a complete sequence solution for part transportation operations is obtained at node m after repeating such branching operations, the sequence solution becomes a temporary optimum solution. The lower bound value LBm at the node is then the objective function value of the makespan. Among the active nodes, the nodes having values equal to or greater than LBm become terminated nodes for which further branching operations are unnecessary. Then, among the active nodes, the branching operation is restarted for the node having the minimum lower bound value. When node n has a new temporary optimum solution and there are no nodes with a lower bound smaller than LBn , the solution at node n becomes the optimum solution for the problem being considered. A key point in the branch-and-bound method is how to formulate lower bounds. The methods for conducting branching and bound operations are not unique, but the constructions must be tailored to the specific problems being considered. In comparison with the enumeration method, where the optimum solution is obtained by selecting the best solution from among all possible combinations of discrete design variables, the branch-and-bound method can effectively avoid unnecessary solution searching processes. If logically appropriate formulations are constructed, the obtained solution can be trusted to be a globally optimum solution. Despite these benefits, significant reductions in computation time cannot be expected for problems that include a large number of discrete design variables, since computation time increases exponentially with the number of discrete design variable


129

combinations. Obtaining an optimum solution in a practical period of computation may become difficult in such cases, and GAs, which will be explained in the next section, are often used instead, as a heuristic optimization method. Nevertheless, the branch-and-bound method is indispensable for evaluating the validity of solutions obtained by a heuristic method.

6.1.5 Genetic Algorithms (GAs) GAs were presented toward the end of the 1960s and beginning of the 1970s by Holland [19], as artificial models that imitate the evolutionary mechanisms of creatures in nature during adaptation processes. Goldberg then developed these artificial models as procedures that can be used during optimizations [20]. When using GAs, design variables are expressed as genetic types that are described by the genetic types of binary numbers, as shown in Fig. 6.5. Usual design variables are termed phenotypes and the mapping from phenotypes to genetic types is called coding. As creatures in nature evolve, individuals having greater fitness for their environment among the set of individuals forming a specific generation reproduce with a higher than average probability of survival. Furthermore, new individuals at the next generation are formed by crossover and mutation processes. Parent 1

1001011011

Child 1

111 0 011 0 11

Parent 2

111 0 01 0 11 0

Child 2

10 0 1 01 0 11 0

Fig. 6.5 Explanatory diagram of genetic crossover

Figure 6.5 shows a diagram of a crossover operation, where “ | ” represents a break symbol on the chromosome of each individual. When the two kinds of individual for design variable b1 correspond to parent 1 and parent 2, child 1 and child 2 are obtained by exchanging the part to the left of the break symbol for each individual chromosome. As a simple example of a mutation operation, Fig. 6.6 shows a mutation-like change, selected with a certain probability. Here, “1” is transformed to “0” at the gene position indicated by an arrow. Position of mutation gene

1001011011 Fig. 6.6 Explanatory diagram of genetic mutation

Transformation from 1 to 0

1001010011

130


As explained above, GAs are used in an optimization method as tools that assist the search for optimal solutions, mimicking the evolutionary processes of living creatures. In applicable GAs here, multiple point searching procedures are used in preference to the single point searching procedures normally used in conventional mathematical programming methods. In GAs using probabilistic operators, multiple points are simultaneously searched for, independently of deterministic rules, and this means that the obtained solution cannot be said to be the global optimum solution. As explained in Sects. 3.1 and 3.2, when optimum design solutions are selected from wider feasibilities, that is, from the conceptual and fundamental design stages, discrete design variables are often included in the design variables of the optimization formulation. However, effectively finding solutions to practical optimization problems that incorporate discrete design variables is a difficult, but very important, problem in product design and manufacturing [21].

6.1.6 Large Scale Optimization Problems As described in Chap. 3, the search for more preferable solutions in product manufacturing requires the implementation of concurrent engineering and collaboration concepts. When limited variety mass production was the norm, the primary evaluative criterion was to minimize manufacturing cost while preserving the required product performances and qualities. That is, the objective function was the product manufacturing cost. After job shop type production became popular, as the product manufacturing paradigm evolved to cope with a wider variety of customer preferences, a broader range of products incorporating various combinations of product performance and manufacturing cost were manufactured. In this scenario, multiobjective optimization must be applied so that both product performance and product manufacturing cost can be included in the objective function. Furthermore, concurrent engineering concepts are fundamental to successfully accomplishing further reductions in product manufacturing costs, additional improvements in product performances and qualities, and necessary reductions in product development periods, since such concepts facilitate the optimization of decision-making factors, at the early stages of product development, that determine how effectively the product manufacturing process can be carried out from start to finish. Here, integration of optimizations at different divisions, stages, and areas is mandatory. Recently, increasing emphasis has been placed on including environmental factors and concerns about dwindling natural resources in product designs, while also considering product lifecycles and mental factors of potential customers, such as aesthetic factors that increase satisfaction levels. Optimization problems in such circumstances naturally increase in size and complexity.

6.2 Fundamental Strategies for Effectively Applying Optimization Methods

131

Design optimization procedures in which the formulation of optimization problems occurs without deep consideration of the problem’s complex requirements, or where conventional optimization techniques are conducted mechanically, usually fail to provide useful results for practical product design problems that include the range of features explained in the foregoing Sects. 6.1.1 ̶ 6.1.6.

6.2 Fundamental Strategies for Effectively Applying Optimization Methods Figure 6.7 shows the flow of a product design optimization using CAD and CAE techniques, from the conceptual design stage. The most important points in the optimization are how to formulate the design problem using all knowledge and intelligence concerning the subjects being examined, and how to evaluate the obtained solutions. The optimization methods support logically obtaining the best solutions for the problems being considered. To make the optimization effective, the following points must be implemented: 1. Support decision-making of designs from the conceptual design stage 2. Understand and evaluate how the global optimum solution was obtained 3. Judge the validity of the obtained optimum solution 4. Support the generation of additional ideas to obtain more preferable solutions

Designer’s idea

Rough sketch

Conceptual design using 3-D CAD

Detailed design using 3-D CAD

Product assembly model using 3-D CAD

CAE using simplified model FOA and Simplified model analysis

Design optimization

CAE using detailed model

Design optimization

Fig. 6.7 Flow of a product design optimization using CAD and CAE techniques

Factor 1 above implies a requirement for simplified or idealized models for the design subjects, which will be used during the conceptual design stage [22]. Furthermore, as explained in Sect. 3.2, simplification or idealization of the optimization subject is indispensable for (1) obtaining the global optimum solution in practical design problems that include many local optimum solutions, and (2)

132


extracting the most preferable design solutions from the abstracted design subject that contains many alternative design ideas. The following are examples of simplification: 1. The value of a specific characteristic is fixed as a constant value that is approximately obtained from experience or prior experiment 2. A number of design variables are transformed into a single or a smaller number of design variables 3. A multi-degree-of freedom system is approximated as a single-degree-of freedom system 4. A complicated characteristic is expressed as combination of several simpler characteristics 5. Characteristic factors that only exercise minor effects are disregarded 6. When two characteristics have a dominant/subordinate relationship, the dominant characteristic is determined before determining the subordinate characteristic 7. A practical model is transformed into an equivalent simplified model 8. Structural members are modeled using simplified elements such as beam elements Table 6.1 shows a methodology where the optimization flow of Fig. 3.2a based on simplification is applied to the optimization of machine structures [23, 24]. In the first phase, Simplification, mathematical models or simulation models having structural characteristics equivalent to practical structures are constructed. The left side of Fig. 6.8 shows an example of the structural cross-section in column members of a practical machine tool, while the right side of the same figure shows the equivalent simplified box-type member having equivalent cross-sectional characteristics. When the model on the left is optimized without simplification, there is an excessive number of design variables, in addition to the exterior dimensions of members, such as the dimensions of internal ribs and partitions in member interiors. In the simplified model on the right, there are only four design variables, consisting of the two cross-sectional widths of the structural members and two plate thicknesses. Furthermore, when optimization is conducted for complex models such as shown on the left side of Fig. 6.8, the obtained optimized results typically preserve the original shapes while only changing dimensions, so the degree of expected design improvement is less than ideal. Using simplified shapes for the optimization of structural members makes it easier to obtain optimal results for the entire structure, and more preferable detailed shapes can later be obtained during the implementation phase. Simplifications of structural members such as shown on the right side of Fig. 6.8 are also conducted for structural joints. The complete simplified machine structural model can then be constructed.

6.2 Fundamental Strategies for Effectively Applying Optimization Methods

133

Table 6.1 Design optimization based on flow of simplification, optimization, and implementation

Phase 1

Phase 2

Phase 3

Simplification

Optimization

Implementation

Step １

Division of the complete machine structure into structural members and joints

Step

2

Simplification of structural members

Step

3

Simplification of joints

Step

4

Construction of the complete machine structural model

Step 5

Optimization of the complete machine structural model

Step 6

Sensitivity analyses of design variables

Step 7

Implementation of structural members

Step 8

Implementation of joints

Step 9

Synthesis of the entire machine structure and evaluation

2580mm

Equivalent simplified box-type cross-section

3 2

0

T1 3(z)

T1

0

2(y)

Α=0.257m2, J=0.116m4 l2=0.0595m4, l3=0.240m4 2580mm

B1

3 0

2

Α=0.318m2, J=0.164m4 l2=0.0861m4, l3=0.259m4

Fig. 6.8 Simplification example for cross-sectional shapes

134


In the second Optimization phase, shown in Table 6.1, optimization is conducted for the complete structural model. Since the number of design variables for the optimization model has been greatly reduced from the original number of design variables included in the practical design, optimization using mathematical programming methods is streamlined and the chances of convergence to undesirable local optimum solutions are minimized. Finally, in the third phase, Implementation, detailed shapes and dimensions for practical machine structures are determined for all of the sub-structures. At this time, additional design and functional requirements are also included in the implementation procedures. The term simplification includes the process of breaking down the problem being regarded into smaller simplified problems, and transforming practical characteristics into simpler characteristics. Wherever possible, the actual optimization procedures are applied to hierarchical structures and groups containing simplified sub-problems. After this type of transformation of the initial optimization problem, sophisticated optimization can be conducted and solutions to the global optimization problem are then obtained step-by-step.

6.3 Fundamental System Optimization Approaches Optimization methods for large-scale optimization problems having many kinds of evaluative factors are categorized into two groups: (1) methods where numerous characteristics and tasks are sequentially organized and processed based on priority and associated interrelationships, and (2), methods where all related characteristics are fully optimized simultaneously in one operation.

6.3.1 Decision-making Sequence Applied to Task Operations and Optimization of Evaluative Characteristics Each item subject to decision-making often has interrelationships with other items. In such cases, these interrelated items cannot be independently determined. Priority relationships among the items to be decided must then be considered. Steward proposed the DSS (Design Structure System) method that gives the optimum sequence when dealing with a group of tasks having complicated priority relationships, to support the construction of optimum planning for large scale design problems [25]. Later, the concept and method were advanced and renamed the DSM (Design Structure Matrix) method. In the DSS and DSM methods, tasks and evaluative characteristics are sequentially ordered based on priority relationships [26]. The flow of the DSM method [27] is as follows:

6.3 Fundamental System Optimization Approaches

135

1. Tasks tk ( k = 1,2,..., r ) are arranged in order in both the extreme left column and the top row of a matrix. When a task located in the top row has to be executed prior to executing a task located in the left column, a “+” symbol is placed in the corresponding position to clarify the precedence relationships among the tasks. Table 6.2 shows the processing order of the ten tasks, t1 , t2 , …, and t10. For example, tasks t1 and t9 must be completed before executing task t3 . Table 6.3 shows the precedence relation matrix. Table 6.2 Order of tasks for processing

Task

Required preceding tasks

t1 t2 t3 t4 t5 t6 t7 t8 t9

t3 t1, t2, t1, t4, t1 t2, t1

t10

t3, t5, t7

None

t9 t8 t7 t8 t5, t9

Table 6.3 Original precedence relation matrix

i

j

t1

t2

t3

t4

t5

t6

t7

t8

t9

t1

+

t2 t3

+

t4 t5

+

+

t7

+

t8 t9 t10

+

+

+

t6

+

+ +

+

+

+

+ +

+

t10

136


2. Rearrangements of matrix columns or rows are then conducted so that the “+” symbols wind up located in the subdiagonal area. Table 6.4 shows the matrix after this rearrangement. When the “+” symbols are located in the subdiagonal area, the reordered tasks in the left column, from top to bottom, represent the most efficient processing sequence. Table 6.4 Precedence relation matrix after rearrangement j

i

t1

t7

t5

t9

t3

t10

t2

t8

t4

+ +

+

t6

t1 t7 t5 t9 t3 t10

+ + + +

+ +

+

+

+ +

t2

+

t8

+

t4 t6

+ +

Table 6.5 shows another example of task precedence relations, and the precedence relation matrix is illustrated in Table 6.6. In a similar fashion, by rearranging columns and rows, the results shown in Table 6.7 are obtained, where three independent blocks, A, B, and C, exist. None of the tasks in these blocks has any precedence relation with tasks in other blocks, so each block’s tasks can be independently processed at the same time, leading to significant improvements in processing efficiency. Table 6.5 Task precedence relations

Task

Required preceding tasks

t1 t2 t3 t4 t5 t6 t7 t8 t9

None

t10

t1, t5, t7

t6, t4, t8 t7, t2, t1 t4

t9 t8 t10 t9

None


137

Table 6.6 Original precedence relation matrix

j

i

t1

t2

t3

t4

t5

t6

t7

t8

t9

t10

t1

+

t2

+

t3 t4

+

t5 t6

+

t7

+

+ +

+

+

+

+

t8 t9 t10

+

+

+

Table 6.7 Precedence relation matrix after rearrangement

j

i

t1

t7

t10

t5

t4

t8

t3

t9

t6

t2

t1 t7 t10 t5 t4 t8 t3

+ +

+ +

+

Block A

+

Block B

+ +

+ Block C

+

t9 t6 t2

+ +

+

+

In block A, tasks t10 and t5 are interrelated, and consequently need to be processed concurrently. Block A can be processed as shown in Fig. 6.9a. Similarly, in block B, tasks t 4 and t8 are interrelated, and so require concurrent processing. Block B can be processed as shown in Fig. 6.9b.

138


t4

t1 t7 t10

t8 t8

t5

(a) Processing flow for block A

(b) Processing flow for block B

Fig. 6.9 Flow of task processing

6.3.2 Two Stage Integrated Optimization

6.3.2.1 Fundamental Methodologies of Two Stage Optimization A system optimization problem where particular jobs are assigned to various divisions or operational units becomes a decision-making problem in which each subsystem problem, which contains one or more jobs, requires optimization. The optimization problem is expressed as a two-stage optimization in which the upper stage corresponds to the system stage where adjustment of the entire system is conducted, while the lower stage corresponds to the sub-system problems where optimizations are conducted individually. In the most fundamental optimization, a system is divided into n sub-systems, and design variable vector d0 pertaining to the whole system, and a design variable vector di pertaining to each sub-system i (i = 1,2, ..., n ) , are defined. d0 is called an adjustment variable. The optimization procedures are as follows: Step 1. In each sub-system i (i = 1,2, ..., n ) , optimization for design variable vector di is conducted using given values of adjustment variable vector d0 . Step 2. Optimization for adjustment variable vector d0 is conducted under the condition that design variable vectors di are fixed to the values obtained in Step 1.


139

Step 3. Solution convergences are evaluated. When the solutions are judged to be convergent, the design variable solutions are considered to be the optimum solutions. If convergence is not obtained, the procedure returns to Step 1. To classify design variables into system level design variables and sub-system design variables, a relational matrix between evaluative characteristics and design variables is often used [28, 29]. Table 6.8 shows an example that expresses whether or not 3 objective functions, f1, f 2 , and f 3 , and 12 constraint functions, g1, g 2 , …, and g12 , have functional relationships with each of 7 design variables, d1 , d 2 ,…, and d 7 . The # symbol shows that a relationship exists. Table 6.8 Original relational matrix between evaluative characteristics and design variables

j

i

f1 f2 f3 g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 g11 g12

d1

# # #

#

d2

# #

d3

# #

# # #

#

d5

d6

d7

#

# #

# #

#

# #

# # #

#

#

#

# #

#

d4

# #

# # #

By adjusting the matrix, the results shown in Table 6.9 are obtained. The design variables are classified into three groups from which two-stage optimization procedures are constructed. Optimization for sub-system 1 at the lower stage uses objective function f 3 , constraint functions g1 , g2 , g6 , g11 , and g12 , and design variables d1 and d 2 , while optimization for sub-system 2 at the lower stage has objective functions f1 and f 2 , constraint functions g3 , g 4 , g5 , g7 , g8 , g9 , and g10 , and design variables d 3 , d 4 , d 5 , and d 6 . For system optimization at the highest level, design variable d 7 is determined for two objective functions, f 2 and f 3 , and the related constraints.

140


Table 6.9 Relational matrix between evaluation characteristics and design variables after the rearrangement i

j

f3 g1 g2 g6 g11 g12 f1 f2 g3 g4 g5 g7 g8 g9 g10

d1

d2

# # # # #

# # # #

Optimization of sub-system 1

d3

d4

d5

d6

Optimization of system level Optimization of sub-system 2

# # # #

# # # #

# #

# #

d7

#

# #

#

# # #

#

#

#

#

# #

6.3.2.2 Methods for Maximizing System Level Goals A large practical enterprise system is usually composed of various sub-systems such as components, divisions, and expert fields. The objective purpose of such a system is to minimize the difference between goal values and practical values that can actually be achieved, considering the contributions of all the various subsystems. Optimization procedures that can be applied to these large systems have been actively studied [30], and several of the proposed methods are described below. In the Bi-Level Integrated System Synthesis method developed by Sobieski et al., optimization of each field broken down from a system is conducted so that the goal value of the entire system is maximized [31]. Two-level optimizations are executed, alternately repeating optimization of lower level sub-systems and optimization at the system level. The Collaborative Optimization (CO) method was developed by Braun et al. for solving large-scale design problems in aerospace fields, particularly among multi-disciplinary engineering divisions [32]. For example, structural and hydrodynamic analyses are separately conducted in their corresponding divisions. Since design and analysis are conducted in different divisions, such data must be reconciled and systematically optimized later. In this method, two-level optimizations consist of an upper level containing the product design goal, and a lower level containing analysis models created by various divisions. At the upper level, the objective function for the entire system is minimized or maximized to achieve the overall goal of the design problem, while consistently satisfying the constraints of several common variables located in the lower level. During lower level optimiza-


141

tions, the lower level design variables approach the design goals of variable values set in the upper level. Optimization procedures at the upper and lower levels are alternately repeated to obtain optimum solutions for the product design as a whole. In the ATC (Analytical Target Cascading) method [33], design variable values used to achieve the product design goal are first obtained at the highest level and these values are then sequentially transferred to lower hierarchical level design problems as goal values, according to the particular hierarchical structure of the problem. The designs variables at the lower levels are optimized so that their goal values will be achieved. The ATC method implements the following practical procedures: Step 1. The total system design problem is broken down into sub-problems and each sub-problem is further broken down into smaller sub-problems using breakdown procedures that are repeated until sub-problems containing the simplest and most essential expressions of the problem characteristics are obtained. Then, using the obtained collection of sub-problems, the design problem is hierarchically constructed using a small number of levels, with the system level located at the highest level. Step 2. The goal value of the objective function located at the highest level is defined. Step 3. At the highest level, the values of design variables are set so that the objective function value approaches the goal value, and these design variables settings are then transferred to the next lowest hierarchical level. These procedures are successively repeated for the lower levels until the lowest hierarchical level is reached. Step 4. At the lowest level, the transferred design variable values are set as goal values for the most detailed design variables and after optimization; the obtained values are then transferred to the next highest hierarchical level where they are used to modify the design variable goal values. These procedures are successively repeated for the remaining higher levels until the highest level is reached. Step 5. If the convergence conditions are satisfied at the highest level, the design variables are considered to be optimum design variables. If the convergence conditions are not satisfied, the procedures in Steps 3 and 4 are repeated. 6.3.2.3 Multidisciplinary Two-Stage Multiobjective Optimization Method In this multidisciplinary two-stage multiobjective optimization method, after obtaining temporary design solutions for each expert field, overall optimum design solutions are obtained in the system division by using the design variable values obtained in each field as the initial design solutions. The objective function, constraint functions, and design variables in each field optimization are included in the multiobjective optimization method for the system division optimization. This optimization method was applied to the design problem of the MicroLabSat piggyback satellite shown in Fig. 6.10 [34]. The optimization problem is

142


divided into two divisions, the structural design division and the component arrangement design division, as shown in Fig. 6.11. In the structural design division, the structural strength and rigidities of the satellite design are evaluated and optimized, while the allocation of the various instrument components installed in the satellite is determined in the component arrangement design division. Figure 6.12 shows the structural analysis model used by the structural design division.

(a) Exterior appearance

(b) Interior appearance

Fig. 6.10 Micro-LabSat piggyback satellite

System Design variables

...

Objective functions

... Structural weight, Natural frequencies,

Structural plate thickness, Component coordinates Moment of inertia ratios

Constraint conditions ... Stress, Natural frequencies,

Side constraints for structural plate thickness, Inclination angle of inertia principal axis, Component coordinates

Plate thickness, Component coordinates

Sub-system(structure analysis division)

Sub-system(component arrangement division)

...

Structural plate thickness

Design variables

...

Component coordinates

Objective functions

...

Structural weight, Natural frequencies

Objective functions

...

Moment of inertia ratios

Constraint conditions

...

Constraint conditions

...

Design variables

Stress, Natural frequencies, Structural weight, Side constraints for structural plate thickness

Inclination angle of inertia principal axis, Component coordinates

Fig. 6.11 Outlines of two-stage multiobjective optimization for multidisciplinary regions


143

z x

y

Fig. 6.12 Model for structural analysis

The following explains the optimization process used in the component arrangement division. The satellite uses a spin control method in which orientation is controlled by having the satellite rotate around the z axis; hence, to ensure stable satellite orientation, its design must consider moments of inertia. When the moment of inertia about the x , y , and z axes are respectively I xx , I yy , and I zz , each of the following moment-of-inertia ratios, S1 and S 2 , must be as large as possible to ensure stability of the satellite during z axis rotation: S1 = I zz / I xx

,

S 2 = I zz / I yy

Unless both of these ratios are above a certain value, counterweights must be installed inside the satellite to increase appropriately the moments of inertia, but doing so increases the satellite’s weight and unacceptably increases the throwweight of the package the rocket must deliver to orbit. The magnitudes of the moments of inertia are influenced by the shapes and positions of panels and the locations of components installed in the satellite. Among these influences, the locations of the components have the largest effect, and changes in various locations can profoundly and effectively alter the moment of inertia values. As shown in Fig. 6.13, the panels that receive nine various components such as batteries, a computer, attitude control equipment, and so on, are joined together in a Y configuration. The component arrangement design problem

144


is to determine the locations of the nine components that will maximize both moments of inertia. The constraints that must be satisfied are that no components should overlap, that the distance between adjacent components should be greater than certain specific magnitudes, that certain combinations of components should not be installed on the same panel (due to panel strength limitations as well as heat dissipation factors), and that certain components have preferred locations and orientations on a given panel. Thus, the number of constraint conditions that need to be satisfied is rather large.

(a) Plan view from the top

(b) Elevation showing arrangement

Fig. 6.13 Features of components installed on the panels

A two-stage optimization procedure was used for this problem. First, in stage 1, rough locations of the components are determined using a GA, since the locations are discrete design variables. Then, using stage 1 solutions as the initial design variables, detailed locations expressed as continuous design variables were determined in stage 2, using the SQP method. The results obtained by these optimizations are shown in Fig. 6.14. The horizontal axis corresponds to the moment of inertia S1 , while the vertical axis corresponds to the moment of inertia S 2 . The o marks show the Pareto optimum solutions obtained by the stage 1 optimization that used a GA, while the □ marks show the Pareto optimum solutions obtained by the stage 2 optimization using SQP in which solutions obtained by the GA are used as the initial design solutions. The straight lines in the figure show correspondence relationships between the GAand SQP-obtained solutions. It can be seen that each moment of inertia is greatly increased from the initial results based on component positions obtained using the GA, compared to the detailed results for the component positions obtained by SQP. Also notable is that S1 and S 2 have a tradeoff relationship.

6.4 System Design Optimization Strategies

145

Fig. 6.14 Optimized results for moment of inertia ratios

The structural design division is in charge of the design concerning the strength and rigidity of the satellite structure itself. The design variables are the plate thicknesses of the four kinds of panels, namely, the upper deck, the lower deck, the panel upon which the components are installed, and the side panels. The objective functions are the natural frequency of the satellite, which must be maximized, and the structural weight of the satellite, which must be minimized. In the system division, the optimum design solutions of the total satellite system were obtained as a multiobjective optimization problem that includes all objective functions used in the lower divisions. A GA was used during the system level optimization.


6.4.1 Features of Machine Product Characteristics and Fundamental Optimization Strategies Practical design optimization problems are usually very complex and include conflicting characteristics. When attempting to solve complex design optimization problems, understanding the features of the characteristics included in the optimi-

146


zation problems and grasping the interrelationships among the characteristics is most important. To achieve this, simplification concepts are first applied, such as breaking down of each performance characteristic into simpler characteristics or extraction of simpler characteristics from the original characteristics, as described in Sect. 6.2. Figure 6.15 shows the flow of system design optimization strategies based on simplification of characteristics [35 ̶ 37].

Simplification of characteristics

Clarification of priority relationships among characteristics

Identification of conflicting relationships at each corresponding hierarchical level Fig. 6.15 Flow of system optimization strategies based on simplification of characteristics

The features of characteristics pertaining to machine product designs are explained using industrial machines as examples. Figure 6.16 shows machine product examples: a machine tool, an industrial robot, a punch press, and an injection molding machine. In Fig. 6.16 a, c, and d, a workpiece or a part is processed between points A and B, and in Fig. 6.16 b, a part or tool is handled at point A. Machine products have designed functions that accomplish specific tasks by the movement and operation of certain machine parts. When designing such devices, the operational accuracies and the times taken to complete specific tasks are evaluated so that overall capability and efficiency of the machine can be assessed. Here, the accuracy and efficiency are concurrently evaluated, and higher values of both are preferable. The product manufacturing cost must always be minimized in actual manufacturing scenarios. In practical product optimization, the accuracy, efficiency, and product manufacturing cost are all considered and are primary performance characteristics. Each performance characteristic is usually quite complex, since it is expressed as compounds or additions of various other component characteristics. Optimum design solutions for these performance characteristics are usually different, implying that the characteristics have conflicting interrelationships, a primary obstacle to obtaining globally optimum solution.


A

147

A B Tool or part

Workpiece

(a) Machine tool

(b) Industrial robot A B

Plate-type workpiece A B

(c) Punch press machine

Mold part

Die plates

(d) Injection molding machine

Fig. 6.16 Examples of machine products

To clarify the interrelationships among characteristics, each is examined so that its composition, dynamic behavior, and mathematical expression can all be clearly expressed, taking the particular details of the optimization problem fully into account. For example, machine accuracies are often expressed in terms of static and/or dynamic displacements at specific points that are determined according to the requirements of specific jobs. Similarly, static rigidities k s can be used to evaluate static displacements, and dynamic rigidities k d can be used when evaluating dynamic displacements. In general, machine products are classified into two groups: those for which static rigidities alone are evaluated, and those that additionally require evaluation of dynamic rigidities. Since machine products accomplish tasks by the movement and operation of various parts, it is usually necessary to evaluate and optimize dynamic rigidities as well as static rigidities. In such cases, the relationships between the static rigidities k s and the dynamic rigidities k d should be clarified. Machine products are composed of structural members, machine elements, and joints. The structural members define the framework of the product, and profoundly affect the structural rigidities k M and material cost C M . The joints, which function as connectors for structural members or machine elements, have their own rigidities and these are particularly important with respect to damping

148


effects that affect vibration characteristics. Ultimately, the machining accuracies of joint contact surfaces determine the accuracies of the machine products, and joint contact surface machining costs occupy increasingly large percentages of the total manufacturing cost for machines capable of higher product accuracies. Product performances such as accuracies and efficiencies are usually quite complex, and are often expressed as compounds of basic characteristics after simplification and idealization of the original characteristics. For example, the total product manufacturing cost is a compound of various costs, each of which has unique features. Similarly, the total operational energy requirement is also a compound of various energy requirements. Obtaining a clear picture of the interrelationships among characteristics just by examining the characteristics in their original form is often difficult, but simplification concepts, when applied to the most important characteristics, can be used as an effective tool for this purpose. In what follows, the priority and conflicting relationships among characteristics are considered based on the simplification or idealization of characteristics, including product design optimizations.

6.4.2 Priority Relationships among Characteristics Since the priority relationships among characteristics are directly related with the sequence in which optimization procedures will be carried out, these priority relations are discussed first. Characteristic A has priority over characteristic B when the following two conditions are satisfied for optimizations that include characteristics A and B: 1. Characteristic A affects characteristic B and improvement in characteristic A causes improvement in characteristic B 2. Changes in characteristic B have no clear influence upon characteristic A Characteristics in system optimization problems are generally categorized into two types: global characteristics (GC) and local characteristics (LC). Structural member rigidities are a good example of GCs. They behave monotonously when the values of design variables specifying structural member dimensions are increased or decreased and structural member rigidities generally have conflicting relationships with structural member weights, which also behave monotonously when the values of design variables specifying structural member dimensions are increased or decreased. With structural member rigidities and weights categorized as GCs, their optimization often yields a single local optimum solution, according to considerations based on the KKT conditions described in Sect. 6.1.2. On the other hand, structural joints are usually modeled using spring and damper combinations that provide certain spring stiffnesses and damping coefficients. The spring stiffnesses and damping coefficients modeling the structural joints have local maxima for dynamic rigidities in the feasible design variable


149

space, so these are usually categorized as LCs. When optimizations include LCs, there may be many local optimum solutions, according to considerations based on the KKT conditions described in Sect. 6.1.2. GCs have priority over LCs during the optimization. When only GCs are included in the optimization problem, simultaneous optimization of all the characteristics is reasonable, but when LCs are included, simultaneous optimization of the characteristics may result in local optimum solutions that prevent obtaining a global optimum solution. Dynamic characteristics are typically evaluated and optimized in machine product design optimizations, since machines usually operate by the movement of machine elements and structural members, and characteristics belonging to both GC and LC types are included and processed. For optimization problems including both types of characteristic, it is important to clarify the priority relationships within the optimization processes. The rules concerning priority relationships among characteristics pertaining to system optimization based on the foregoing discussion are as follows: 6.4.2.1 General Rules for System Optimization The following priority rules exist for priority relationships among characteristics for product system optimization: “Optimization of characteristics pertaining to the total system has priority over optimization of characteristics pertaining to a local area. Optimization of total system characteristics should precede optimization of local area characteristics.” These rules can be alternatively expressed as: “Optimization of characteristics whose influence extends over a wide area has priority over optimization of characteristics that only influence local areas” and “Characteristics that have a greater influence on other undetermined characteristics should be determined earlier in the decision sequence than other undetermined characteristics.” 6.4.2.2 Specific Rules for System Optimization The following are rules for structural optimizations including structural members and their joints, in which static rigidities, dynamic rigidities and stresses must be evaluated. (1) Relationship between static rigidity and stress distribution Figure 6.17a shows a conceptual diagram of the framework model where an external force ( F ) is applied at points A and B, and the relative displacement of X ( X = X A − X B ) ) between points A and B is evaluated. The static rigidity k s is expressed as F / X . The static rigidity can be obtained even when only the framework model of the product, such as beam models, is given, that is, prior to specifying detailed design information. Rigidity analysis of the model design framework

150


returns values for the internal force magnitudes ( I F ) at a specific framework point, such as point I shown in Fig. 6.17b, and after defining the detailed surface shape at that point, the surface stress distributions ( σ ) can be determined. The design of the total structural system greatly affects the static rigidity and can be obtained without detailed design information, so optimization of the static rigidity has priority over optimization of the stress distributions. This priority relationship is a basic rule for items (3) and (4) that follow, which pertain to internal forces of the structure and detailed stress distributions. I A F XB B

XA

Joint J

I1

F Surface

Contact Surface

(b)

(c)

(a)

I2

Fig. 6.17 Conceptual framework of machine products

(2) Relationship between static and dynamic rigidities Figure 6.18 shows a conceptual diagram of the frequency response of the receptance X / F between points A and B in Fig. 6.17a. The applied force F acts between points A and B when the processing action is performed. To evaluate the process accuracy, the relative displacement of X ( X = X A − X B ) is measured. The receptance frequency response between the applied force F and the displacement X is expressed as follows: r (ω ) =

X F

∞

(ω ) = ∑ [

fm

m =1 1 − ( ω ) 2 + 2 j ( ω )ς ωm ωm m

]

(6.21)

where ω is the frequency, and ωm , f m , ζ m are respectively the natural frequency, modal flexibility, and damping ratio at the m-th natural mode. " j" denotes an imaginary unit. The static rigidity k s is obtained using the reciprocal of the static compliance f s , while the dynamic rigidity kd is obtained using the reciprocal of the maximum receptance value rmax over the entire frequency range.


151

Fig. 6.18 Example of frequency response at working points between points A and B in Fig. 6.17a.

When the frequency ω is set to 0 in (6.21), the following simple relationship between f s and modal flexibility f m is established [38]: ∞

f s = ∑ fm m =1

(6.22)

Both f s and f m have positive values. The modal flexibility f m ( m = 1, 2, ..., ∞ ) expresses the distributed magnitude of the static compliance f s for each natural mode. Equation (6.22) indicates that minimizing the static compliance f s , which is equivalent to maximizing the static rigidity, reduces the modal flexibility at the natural mode where the modal flexibility value is highest. The term j (ω / ωm )ζ m in the denominator of (6.21) has an effect only in the neighborhood of each natural frequency ωm . Thus, a graph of the frequency response over the entire frequency rage can be determined without considering damping effects. Reducing the static compliance reduces the magnitude of the receptance over the entire frequency range and generally increases the natural frequency of each natural mode, as illustrated by the dotted line in Fig. 6.18. Examination of the related characteristics shows that increasing the static rigidity k s increases the dynamic rigidity kd . Optimization of the static rigidity should therefore have priority over optimization of the dynamic rigidity.

152


(3) Relationship between structural members and joint rigidities Figure 6.17c is a structural model similar to Fig. 6.17b, but has structural members connected by a joint at point J. The internal shape details of structural members are not always required when evaluating structural member rigidity k M , but detailed contact surface shapes are required when evaluating the structural joint rigidity k J . The dimensions of a product’s structural members determine its outline as a whole, but the location of the joints seldom influences the external contour. The dimensions of the structural members obviously affect the internal forces operating between structural members. Although joint design variables do not fundamentally alter these internal forces, they do change the stress distributions of joint contact surfaces. Optimization of the structural member rigidity should therefore have priority over optimization of the structural joint rigidity. (4) Relationship between structural joint rigidity and joint damping coefficients Structural joint rigidities significantly influence the static rigidity k s , and also influence the entire spectrum of receptance frequency responses such as shown in Fig. 6.18. On the other hand, each joint damping coefficient locally influences the receptance frequency response near specific natural modes, indicated by the small circles, and the damping ratio ζ m at each natural mode is thereby determined. Optimization of the structural joint rigidity should thus have priority over optimization of the joint damping coefficients. Damping effects are much larger at the joints, and the magnitude of these effects largely depends on the contact conditions and shapes of the joints. Furthermore, the precise manner in which damping arises is completely different for each natural mode, since each has a different mode shape and different stress distributions at each contact location. These particular phenomena are what cause the many local optimum solutions in design optimization problems that include dynamic characteristics.

6.4.3 Creation of Hierarchical Optimization The priority relationships (relationships among simplified characteristics that will be optimized before other characteristics), which were obtained as described in Sect. 6.4.2, determine the order of hierarchically applied optimization procedures. When the performance characteristic is the dynamic rigidity kd , this can be broken down into three simpler characteristics, namely, structural member rigidity k M , joint rigidity k J , and the damping ratio ς m . The value of k M is independent of the values of k J and ς m , and, similarly, k J is independent of ς m . However, the value of k M does influence the evaluation of k J and ς m , and the value


153

of k J influences the evaluation of ς m . That is, increases in structural member rigidities cause increases in their cross-sectional widths, which increases the contact surface area of joints, increasing the joint rigidities. As structural member rigidities increase, the joints become more influential in the overall design, due to damping effects that are important in machine products, and the damping ratios at each natural mode are also increased. In this scenario, maximizing k M generally results in a corresponding maximization of k J and ς m . Therefore, to maximize the magnitude of kd , the multi-step decision-making process should handle these three evaluative factors in the order of k M , k J , and ςm .

6.4.4 Conflicting Relationships Between Characteristics Basic hierarchical optimization procedures are now constructed based on the priority rules explained above. Each characteristic in a hierarchical optimization problem generally has a conflicting relationship with the simplified characteristics included in the optimization problem, and these conflicting characteristics are included in the hierarchical formulation of the optimization procedures. Characteristics that have conflicting relationships are arrayed in various sub-optimization groups. For example, maximization of structural member rigidity has a conflicting relationship with minimization of the structural member weight. Maximization of the joint rigidity has a conflicting relationship with minimization of the machining cost for the joint contact surfaces. Characteristics having conflicting interrelationships are grouped as sub-optimization problems and are optimized as a multiobjective optimization problem.

6.4.5 Construction of Hierarchical Optimization Procedures Based on the foregoing considerations and discussion, the optimization methodologies discussed here have the following features: 1. The order in which the hierarchical optimization procedures are carried out follows the priority relationships established among the simplified characteristics . 2. Since optimizations at each hierarchical stage have conflicting interrelationships among the characteristics, multiobjective optimization methods are used Multiobjective problems are solved as a series of optimizations carried out at each hierarchical level. Each multiobjective optimization problem is called a SubOptimization Problem (hereafter SOP).

154


The optimization procedure begins at the bottom level of the simplest characteristics and the results, in the form of Pareto optimum solutions, are transferred to the next higher hierarchical level. In this way, the Pareto optimum design solutions obtained in SOPs are included in the input variables for optimization of the SOP located at the next higher level, as shown in Fig. 6.19. Design solutions at discrete points on the Pareto optimum solution set are then transferred for use in an upper level optimization. The results of Pareto optimum solutions obtained by each optimization are cumulatively utilized to obtain Pareto optimum solutions for each SOP and, finally, Pareto optimum solutions at the top hierarchical level. Product designers can then select the most appropriate design solution from a range of alternative designs included in the highest level Pareto optimum solution set, enabling design decisions from a wider viewpoint. When the product design requirements are decided, the environment where the product will be used is taken into account, so the most suitable design solution can be selected from among a number of Pareto optimum design solutions.

fc

fd

Transfer data and create initial Pareto solutions at upper level

fa

fb

Characteristic

Sub-optimization problem (SOP)

Fig. 6.19 Transmission of Pareto solutions to the next upper level

Appropriate settings for the initial values of the design variables facilitate the step-wise procedure of the hierarchical optimization strategies, and optimum solutions can then be effectively obtained.


155

6.4.6 Practical Procedures for Product Optimization Practical procedures are now explained using applied examples. Figure 6.20 shows a framework model of a machine tool composed of structural members and joints, Model A. The performance characteristics to be considered are the static and dynamic rigidities at the cutting point and the manufacturing cost of the machine tool. The objective functions are the maximum receptance value rmax and the machine’s manufacturing cost CT , each of which should be minimized. The characteristic of the maximum receptance value rmax is broken down into two characteristics, namely, the static compliance f s and the damping ratio ς . The manufacturing cost CT is broken down into the material cost C M of the structural members and the machining cost C J of the joints.

Motor 1 m[kg] Structural member 4

Joint 6 Joint 3

Joint 4

Structural member 5 F

A

Table 1 M[kg]

Cutting point B

Joint 5

F

Structural member 1

Joint 2

z y

Joint 1 Structural member 2

x Fig. 6.20 Framework model of a milling machine (Model A)

Structural member 3

156


Fig. 6.21 Hierarchical multiobjective optimization procedures

Structural member 4

Structural member 5 F

A

Cutting point

B Structural member 1

F

Structural member 2 Fig. 6.22 Structural model of the static force loop (Model B)

Structural member 3


157

The hierarchical optimization procedures are shown in Fig. 6.21, and the optimization is conducted as follows: Step 1. Using the simplified Model B shown in Fig. 6.22, the multiobjective optimization problem for the static rigidity k M and the total structural weight Ws of the structural members on the static force loop is solved in SOP 3 at the bottom level in Fig. 6.21, since the structural member rigidity k M has the highest priority among the characteristics, and Ws clearly has a conflicting relationship with k M . Since obtaining the desired static rigidity between points A and B only involves the structural members and joints on the static force loop shown in Fig. 6.22, the other structural members, machine elements, and joints have been removed from Model A to simplify the optimization modeling. Furthermore, each joint in the structural model is treated as a rigid joint, since structural member rigidity optimization has priority over the structural joint rigidity optimization. The design variables d i are the cross-sectional dimensions of each structural member. A Pareto optimum solution set of cross-sectional dimensions is then obtained. Step 2. In SOP 2 shown in Fig. 6.21, the multiobjective optimization problem is solved for the static compliance f s (the reciprocal of the static rigidity k s ) and the total manufacturing cost CT of the structural members on the static force loop, which is the sum of the material cost C M and the machining cost C J of the joints. Each joint in Model B is now treated as a flexible joint, modeled as a spring, and the maximum surface roughness R max of the joint contact surface is included in the design variables. The results of the cross-sectional dimensions obtained in Step 1 are used as initial design variables for this optimization. In this optimization, the relationships between the surface roughness Rmax and the machining cost cu per unit contact surface, shown in Fig. 6.23, are used, and three kinds of machining methods, milling, grinding, and super finishing, are considered. The joint rigidities are calculated according to their surface roughness values and contact surface areas. The material cost C M is calculated by multiplying the material cost per unit weight by Ws . Step 3. In SOP 1 shown in Fig. 6.21, the multiobjective optimization problem for the maximum receptance value rmax and the total manufacturing cost CT is solved and a Pareto optimum solution set is obtained. The comprehensive structural model now used is shown in Fig. 6.20, where structural members, machine elements, and joints other than those on the static force loop are included. The optimization results for the cross-sectional dimensions and spring stiffnesses obtained in Step 2 are used as initial design variables. The damping ratio ζ at each natural mode is held to a constant value. Step 4. At the highest hierarchical level in Fig. 6.21, the multiobjective optimization problem for the maximum receptance value rmax and the total manufacturing cost CT is solved, using the optimization results for the cross-sectional dimen-

158


sions and spring stiffnesses obtained in Step 3 as initial design variables, with the damping coefficients of the joints included in the design variables. The final Pareto optimum solution set between the maximum receptance value rmax and the total manufacturing cost CT is then obtained. 30

25

Super Finishing

Grinding

Milling

20

15 10

5

Smoothly modified original Pareto optimum solution line 0 0.1×10-8

0.1×10-7

0.1×10-6

0.1×10-5

0.1×10-4

Surface roughness R max [m] Fig. 6.23 Relationship between surface roughness and machining cost per unit contact area

Figure 6.24, illustrating the result of Step 1, shows the Pareto optimum solution set line between the static rigidity k M and the total structural weight Ws of the structural members on the static force loop. Figure 6.25, displaying the Step 2 result, shows the Pareto optimum solution set line between the static compliance f s and the total manufacturing cost CT of the structural members and joints on the static force loop. Figure 6.26, the result of Step 3, shows the Pareto optimum solution set line between the maximum receptance value rmax and the total manufacturing cost CT of the comprehensive structural model, where the damping ratio at each natural mode is kept constant.


159

1600 1,600 Total Weight WM [kg]

Total weight WM [kg]

1400 1,400 1200 1,200 1000 1,000 800 800 600 600 Structural member 4

400 400

Structural member 5 Structural Cutting point member 3 F

200 200 0

F

0 1.00×10 1.00E-07-7

Structural member 1 Structural member 2

1.00×10 1.00E-06-6

1.00×10-5 1.00E-05

1.00×10 1.00E-04-4

Staticcompliance Compliance f s [m/N] Static fs [m/N] Fig. 6.24 Pareto optimum solution line for Step 1

1800000 1,800,000

Total manufacturing cost CT [yen]


2000000 2,000,000 1600000 1,600,000 1,400,000 1400000 1,200,000 1200000 1,000,000 1000000 800,000 800000 600,000 600000 400,000 400000 200,000 200000 00

1.E-07 -7 1.00×10

1.E-06 -6 1.00×10

1.E-05 -5 1.00×10

1.E-04 -4 1.00×10

Static fs [m/N] Staticcompliance Compliance f s [m/N] Fig. 6.25 Pareto optimum solution line for Step 2

1.E-03 -3 1.00×10

160


1800000 1,800,000 1600000 1,600,000 Total Cost CT [yen]


2000000 2,000,000

1400000 1,400,000 1200000 1,200,000 1000000 1,000,000 800000 800,000 600000 600,000

400,000 400000 200,000 200000 0 0 1.00×10 1.E-07 -7

1.00×10 1.E-06 -6

1.00×10 1.E-05 -5

1.00×10 1.E-04 -4

1.00×10 1.E-03 -3

Maximum Frequency Response r max [m/N] Maximum frequency response rmax [m/N]

Fig. 6.26 Pareto optimum solution line for Step 3



3000000 3,000,000

Conventional method Proposed method

2500000 2,500,000 2000000 2,000,000 1500000 1,500,000 1000000 1,000,000 500,000 500000 00 -7 1.00×10 1.E-07

Q P -6 1.00×10 1.E-06

-5 1.00×10 1.E-05

-4 1.00×10 1.E-04

Maximum frequency response r max [m/N]

Fig. 6.27 Pareto optimum solution line for Step 4

-3 1.00×10 1.E-03


161

Figure 6.27, the Step 4 result, shows the Pareto optimum solution set line between the maximum receptance value rmax and the total manufacturing cost CT , with the damping coefficients at the joints included in the design variables. To demonstrate the effectiveness of the proposed method, these results are compared with results obtained using a conventional method where the performance characteristics (Step 4 objective functions) are directly optimized using the feasible direction method, but without using the proposed hierarchical optimization procedures. The conventional method results are plotted with “” marks in Fig. 6.27, while the proposed method’s results are shown with “▪” marks. The thin line indicates the Pareto optimum solution line, which is the optimum solution frontier. The comparison shows that more preferable solutions are more reliably obtained when the proposed method is used. In conventional optimization procedures, solutions are searched for within an initially large feasible region. However, if inappropriate feasible design regions are included when the searching process is carried out, satisfactory convergence to useful solutions may be hindered, or even prevented, by the attraction of design solutions in these inappropriate regions. In the applied example described above, machine product models are updated from the simplest initial model to one that is more detailed, as shown in Fig. 6.28, as the hierarchical optimization procedures progress. To simplify each optimization problem, only the characteristics being evaluated are modeled. The main feature of this method is that preferable initial design solutions become the basis for subsequent hierarchical phases of the optimization, and the overall sequential process continues to a useful convergence.

(a)

(b)

Rigid joint

Flexible joint

(c) Modeling of total system

Modeling on the static force loop Fig. 6.28 Model changes during the hierarchical optimization process

162


6.4.7 Discussion Concerning System Design Optimization The advantages of the system optimization procedures described in Sect. 6.4 above are: 1. The possibility of obtaining global optimum solutions is increased for complex system design optimization problems 2. Evaluations of the obtained optimum solutions are facilitated, and solution features can be more easily and systematically grasped Most machine products are composed of structural members, machine elements, and joints, so achieving good practical results requires the evaluation and optimization of dynamic rigidities. The fundamental optimization methodologies described above are therefore applicable to a broad range of machine product designs. In addition, the fundamental techniques of hierarchical optimization methodologies based on simplification concepts for characteristics can be applied to many general optimization problems. When a characteristic is effectively simplified, processing and optimization of the simplified characteristics should almost always be carried out before processing and optimizing the original characteristic. In practice, effective hierarchical optimization procedures develop naturally according to the relationship between the original characteristic and the simplified characteristics that are derived or extracted.

6.5 Optimum Selection Method for Alternative Design Solutions To make the end result of product design optimizations more effective, it is necessary to prepare as many alternative design candidates as possible and then select the best one. Designed products are usually composed of various parts, elements and structural members that are most often connected in a hierarchical manner. A small number of alternatives, or many, may exist for each component at a given hierarchical level. Figure 6.29 shows a conceptual diagram that expresses the relationship between a hierarchical structure composed of components and various alternatives. Product design optimizations that include discrete design variables, such as the selection of one property from a range of alternatives, tend to be highly complex, and Hierarchical Genetic Algorithms (abbreviated HGAs) [39] are useful tools in optimization methodologies that aim to provide practical solutions to such design problems.

6.5 Optimum Selection Method for Alternative Design Solutions

163

Product Component Attribute Component alternatives

Attribute alternatives

Fig. 6.29 Conceptual diagram of HGAs

In conventional GAs, one-dimensional vector genotype arrays of design variables are used, but with HGAs, hierarchical genotype coding representations are used. This allows alternatives for both detailed designs at the lower hierarchical levels and functions and mechanisms at the higher levels to be concurrently included in the design variables. Therefore, problems in which decision- making processes are hierarchically expressed can also be formulated easily and solutions effectively obtained. To explain the principle of the methodology here, consider the problem of selecting an optimized design solution for a simple machine structure that has a number of alternatives at the level of substructure components in the hierarchical representation shown in Fig. 6.30. The represented machine is composed of substructures A and B, which respectively have alternatives A1 and A2, and B1, B2, and B3.

164


Sub-structures

A

Sub-structure A alternatives

A1 a

Sub-structures

a1

a2

A2

a3

Sub-structure a alternatives

B

Sub-structure B alternatives

B1

B2

B3

b

b1

b2

Sub-structure b alternatives

Fig. 6.30 Machine structure having alternative sub-structures

Furthermore, substructure A2 is composed of two elements, a and b, and element a has three alternatives, a1, a2, and a3, while element b has two alternatives, b1 and b2. To enable the optimization, the foregoing relations are represented as a hierarchical structure using symbols as shown in Fig. 6.31, where the numbers within parentheses “( )” express the number of alternatives for each substructure. That is, when a node has n substructures at the next lowest level and each substructure i has hi ( i = 1,2, ..., n ) alternatives, these relationships are represented as ( h1, h2 , ..., hn ) . When the substructure corresponds to the t-th alternative for the sth substructure at the substructure’s upper level, the prefix symbol “s, t-” is added to the original notation.

(2 　,　3)

1 , 2 -(3 　,　2) Fig. 6.31 Hierarchical symbolic expression of machine structures having alternative substructures

6.5 Optimum Selection Method for Alternative Design Solutions

165

If there are further upper level substructures, these procedures are repeated, and additional symbols are used in the notation. The design variable used to select an alternative from among many alternatives of a substructure corresponds to a gene in the HGA. Consider a case where the alternative shown in Fig. 6.32 is selected from the hierarchical expression of alternatives shown in Fig. 6.30. Since A2, the second alternative of A, and B1, the first alternative of B, are first selected at the top level, the notation is [2,1] as shown in Fig. 6.33. The numbers in the brackets “[ ]” indicate the design variables selected from alternatives at each substructure. Since a3 and b1 are selected in A2, the notation becomes [3,1]. To represent that the substructure belongs to A2 in the hierarchy, the prefix “1,2-” is added, giving 1,2[3,1]. That is, when the substructure corresponds to the t-th alternative at the s-th substructure at the higher level, the prefix “s,t-” is added to the original notation [ g1, g 2 , ..., g n ]. If further higher level substructures exist, similar procedures are repeated, and the corresponding prefixes are added to the notation. Thus, the alternatives selected in the hierarchical structure have notational expressions as shown in Fig. 6.33.

A A2 a a3

B B1

b b1

Fig. 6.32 Hierarchical expression of individual genotypes

166


〔2 　,　1〕

1 , 2 -〔3 　,　1〕 Fig. 6.33 Hierarchical notational expression of individual genotypes

Figure 6.34 shows an example of crossover operations in the foregoing hierarchical structural model, where genes are exchanged among substructures. Here, a crossover operation is conducted for parent 1 and parent 2. Parent 1 and parent 2 have different genes at the circled gene positions. The right portion of Fig. 6.34 illustrates the exchange of genes. When such genes are exchanged, the lower genes belonging to the original genes are also exchanged. That is, parent 1 has the “2” gene which contains the “1,2-[3,1]” gene at the lower level, so when child 2 is made during the crossover operation, the “1,2-[3,1]” gene is accompanied by the “2” gene. Different

Exchange

〔 2 　,　1〕〔 1 　,　3〕 Parent 1

Parent 2

1 , 2 -〔3 　,　1〕

〔1 　,　-1〕〔2 　,　3〕 Child 1

Child 2

1 , 2 -〔3 　,　1〕

Fig. 6.34 Example 1 of crossover operations

Such crossover operation rules are not only applied to the gene at the highest level, but also to lower level gene crossover operations. In the case of the example shown in Fig. 6.35, since the gene positions circled at the higher level have the same gene, the gene exchanges are conducted between the gene positions circled at the lower level. The mutation operations are first conducted for the set of genes at the highest level of the hierarchical system, and similar operations are applied serially to the sets of genes at the lower levels.

Exercises

167

Same

〔 2 　,　1〕〔 2 　,　3〕 Parent 1

Parent 2

1 , 2 -〔 3 　,　1〕 1 , 2 -〔 2 　,　2〕 Different

〔2 　,　1〕

〔2 　,　3〕

Child 1

Child 2

1 , 2 -〔2 　,　1〕 1 , 2 -〔3 　,　2〕 Exchange

Fig. 6.35 Example 2 of crossover operations

Exercises 6.1 Explain the factors that make optimization practices difficult for each of the following standpoints: nonlinear optimization problems, multiobjective optimization problems, discrete design variable optimization problems, and largescale optimization problems. Discuss accepted methodologies for overcoming the above difficulties. 6.2 Discuss the features, differences, advantages, and disadvantages of GAs and branch-and-bound methods for optimization of design problems that include discrete design variables. 6.3 Discuss the specific difficulties when trying to obtain optimum design solutions for optimization problems of machine products that require higher operational precision and higher operational efficiency from the standpoints of (1) nonlinear optimization problems, (2) discrete design variable optimization problems, and (3) multiobjective optimization problems. Then, discus optimization methodologies that can resolve such difficulties for the following hypothetical design optimization problem: Criteria: product weight, static rigidity Design variables: three alternative structural materials, A, B, and C N design variables for structural member dimensions Constraint: Upper and lower bounds of design dimensions 6.4 Discuss how structural rigidities and joint rigidities differ when performing optimization of structural designs. 6.5 Discuss how structural member material costs and machining costs of structural joints differ when performing optimization of structural designs. 6.6 Why are optimization procedures based on simplification concepts particularly useful for obtaining optimum product design solutions?

168


References 1. 2. 3. 4. 5.

Johnson RC (1961) Optimum design of mechanical elements. 2nd edn. Wiley, New York Wilde DJ (1978) Globally optimal design. Wiley-Interscience Publication Haug EJ, Arora JS (1979) Applied optimal design. Wiley-Interscience Publication, New York Siddall JN (1982) Optimal engineering design. Marcel Dekker, New York Vanderplaats GN (1984) Numerical optimization techniques for engineering design: with applications. McGraw-Hill, New York 6. Haftka RT, Kamat MP (1985) Elements of structural optimization. Martinus Nijhoff 7. Papalambros PY, Wilde DJ (2000) Principles of optimal design. 2nd edn. Cambridge University Press, Cambridge 8. Arora JS (2004) Introduction to optimum design. 2nd edn. Elsevier 9. Mastinu G, Gobbi M, Miano C (2006) Optimal design of complex mechanical systems, Springer 10. Karush W (1939) Minima of functions of several variables with inequalities as side conditions. M.S. thesis, Department of Mathematics, University of Chicago 11. Kuhn HW, Tucker AW (1951) Nonlinear programming. In: Proceedings of the 2nd Berkeley Symposium on Mathematical Statistics and Probability, J. Neymmmaan (Ed.), University of California Press, Berkeley, CA:481 ̶ 492 12. Yoshimura M (2008) System design optimization strategies for machine products. In: The 5th China-Japan-Korea Joint Symposium on Optimization of Structural and Mechanical Systems (CJK-OSM5), Korea:1 ̶ 6 13. Eschenauer H, Koski J, Osyczka A (eds) (1990) Multicriteria design optimization. Springer 14. Papadimitriou CH (1982) Combinatorial optimization: algorithms and complexity. PrenticeHall 15. Baker KR (1974) Introduction to sequencing and scheduling. Wiley, New York 16. French S (1982) Sequencing and scheduling: an introduciton to the mathematics of the jobshop. Ellis Horwood 17. Gutkowski W, Bauer J. (eds) (1994) Discrete structural optimization. Springer 18. Hitomi K, Yoshimura M (1986) Operations Scheduling for Work Transportation by Industrial Robots in Automated Manufacturing Systems. Material Flow, 3:131 ̶ 139 19. Holland J (1975) Adaptation in natural and artificial systems. University of Michigan Press 20. Goldberg DE (1989) Genetic algorithms in search, Optimization and machine learning. Addison Wesley. 21. Yoshimura M, Yoshimura Y, Izui K, Nishiwaki S (2009) Product optimization incorporating discrete design variables based on decomposition of performance characteristics. ASME Journal of mechanical design, Vol.131:031004,1 ̶ 10 22. Nishigaki J, Nishiwaki S et al. (2001) First order analysis – New CAE tools for automotive body designers. SAE Paper, No.2001-01-0768 23. Yoshimura M, Hamada T, Yura K, Hitomi K (1983) Design optimization of machine-tool structures with respect to dynamic characteristics. Transactions of the ASME, Journal of mechanisms, transmissions, and automation in design, 105(1):88 ̶ 96 24. Yoshimura M, Takeuchi Y, Hitomi K (1984) Design optimization of machine-tool structures considering manufacturing cost, accuracy and productivity. Transactions of the ASME, Journal of mechanisms, transmissions, and automation in design, 106(4):531 ̶ 537 25. Steward DV (1981) The design structure system: a method for managing the design of complex systems. IEEE transactions on engineering management, 28(3):71 ̶ 74 26. Yoshimura M, Izui K, Fujimi Y (2003) Optimizing the decision making process for largescale design problems according to criteria interrelationships. International journal of production research, 41(9):1987 ̶ 2002

References

169

27. Kusiak A, Wang J (1993) Decomposition of the design process. Transaction of the ASME journal of mechanical design, 115:687 ̶ 694 28. Krishnamachari RS, Papalambros PY (1997) Hierarchical decomposition systhesis in optimal systems design. ASME Journal of Mechanical Design, December, 119:448 ̶ 457 29. Michelena NF, Papalambros PY (1997) A Hypergraph framework for optimal model-based decomposition of design problems. Computational optimization and applications, 8(2):173 ̶ 196 30. Park GJ (2007) Multidisciplinary design optimization. In: Arora, JS (ed) Optimization of structural and mechanical systems, World Scientific 361 ̶ 388 31. Sobieszczanski-Sobieski J, Altus TD, Phillips W, Sandusky R (2003) Bilevel integrated system synthesis for concurrent and distributed processing. AIAA Journal, 41(10):1996 32. Braun RD, Moore AA, Kroo IM (1997) Collaborative approach to launch vehicle design. Journal of spacecraft and rockets, 34(4):478 33. Kim HM, Rideout DG, Papalambros PY, Stein JL (2003) Analytical target cascading in automotive vehicle design. Transaction of ASME, journal of mechanical design, 125(3):481 34. Yoshimura M, Murase Y, Yamauchi M, Izui K, Nisiwaki S, Moriya S (2004) A theoretical consideration of organized analysis environment for optimum design solutions. In: Proceedings of International Conference on Cybernetics and Informatics Technologies, Systems and Applications (CITSA2004), Vol.IV:76 35. Yoshimura M, Taniguchi M, Izui K, Nishiwaki S (2006) Hierarchical arrangement of characteristics in product design optimization. ASME journal of mechanical design, 128:701 ̶ 709 36. Yoshimura M (2007) Global product design optimization strategies based on simplification of product characteristics. In: Conference Proceedings of the 16th International Conference on Engineering Design, ICED’07, August, Paris, France:1 ̶ 12 37. Yoshimura M, Yoshimura Y, Izui K, Nishiwaki S (2008) Fundamental strategies for system optimization of machine product designs, In: Proceedings of ASME 2008 Design Engineering Technical Conferences and Computers and Information in Engineering Conference, DETC2008-49435:1 ̶ 11 38. Yoshimura M (1987) Design optimization of machine-tool dynamics based on clarification of competitive-cooperative relationships between characteristics. Transactions of the ASME, Journal of mechanisms, transmissions, and automation in design, 109(1):143 ̶ 150 39. Yoshimura M, Izui K (2004) Hierarchical parallel processes of genetic algorithms for design optimizations for design optimization of large-scale products. ASME Journal of mechanical design, 126:217 ̶ 224

7 Decision-making Methods

Achieving more preferable product manufacturing depends on obtaining more preferable decision-making results. The supporting technologies explained in Chap. 5 have been developed and implemented to enable the people involved in product manufacturing to exercise more sophisticated and insightful consideration and judgment of various factors and details. The optimization technologies for product manufacturing explained in Chap. 6 have been developed and advanced so that optimized product designs that provide higher satisfaction of a variety of needs can be more efficiently and logically produced. The combination of these technologies are applied to decision-making concerning every aspect of product manufacturing. This chapter describes fundamental decision-making methods for obtaining more satisfactory decision-making results. The use of decision-making support systems and optimization methods improves the likelihood of obtaining more preferable results, compared with results obtained by simply relying upon experience and intuition. Furthermore, the methods described here enable thorough evaluation and examination of the decision-making processes as well as the obtained design results, so that even more preferable results can be developed over time.

7.1 Decision-making Difficulties and Fundamentals of Decision Making

7.1.1 Decision-making Difficulties Situations where decision making is especially difficult include the following: 1. Cases where there are many alternatives

172


2. Cases where there are many evaluative criteria or attributes 3. Cases where there are uncertain factors such as business or weather fluctuations 4. Cases where there are many decision makers, each with a different field of expertise, who must reach a decision-making consensus The last situation corresponds to decision-making in collaborative circumstances, and, in such cases, one decision maker may, one way or another, attempt to maximize his or her influence at the expense of other group members, which may drastically lower the quality of the outcome. In such cases, negative synergy effects may intrude.

7.1.2 Fundamental Schemes to Facilitate Decision-making The following schemes can help make decision making as effective and easy as possible: 1. As the number of design alternatives increases, the difficulty of simultaneously comparing the alternatives increases geometrically. This relationship also holds for criteria that must be evaluated; however, a comparison between two alternatives for a specific criterion (attribute), that is, a pair comparison, is usually easy, and can be performed serially. A systematic evaluation where multiple pair comparisons are integrated can then be employed. 2. Qualitative attributes resist comparison. To allow effective evaluations, qualitative attributes must be replaced by quantitative values, which can then be used in effective decision making. 3. Even when a single solution unambiguously appears to be superior, judging whether or not the decision-making that provided this solution is reasonable or not can be difficult; however, a comparison using some other obtained solutions may facilitate accurate judgment. For example, presentation of a set of comparable solutions, such as Pareto optimum solutions, rather than presentation of just one solution, may make decision-making more reliable and effective. 4. Displays using diagrams that have visual appeal can help decision-makers reach valid conclusions more efficiently and effectively. 5. Since people are usually the final arbiters of design decisions, quantitatively expressing the preference features of the decision-makers themselves may help clarify decision-making details. 6. In collaboration scenarios, construction of a decision supporting system may be of great benefit, and lead to rational compromises and potential synergy effects. 7. Tracing the decision making flow may reveal the cause or nature obstacles preventing fully satisfactory decision-making. Such information may then be used to improve the decision-making support system.

7.2 Fundamentals of Decision-making

173

7.2 Fundamentals of Decision-making When selecting a solution from among a considerable number of alternatives, the results of simulations may be useful, even if such information is not directly used when making the final decision. The most fundamental methods that support these decision-making scenarios are explained below.

7.2.1 Method for Selecting the Best Alternatives when There Are Many Evaluative Factors Consider n alternatives expressed as di (i = 1,2,..., n ) , with m attributes (for example, accuracy, efficiency, cost, required delivery date, mechanical function, and so on) denoted j ( j = 1,2,..., m) . Next, weighting coefficients for each attribute are m

expressed as ω1, ω2 , ..., ωm , where ∑ ω j = 1, (ω j > 0) . The value to be evaluated j =1

n

concerning attribute j for alternative d i is denoted aij ( ≥ 0, ∑ aij = 1) ), where i =1

larger values of the attribute value are more preferable. The value to be evaluated m

concerning alternative d i is then obtained as ∑ ωij aij . The alternative that yields j =1

the largest value is ultimately selected as the best one.

7.2.2 Calculation of Weighting Coefficients for Attributes Using the Pair Comparison Method The pair comparison method is often used to define rationally the weighting coefficients for m attributes (evaluative factors) [1]. Table 7.1 shows an example of a pair comparison matrix for four attributes. Each factor is arranged in a vertical sequence as i ( i = 1, 2, 3, 4 ) and the same factors are deployed horizontally as j ( j = 1, 2, 3, 4 ). The decision-maker assigns levels of importance for the relationships among the set of factors i and j , and these values appear in the table. Table 7.2 shows pair comparison importance values aij between factors i and j . The most suitable value is selected by the decision-maker for each pair comparison. When more fine-grained evaluations are required, intermediate values of the values appearing in the table, such as 2, 4, 6, and 8, can be used. In Table 7.1, aii = 1 and aij = 1 / a ji .

174


The pair comparison method exploits the ease with which human beings can make choices when only two evaluative factors are in play, and avoids the difficulties of selecting the best choice when many evaluative factors must be simultaneously considered. Table 7.1 Example of a pair comparison matrix（case where m = 4 ）(where aii ＝１，

a ji = 1 / aij ) where aii ＝１，

a ji = 1 / aij

Factor１

Factor２

Factor３

Factor４

Factor１

a11

a12

a13

a14

Factor２

a21

a22

a23

a24

Factor３

a31

a32

a33

a34

Factor４

a41

a42

a43

a34

i j

Table 7.2 Pair comparison importance values between factors i and j

Comparison of factor i with factor j

aij

Equally important Slightly more important Rather more important Much more important Extremely more important

１３５７９

When m weighting coefficients are expressed as (ω1, ω2 ,..., ωm ) (= {ω} T) and a pair comparison matrix such as shown in Table 7.1 is denoted A , the weighting coefficients are obtained by solving the following eigenvalue equation, where m is the number of evaluative factors: A{ω} = m{ω}

(7.1)

Matrix A is logically expressed using weighting coefficients (ω1, ω2 ,..., ωm ) as follows:

ω1 / ω2 ⎡ 1 ⎢ω / ω 1 ⎢ 2 1 ⋅ A= ⎢ ⋅ ⎢ ⋅ ⎢ ⋅ ⎢ωm / ω1 ωm / ω2 ⎣

⋅ ⋅

ω1 / ωm ⎤ ⋅ ω2 /ωm ⎥⎥ ⋅

⋅ ⋅

⋅

⋅

1

⎥ ⎥ ⎥ ⎥ ⎦


175

The above matrix A expresses an ideal set of pair comparisons that is fully consistent and without redundancies and eigenvalue λ is thus coincident with m . The magnitudes of the elements on the second and lower lines are larger than the values of the elements in the topmost line by fixed multiplicative values, and the rank of the matrix is 1. Among m eigenvalues, only one eigenvalue has a positive value m , while the others are equal to zero. A practical matrix A which consists of input pair comparisons likely differs from the above ideal case, and eigenvalue λmax is not coincident with m . Whether or not the pair comparison results are sufficiently consistent can be evaluated by considering the magnitude of the consistency index (CI), which expresses the magnitude of the difference between λmax and m , as shown below: CI＝

λmax − m

(7.2)

m −1

When the matrix is completely consistent, the magnitude of CI is 0, and increasing values of CI imply lower consistencies. When CI exceeds 0.1, the pair comparison results should be reconsidered and, if necessary, modified. m

Usually, the obtained weighting coefficients values are normalized as ∑ ω j = 1 j =1

( ω j ≥ 0 ). When the number of evaluative factors is m , the number of pair comparisons that must be conducted is equal to m(m − 1) / 2 . Thus, the weighting coefficients are determined using fairly redundant information, and the quality of the final results will be unaffected by less than ideal consistency. Table 7.3 shows a pair comparison example where four evaluative factors of (1) conveyable weight, (2) operational accuracy, (3) operational efficiency, and (4) product cost are considered for a selection problem for a robot used in a specific environment. Table 7.3 Pair comparison matrix between attributes

i j Conveyable weight Operational accuracy Operational efficiency Product cost

Conveyable weight

Operational accuracy

Operational efficiency

Product cost

1

3

1/3

3

1/3

1

1/3

3

3

3

1

5

1/3

1/3

1/5

1

176


The eigenvalue for the problem is obtained as 4.197. From the eigenvector, the normalized weighting coefficients for the four evaluative factors are （ ω1, ω2 , ω3 , ω4 ) =(0.265, 0.151, 0.508, 0.075). CI for the weighting coefficients is obtained as follows: CI=

λmax − m m −1

=

4.197 - 4 =0.0657 4 -1

Since CI is less than 0.1, the consistency is judged sufficient.

7.2.3 Finding the Best Alternative from Among Several Alternatives Using the Analytic Hierarchy Process (AHP) Method The Analytic Hierarchy Process (AHP) method presented in the 1970s by T.L. Saaty is a method for selecting alternatives so that the final goal is most closely achieved [1]. As shown in Fig. 7.1, the decision-making problem is considered as a hierarchical structure composed of three layers, namely, (1) the final goal, (2) the evaluative criteria, and (3) alternatives.

Level 1 : Final goal

Level 2 : Evaluative criteria

Level 3 : Alternatives Fig. 7.1 Hierarchical structure in the AHP method

For example, in the industrial robot selection problem explained above, where the final goal is to maximize the user’s satisfaction level, the evaluative criteria are the conveyable weight, operational accuracy, operational efficiency, and product cost, and the alternative designs D1 , D2 ,…, D5 are considered. First, weighting coefficients ω A j ( j = 1,2, ..., m) expressing the importance levels of the evaluative criteria for selecting the industrial robot are obtained using the pair comparison method as explained above. Next, the adaptability of each alternative product from the standpoint of each evaluative criterion is defined as an additional weighting coefficient. That is, alternative weighting coefficients ωBi A j are obtained using


177

the pair comparison method so that weighting coefficients of m evaluative criteria for n alternative products can be defined. In the robot selection problem, the adaptability levels are first obtained for each alternative product as weighting coefficients with respect to the conveyable weight. Next, from the standpoint of the accuracy, the adaptability levels are similarly obtained, and these procedures are repeated for the remaining evaluative criteria. Finally, the combined weighting coefficient ωCi for alternative product i is obtained by multiplying weighting coefficients ωB

i Aj

and ω A j as shown in (7.3).

The alternative product having the greatest magnitude of ωCi is selected as the best alternative for achieving the final goal. The AHP method thus consists of hierarchical weighting procedures.

⎡ω B1 A1 ⎢ω ⎢ B2 A1 ⎢ ⋅ ⎢ ⋅ ⎢ ⎣⎢ω Bn A1

ωB1 A2 ω B2 A2 ⋅ ⋅

ω Bn A2

...... ωB1 Am ⎤ ...... ω B2 Am ⎥⎥ ⋅ ⋅ ⎥ ⋅ ⋅ ⎥⎥ ...... ω Bn Am ⎦⎥

⎧ω A1 ⎫ ⎧ωC1 ⎫ ⎪ ⎪ ⎪ ⎪ ⎪ω A2 ⎪ ⎪ωC 2 ⎪ ⎪ ⎪ ⎪ ⎪ ⎨⋅ ⎬ ＝ ⎨⋅ ⎬ ⎪⋅ ⎪ ⎪⋅ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ω Am ⎪ ⎪ωC n ⎪ ⎩ ⎭ ⎩ ⎭

(7.3)

7.2.4 Decision-making Using Subjective Probability Under Uncertain Circumstances Uncertain factors are almost always present in real world situations, and include environmental conditions such as weather (e.g., tomorrow’s weather or seasonal temperature changes) as well as business conditions (e.g., short-term or long-term). Concerning uncertain business conditions, there are three possible cases: economically improving business conditions ( j = 1 ), flat business conditions ( j = 2 ), and economically deteriorating business conditions ( j = 3 ). Here, a selection problem to find the best alternative from among two alternatives d i (i=1 and 2), where a new factory is constructed for i = 1 , and the existing factory is repaired for i = 2 , is considered. When the utility of evaluative factor j for alternative i ( d i ) is uij (d i ) , the expected utility Ei for alternative d i is ex3

pressed as Ei = ∑ p j uij ( d i ) , where p j is a probability value subjectively choj =1

178

7 Decision-making Methods 3

sen that expresses the probability that for evaluative factor j , ∑ p j = 1 holds. In j =1

this method, the probability of occurrence for a given economic condition acts as a kind of weighting coefficient that is multiplied by the utility uij (d i ) , and the alternative with the greatest expected utility value is selected as the best one. A simple example is explained below. 1. For the scenario of constructing a new factory, the decision-maker defines the probability of occurrence and utility for each evaluative factor as shown in Table 7.4. For these data, E1 = 0.3 × 5 + 0.2 × 1 + 0.5 × ( −5) = −0.8 . Table 7.4 Data for construction of a new factory

Uncertain factor j

Probability of occurrence pj

Economically improving business conditions Flat business conditions Economically deteriorating business conditions

Utility u j (d1 )

0.3

5

0.2 0.5

1 ̶5

2. For the scenario of repairing the existing factory; the decision-maker defines the probability of occurrence and utility for each evaluative factor as shown in Table 7.5. For these data, E1 = 0.3 × ( −3) + 0.2 × 2 + 0.5 × 3 = 1 .

Table 7.5 Data for repair of an existing factory

Uncertain factor j

Probability of occurrence pj

Utility u j (d 2 )

Economically improving business conditions

0.3

̶3

Flat business conditions Economically deteriorating business conditions

0.2 0.5

2 3

Based on the above results, the expected utility value of repairing the existing factory is selected as the best solution.


179

7.2.5 Decision-making Considering Personal Preferences of a Decision-maker The personality of a decision-maker can profoundly influence the decisions made. Table 7.6 shows the benefit gij of selecting alternative i for the three uncertain cases of favorable business conditions ( j = 1 ), flat business conditions ( j = 2 ), and deteriorating business conditions ( j = 3 ). Table 7.6 Benefit table ( gij )

Uncertain factor

j

Alternative i

Construction of a new factory Repair of the existing factory

Economically Flat busi- Economically improving business condi- deteriorating ness conditions tions business conditions 4

1

̶4

̶2

2

3

(1) Max-mean criterion The alternative having the maximum mean magnitude of the benefit values for all uncertain cases is selected. In the foregoing example concerning the construction or repair of a factory, the mean of the benefits for all factors for construction of the new factory is 1/3, while the corresponding value for repair of the existing factory is 1, so repair of the existing factory is the obvious choice. This method appears reasonable, but when mean values are strongly affected by unusual values in the benefit table, the final result may be skewed and inappropriate. When the benefit table values are appropriate and decision processes are iterated, this method is reasonable and reliable. (2) Max-min criterion For each alternative, attention is given to cases where the business conditions are least favorable, and the alternative having the greatest utility for this case is selected as the best alternative. That is, the utility value Ui for alternative i under least favorable conditions is first obtained as follows: Ui = min gij j

The alternative i having the greatest utility U i is selected as the best one. The max-min criterion is, as it were, pessimistic. In the foregoing example, the utility of the worst case (unfavorable business conditions) for the construction of a new

180


factory is ̶ 4, while the utility of the worst case (favorable business conditions) for the repair of the existing factory is ̶ 2. Here, the alternative with the greater utility, the repair of the existing factory, is, again, the best choice. (3) Max-max criterion Attention here is given to each alternative in favorable economic conditions and the alternative having the greatest utility is selected as the best alternative. That is, the best utility Ui for alternative i is first obtained as follows: U i = max gij j

The alternative i having the greatest utility U i is then selected as the best one. The max-max criterion is optimistic, and tries to hit a jackpot. In the foregoing example, the utility of the best case (favorable business conditions) for the construction of a new factory is 4, while the utility of the best case (unfavorable business conditions) for the repair of the existing factory is 3. Using this criterion, construction of a new factory is seen as the best choice. (4) Regret minimax criterion Here, the regret level is equal to the difference between the utility obtained from an alternative selected when uncertain environmental conditions occur, and the utility obtained from the alternative actually selected. The worst case for each alternative is considered. The alternative having the lowest regret level for the worst case is selected as the best alternative. That is, the maximum regret level U i for alternative i is first obtained as follows:

U i = max(max g ik − g ij ) j

k

where ( max gik − gij ) corresponds to the regret level for alternative i . k

Alternative i having the minimum value of U i is then selected as the best one. This criterion can be considered pessimistic in nature. In the foregoing example, the worst case for the construction of a new factory is for unfavorable business conditions and has a regret level of 7 [3 ̶ ( ̶ 4)=7], while the worst case for the repair of the existing factory is for rising business conditions and has a regret level of 6 [4 ̶ ( ̶ 2)=6]. The alternative with the lowest regret level, repair of the existing factory, is thus the best choice using this criterion. (5) Intermediate criterion To express an intermediate personal preference between pessimistic and optimistic criteria, the following equation can be used, together with an appropriate weighting coefficient ω :

7.3 Methodologies for Decision-making in Collaborative Circumstances

U i = ω min gij + (1 − ω ) max gij j

181

0 ≤ω ≤1

j

In practical decision-making scenarios, the above criteria may need to be appropriately modified. In almost all cases, however, logical evaluation and examination of optimization problems using the methods detailed above can almost always yield reference data that are useful for making final decisions.

7.3 Methodologies for Decision-making in Collaborative Circumstances Decision-making in collaborative circumstances generally consists of consensus decisions made as a group by the people who are members of it. Group decisionmaking offers high potential for obtaining more preferable solutions due to the contributive actions of each member, and the combination of expert knowledge, technology, and wisdom. However, individual members may attempt to further their own agenda, seek personal benefit of one kind or another, or behave inflexibly, all of which may jeopardize a developing project’s potential for success. Persons whose voices are louder or who have higher professional positions or status may wind up leading the process of reaching collaborative decisions, even though they may not be the most suitable individuals for such roles. In such cases, the potential for negative synergy effects can increase. Therefore, the construction and use of systems that support rational compromises and increase positive synergy effects are advantageous. For example, collaborations will result in better solutions and be more profitable for all if negative personal dynamics, such as unreasonable stubbornness or grandstanding, can be minimized or prevented entirely. Figure 7.2 shows a conceptual diagram for four decision-makers, A, B, C, and D, who take part in group decision-making, to evaluation three attributes, a product’s operational accuracy, operational efficiency, and manufacturing cost [2]. In the following, the most basic decision-making method is explained.

182


Decision maker A

Decision maker B Criteria Operational efficiency

Operational accuracy

Product manufacturing cost

Decision maker C

Decision maker D

Fig. 7.2 Conceptual diagram of group decision-making in a design problem

Each decision-maker i obtains a weighting coefficient ωij for each attribute j by using a pair comparison method. Often, a decision-maker who places great value on a particular attribute is an expert in the field pertaining to that attribute. Next, each decision-maker i defines a satisfaction function Si , and a utility function sij for each attribute j , and the design variables are included in the attributes of sij [3, 4]. The satisfaction function Si of decision-maker i is obtained as follows: n

Si = ∑ ωij sij j =1

(7.4)

where n is the number of attributes. Design variables d k ( k = 1,2,..., N ) are then determined so that the sum of the satisfaction functions for M decision-makers is maximized as follows: M

S g = ∑ Si → maximize i =1

where S g is the group satisfaction level. At this time in the decision-making process, it is important that the satisfaction functions and weighting coefficients for each decision-maker for each attribute be visible, and that this information is shared among all decision-makers. Initially, each decision-maker may tend to set unreasonably strong weighting coefficients

7.3 Methodologies for Decision-making in Collaborative Circumstances

183

and satisfaction functions that are too demanding. In such cases, the group satisfaction level S g will tend to be unsatisfactorily low. To maximize the group satisfaction level S g , each decision-maker should cooperatively consider the following compromise processes: 1. Reconsider the weighting coefficient for each attribute and if necessary, adjust it 2. Reconsider the satisfaction function for each attribute, and if necessary adjust it The horizontal axis in Fig. 7.3 shows a graph of values for attribute zij , similar to ε in Fig. 2.7, The figure conceptually illustrates that the satisfaction level for zij

2

increases from point A to point B when the goal value zij∗1 for zij is adjusted

to zij∗2 , as might be done by a decision-maker to improve the group satisfaction level.

1

B

0.8 0.6

　Improved magnitude of satisfaction level A

0.4

　Compromised magnitude of goal value

0.2 0.0 z*ij1 z*ij2 　Attribute Zij Fig. 7.3 Conceptual diagram of changes in satisfaction level function

A higher priority for the compromise is given to the decision-maker who presents the greater weighting coefficient or the stricter satisfaction function. Figure 7.4 is a conceptual figure showing the step-by-step improvements in the group satisfaction level. By conducting the foregoing procedures 1 and 2, the group satisfaction level can be increased, as shown in Fig. 7.4.

184


0

0

1

2

3

4

5

Number of cooperative compromises Fig. 7.4 Improvement in group satisfaction level by cooperative compromise

Exercises 7.1 Explain the merits of, and difficulties in, group decision-making, and summarize successful group decision-making methodologies. 7.2 Discuss factors that should be considered when constructing decision-making support systems that can be effectively applied to similar decision-making problems.

References 1. Saaty TL (1989) The analytic hierarchy process, McGraw-Hill 2. Yoshimura M, Kondo H (1997) Group decision making in product design and manufacturing. In: Proceedings of 1997 ASME Design Engineering Technical Conferences, Sacramento, California:1 ̶ 7 3. Keeney RL, Raiffa H (1976) Decisions with multiple objectives: preferences and value tradeoffs, Wiley, New York 4. Yoshimura M, Kondo H (1996) Product design based on concurrent processing of design and manufacturing information by utility analysis. International journal of concurrent engineering: Research and applications, 4(4):379 ̶ 388

8 Design Optimization for Creativity and Balance

Effective collaboration in product design and manufacturing is one of the most promising methodologies for obtaining superior product design solutions, as explained in earlier sections. The system design optimization strategies explained in Sect. 6.4 can be applied in creative optimization scenarios where a number of group members work collaboratively. In Sect. 8.1, product design decisionmaking methodologies and procedures based on collaboration are described and contrasted with group decision-making based on game theory. Since product manufacturing depends upon a broad range of industries, the success or failure of such groups of people and corporations has considerable cultural impact. The manufacturing techniques used and the quality of the designs produced are crucially important to the sustainable success of such endeavors, and the cooperation of various entities, despite conflicts, is seen as being more effective in the long run than the maintenance of narrow self interests. In Sect. 8.2, the interrelated nature of environmental and cultural impacts of product design and manufacturing are discussed, in terms of the need for increasingly sophisticated and practical product design optimization methods and strategies.

8.1 Creativity Optimization Based on Collaborative Effort Product design optimizations generally require multidisciplinary optimization, and successful results depend on a variety of high-level technologies and knowledge. To achieve the best possible results, design engineers who individually have specific expert knowledge and access to advanced technologies should work collaboratively to complete the design process. In Sect. 3.4, the usefulness of the collaboration concept and its role in the technology of innovative product manufacturing were explained, and the importance of selecting appropriate collaboration partners was discussed. Section 4.4 discussed criteria for the optimum selection of the hu-

186


man resources who will work together on product manufacturing projects. In activities based on collaboration concepts, the knowledge and technologies that each individual possesses can be synergistically combined, which increases the likelihood of an increase in the profit level of the design process result. In Sect. 5.1.6, the technologies used to gather information concerning customer needs for product designs intended for sale were explained. Market conditions and product design requirements often change rapidly, but there is an overall trend toward the use of increasingly sophisticated design solutions. The highest goal of product design optimization is not to obtain a single optimum solution that is fixed in time, but to provide a basis for effective optimum solutions that can be evolved further, to cope best with complex market demands over time. Analogously, the goal of system design optimization should not be limited to obtaining the global optimum solution, but rather to enable breaking through an obtained global optimum solution so that more preferable design solutions can be selected according to dynamic requirements. In Section 6.4，system design optimization strategies to conduct product design optimizations effectively were explained, namely, the use of simplified characteristics, priority relationships, and conflicting relationships among simplified characteristics. The methodologies and procedures based on hierarchical optimizations are also useful in collaborative circumstances where members of a working group can actively combine their knowledge and expertise. In general product design optimizations, related characteristics are included in the formulation of the objective function and the optimization constraints and the formulated problems are then solved using mathematical optimization methods. In the conceptual diagram shown in Fig. 8.1, these original characteristics appear as surface level characteristics at the highest hierarchical level. However, as explained in Sect. 6.4, obtaining a true globally optimum design solution is aided by developing a deeper understanding of the design problem being regarded, by grasping the characteristics of the design problem at theirs and analyzing their interrelationships [1, 2]. The system structure of a hierarchical optimization procedure is constructed by breaking down each original, highest-level characteristic located at the surface level into simpler characteristics and/or by extracting simpler characteristics from each of these characteristics. The breakdown and extraction procedures are conducted until the simplest fundamental characteristics are obtained, and these will be located at the lowest level of the hierarchy. The conflicting relationships among these simplified characteristics and the order, or priority, of their optimization are then examined. Simplified characteristics that have conflicting relationships and are located in the same lower hierarchical level are then simultaneously optimized as a multiobjective optimization problem. The optimum solutions for original surface characteristics located at the top hierarchical level usually form a Pareto optimum solution line or higher dimensional surface, depending on the number of optimized objectives. Such solution lines or surfaces are frontiers that also act as a

8.1 Creativity Optimization Based on Collaborative Effort

187

kind of barrier for improvements in the performance characteristics, a barrier that ideally should be broken through. To break through an existing optimum solution frontier, pairs of characteristics located in deeper hierarchical levels that have conflicting interrelationships that dominate surface level characteristics should be focused on. Characteristics that are located at the bottom or lower levels of a hierarchical optimization model are generally simple and the interrelationships of these simple characteristics are very clear. It is therefore easier to find effective breakthrough alternative design ideas that will form a new Pareto optimum solution line or surface by working with characteristics at the lower possible hierarchical level. Experts in the collaborative group can consequently develop effective design improvements more easily by focusing on lower level, simplified characteristics.

Surface level characteristics

Deeper level characteristics

Fig. 8.1 Surface level characteristics and deeper level characteristics

A collaboration project for effectively conducting system design optimizations consists of a leader with wide knowledge and good organizational skills, and a number of experts each of whom has specific expert knowledge in one or more fields useful to the project. Customers for the product being designed make purchasing decisions based on two or three original characteristics located at the highest level of the hierarchical design optimization system, here called surface characteristics. As shown in Fig. 8.2, an initial Pareto optimum solution line is obtained, using the technologies and knowledge of the collaboration members, by solving the above hierarchical optimization problem. The results are then displayed in the objective function space and the profit that this first design solution is expected to deliver is estimated. To increase the profit, the leader and the collaborative mem-

188


bers discuss alternatives and determine a preliminary improved product design goal point in the objective function space. The expected additional profit increases the incentive of the collaboration members to pursue their work actively. This process is repeated until the most effective product design is obtained. In such collaboration scenarios, each member of the group can maintain a feeling of contributing to the increased profit achieved by the group as a whole. Here, each collaborative individual can expect a reward that is proportional to his or her level of contribution to the collaborative project and this expectation naturally motivates each member to give wholeheartedly their best to the job at hand. Design solution without collaborative effort

P Initial Pareto optimum solution set for collaborative effort

×

f2

Initial design solution by the collaborative work More preferable direction

Q

Goal 1 Goal 2 Goal 3

f1 Fig. 8.2 Evolutional processes of design solutions based on collaborative effort

Game theory [3, 4] has been widely discussed as a theory that can provide useful support in group decision-making scenarios during a collaborative project. Here, the methodologies and logical procedures based on the collaborative concepts discussed above are collectively termed collaboration theory [1, 2], and will be contrasted with game theory methodologies. Figure 8.3 shows a comparison of product design activities based on both game theory and collaboration theory. The size of the circles corresponds to the amount of expected profit and the gray areas indicate the expected profit prior to the design activity carried out by the group. In the design activities based on basic game theory, each active member considers strategies that will maximize personal gain, and harbors suspicious feelings toward other members of the group. Such scenarios are likely to lead to a reduction in the total profit that can be expected from the result of the design process.

8.2 Cultural Impact of Product Manufacturing

189

Expansion of profit Reduction of profit

Activities based on game theory

Activities based on collaboration theory

Fig. 8.3 Comparison of design activities based on game theory and collaboration theory

On the other hand, in activities based on collaboration theory, the knowledge and technologies that each individual possesses can be synergistically combined, which increases the probability that the design result will yield higher profit levels. In such scenarios, each member of the group can maintain a feeling of contributing to the increased profit achieved by the group as a whole. Here, each collaborating person can expect a reward that is proportional to his or her level of contribution to the project and this expectation naturally motivates each member to give wholeheartedly their best, without the strain of interpersonal competition, as explained above.

8.2 Cultural Impact of Product Manufacturing As explained in Chaps. 1 and 2, product manufacturing has evolved and become more sophisticated so that a larger number of factors are simultaneously evaluated. That is, the concerns that product manufacturing must now address are no longer limited to creating products that simply satisfy people’s physical or functional requirements. Current product manufacturing must also include the improvement of customers’ mental or cultural satisfaction levels by considering the aesthetic qualities and ergonomics of products, as explained in Chap. 4. Ideally, an important goal of product manufacturing is to design and manufacture attractive products that harmonize with the environment in which they are used and minimize negative impacts on climate, natural resources, and local culture [5, 6], in addition to satisfying personal preferences and tastes. This, and other goals, can be achieved by systematically considering a range of evaluative factors. Many industries are

190


starting to realize that their long-term success depends on addressing factors beyond the design of products that merely satisfy minimum, isolated, short-term requirements. As shown in Fig. 8.4, for industries truly to flourish, product manufacturers must be aware of the cultural impact of their products, and work to achieve balanced approaches that deal with broader issues pertaining to natural environments, climate, and the personality of the people who purchase and use their products, so that customer satisfaction can be truly maximized. Skillful diversification in product manufacturing can increase the personal satisfaction of customers, and drive the creation of new products that cope better with a variety of local environments. The application of optimization techniques to product designs is important not simply from the standpoint of obtaining a single superior design solution, but because such techniques can provide a useful variety of design solutions. Using this variety, the most appropriate global solution for particular needs can be selected from a number of alternative solutions, according to the detailed requirements pertaining to specific products in specific locations and times. Thus, optimization techniques can potentially play important roles in both creating products that deliver greater satisfaction levels, and in manufacturing products that achieve greater harmony with their surroundings, by considering a broader range of factors.

Flourishing of industries

Creativity and balance in manufacturing

Cultural impact Fig. 8.4 Conceptual diagram of the relationship between creativity and balance in manufacturing, and cultural impact and flourishing of industries

Exercises 8.1 Explain the role of collaboration for obtaining design ideas that break through present levels of product performance and manufacturing cost.

References

191

8.2 Discuss the features of product manufacturing in terms of both its beneficial and destructive cultural impact.

References 1. Yoshimura M, Kikuchi S, Kizu M, Saitou Y (2009) Optimum system design of machine products based on collaboration theory. In: Proceedings of the 8th World Congress on Structural and Multidisciplinary Optimization, Portugal 2. Yoshimura M, Kikuchi S (2009) Optimization of machine product designs from deeper level characteristics using collaboration theory concepts. In: Proceedings of ASME 2009 Design Engineering Technical Conferences and Computers and Information in Engineering Conference, DETC2009-86928 3. Myerson RB (1991) Game theory – analysis of conflict. Harvard University Press, Cambridge 4. Osborne MJ and Rubinstein A (1994) A course in game theory. The MIT Press, Cambridge 5. Watsuji T (1998) Climate and culture – a philosophical study. Translated by Bownas G. Yushodo, Tokyo 6. Nisbett RE (2003) The geography of thought – how Asians and westerners think differently and why. Free Press, New York

Index

A Accuracy, 146 Active, 121 Adaptation processes, 129 Adjective word pairs, 63 Adjustment factors, 92 Adjustment values, 92 Adjustment variable, 138 Aesthetic characteristics, 21 Aesthetic factors, 58, 60 Agents, 113 Alternative design candidates, 162 Alternatives, 162, 163, 164 Analytic Hierarchy Process (AHP), 176 Analytical Target Cascading (ATC), 141 Approach, 88 Assembly, 95 Assembly difficulties, 95 Attributes, 23 Availability, 19

B Barrier, 187 Benefit levels, 51 Bi-Level Integrated System Synthesis method, 140 Binary numbers, 129 Bottlenecks, 14, 91 Bounded objective function method, 125 Bounding operations, 127 Box-type member, 132 Brainstormings, 85 Branch-and-bound method, 105, 126, 128 Branching, 128 Branching operations, 127, 128 Breakdown, 141, 186 Breakthrough alternative design, 187 Breakthrough design solutions, 47 Breakthrough designs, 35

Breakthroughs, 32

C CALS paradigm (Continuous Acquisition and Life-cycle Support), 6 CAM information, 81 Candidate partners, 50, 52 Career path, 74, 76 Career satisfaction, 80 Casting, 100 CAT (Computer-Aided Testing), 81 Cell production, 17 Chain reaction process, 85 Chromosome, 129 Climate, 189 Cluster analysis, 107 Coding, 129, 163 Coding systems, 96 Collaboration, 36, 46, 47, 50, 55, 76 Collaboration partner, 50, 52 Collaboration project, 187 Collaboration theory, 188, 189 Collaborations, 112 Collaborative circumstances, 181 Collaborative decision-making, 109 Collaborative decisions, 181 Collaborative design, 86 Collaborative design process, 88 Collaborative efforts, 55 Collaborative Optimization (CO), 140 Combinatorial optimization problems, 126 Commerce At Light Speed (CALS), 6 Common databases, 111 Common parts, 112 Component arrangement design division, 143 Compromise, 181 Computer networks, 44, 50 Computer simulation, 83 Computer-Aided Design (CAD), 37

194 Computer-Aided Engineering (CAE), 83 Computer-Aided Manufacturing (CAM), 37 Computer-Aided Planning (CAP), 37 Computer-Aided Process Planning (CAPP), 37 Computer-aided technologies, 37 Computer-Aided Testing (CAT), 37 Computer-Integrated Manufacturing (CIM), 38, 41 Concave shape, 125 Concavity, 125 Conceptual design stages, 35 Conceptual product design, 37 Concurrent engineering, 15, 36, 37, 40, 81,, 90, 96,106,112 Concurrent evaluations, 66 Concurrent optimization, 40 Conflicting interrelationships, 146 Conflicting relationship, 148, 186 Consensus decisions, 181 Consistency index, 175 Constraint functions, 26 Consumer products, 4 Contact conditions, 152 Continuous design variables, 126 Contour lines, 119 Conveyor production, 15 Cooperative collaboration, 19 Cooperative project, 49 Creative group activity, 86 Creative process, 88 Criteria, 9, 26 Criteria requirements, 22 Crossover, 129 Crossover operation, 129, 166 CSCW (Computer Support Cooperative Work), 50 Cultural impact, 189 Customer desires, 6 Customer needs, 102, 107 Customer requirements, 11 Customer satisfaction, 107, 109, 190 Customer satisfaction levels, 58, 59 Customer-maker collaborative manufacturing, 6 Customers, 59, 68 Cutthroat competition, 55

D Damping coefficients, 148, 152, 158, 161

Index Damping effects, 151, 152, 153 Damping ratio, 150, 152, 153, 155 Data mining, 111 Decision making support system, 172 Decision support system (DSS), 111 Decision supporting system, 172 Decomposition, 127 Deeper understanding of the design problem, 186 Deepest levels, 186 Delay in production, 12 Delivery times, 11 Demand, 12 Demand and supply curves, 12 Demand curve, 12 Demand leading paradigm, 14 Design and manufacturing information, 81 Design for assembly, 41 Design for distribution, 41 Design for environment, 41 Design for lifecycle, 41 Design for maintenance, 41 Design for manufacturing, 41 Design for quality, 41 Design for reliability, 41 Design qualities, 10 Design Structure Matrix (DSM), 134 Design Structure System (DSS), 134 Design variable space, 27, 94 Design variables, 26, 27 Destruction of nature, 55 Die casting, 101 Digital engineering, 81 Digital-mockups, 82 Dimensional constraints, 27 Dimensional variances, 10 Disassembly, 96 Discrete design variable optimization, 126 Discrete design variables, 126, 128, 162 Dissemination of proprietary technologies, 47 Dynamic characteristics, 149 Dynamic rigidity, 122, 123, 150, 151

E Economically deteriorating business conditions, 177 Economically improving business conditions, 177 Economies of scale, 6 Efficiency, 146

Index Eigenvalue equation, 174 Energy-efficient products, 59 Enterprise management decision-making, 111 Enterprise Resource Planning (ERP), 110 Enumeration method, 126, 128 Environmental impact, 21, 59 Epsilon-constraint method, 125 Equality constraint, 27 Ergonomic factors, 59 Ergonomics, 60, 69, 70 Evaluation map, 64, 65, 66 Evaluative criteria, 172, 176 Evolutionary mechanisms, 129 Excess inventory, 13 Existing technologies, 47 Expected utility, 177 Expert abilities, 74 Experts, 46 Extraction of simpler characteristics, 146 Extraction procedures, 186

F Failure rate, 18 Feasible design region, 27 Feasible design variable space, 118, 119 Feasible direction method, 161 Finite element method (FEM), 83 Five physical senses, 60 Flat business conditions, 177 Flexible joint, 157 Flexible manufacturing, 15 Flexible Manufacturing Cell (FMC), 15, 104, 127 Flexible Manufacturing System (FMS), 15 Ford paradigm, 5 Forging, 100, 101 Forward problems, 27 Functional quality, 108

G Game theory, 188 Gantt chart, 105 Gene, 166 Gene position, 129, 166 General rules, 149 Generative approach, 102 Genetic algorithms (GAs), 129, 162 Genetic types, 129 Global characteristics, 148

195 Global optimal solution, 40 Global optimization, 29, 134 Global optimum solution, 32, 122, 130, 131, 149, 162, 186 Globally optimum, 46 Globally optimum solution, 118, 119, 122, 128, 146, 149 Grinding, 100, 157 Group analysis, 108 Group decision-making, 181 Group satisfaction level, 182, 183 Group technology (GT), 96 Groupware, 50

H Hierarchical Genetic algorithms (HGAs), 162 Hierarchical genotype, 167 Hierarchical optimization methodologies, 162 Hierarchical optimization procedures, 152, 153, 162, 186 Hierarchical optimization strategies, 154 Hierarchical structural model, 166 Hierarchical structure, 141, 176 Hierarchical weighting procedures, 177 Histogram, 17 Human abilities, 58 Human body, 69 Human judgment, 58 Human logic, 57 Human relationships, 75, 76 Human resource (HR), 76 Human scale, 3

I Ideal consistency, 175 Ideal point, 124 Idealized models, 131 Implementation, 134 Implementation phase, 132 Importance levels, 176 Improvement procedures, 28 Incentive, 188 Industrial machines, 3 Industrial robot, 127, 146 Inequality constraint, 27 Information and material flow, 38 Inherent abilities, 58 Initial design variables, 118, 144, 157, 158

196 Injection molding machine, 146 Injection-molding, 146 Innovation, 35 Inspiration, 85 Integrated preference function, 68 Intercommunication, 86 Intermediate personal preference, 180 Inventory, 13 Inventory control, 12 Inventory levels, 13 Inventory shortages, 13 Inverse problems, 27

J Job shop manufacturing, 15 Job shop paradigm, 101 Job shop production, 11, 12, 17, 96 Job shop type of production, 6 Job shop type production, 130 Joint contact surfaces, 148 Joint damping coefficient, 152 Joint design variables, 152 Joint rigidity, 152 Joints, 152, 153 Just in time (JIT), 6, 13

K Kanban, 14 Kansei, 60 Kansei attributes, 67 Kansei characteristics, 21 Kansei engineering, 60 Kansei evaluation, 61 Kansei feelings, 63 Kansei impressions, 64 Kansei preference levels, 66 Karush-Kuhn-Tucker condition (KKT condition), 120 KJ-method, 85 Knowledge sharing, 48, 50

L Laborious hand honing, 100 Lagrangian function, 121 Lagrangian multiplier, 121 Lagrangian multiplier coefficient, 121 Lead time, 11 Lean production, 6 LifeCycle Assessment (LCA), 21

Index LifeCycle Costing (LCC), 20 Lifecycle design, 41 Lifestyles, 21 Line production, 15 Linear optimization problems, 27 Linear planning problems, 27 Linear programming, 118 Linear programming methods, 118 Link-structure, 88 Local characteristics, 148 Local optimality, 120 Local optimum solution, 118, 119, 120, 134, 149, 152 Locally optimum solution, 118 Long-term economic benefits, 20 Loss, 92 Loss function, 92 Lot size, 14 Lower bound, 127 Lowest level of the hierarchy, 186

M Machine elements, 147 Machine product, 148 Machine product design optimizations, 149 Machine tool, 127, 146, 155 Machining, 100 Machining accuracy, 99, 100, 148 Machining cost, 98, 148, 153, 157 Maintainability, 20 Maintenance, 45 Makespan, 104, 127 Making products to order, 6 Management information system (MIS), 111 Management of the enterprise, 110 Manipulator, 61 Manufacturing cost, 10, 11, 81, 92, 95, 96,100, 103, 104 Manufacturing flexibility, 15 Manufacturing losses, 14 Manufacturing paradigm, 59, 96, 97 Manufacturing qualities, 10 Market demands, 186 Market research, 4 Market share, 31 Marketplace competition, 32 Mass production, 6, 14, 17, 20 Mass production paradigm, 97 Mass-produced items, 6 Material cost, 147, 155

Index Material Requirements Planning (MRP), 104 Mathematical programming methods, 117 Matrix tables, 108 Maximum receptance value, 150 Max-max criterion, 180 Max-mean criterion, 179 Max-min criterion, 179 Mean time between failures (MTBF), 20 Mean time to failure (MTTF), 20 Mean time to repair (MTTR), 20 Mental abilities, 57 Mental and emotional satisfaction, 60 Mental factors, 21, 58 Milling, 157 Mobile agents, 113 Modal flexibility, 151 Mode shape, 152 Modular parts, 102 Module technology, 102, 103 Modules, 102 Mold clamping unit, 83 Mold press, 101 Monotonous relationship, 122 Motions of the limbs, 71 Motivation, 60 Motivational objective function, 76 Multi-attribute utility functions, 26 Multidisciplinary collaboration, 46 Multi-disciplinary engineering divisions, 140 Multidisciplinary optimization, 29, 185 Multidisciplinary two-stage multiobjective optimization method, 141 Multiobjective optimization, 130 Multiobjective optimization methods, 26, 153 Multiobjective optimization problem, 30, 43, 123, 124, 145, 153, 157, 186 Multiple point searching procedures, 130 Multi-step decision-making, 153 Muscular forces, 71 Mutation, 129 Mutation operation, 129, 166

N Natural environments, 7, 20, 59, 190 Natural frequency, 150 Natural resources, 22, 55 NC (Numerical Control), 15 NC programs, 97

197 Negative impacts, 189 Negative synergy effects, 181 Networked database, 45 Networked information systems, 112 Networked information technology, 111 Networked systems, 50 Networking technologies, 47 Non-inferior solution set, 27 Nonlinear optimization problem, 27, 119 Nonlinear planning problems, 27 Nonlinear programming problems, 118 Non-monotonous relationship, 123 Normal distribution, 18 Notational expressions, 165

O Objective function, 26, 118 Objective function space, 27, 31 One-way sequential decision-making processes, 39 Operational accuracy, 49, 146 Operational efficiency, 49 Optimal allocation, 74 Optimization methods, 26 Optimization problem, 26, 71, 148 Optimization techniques, 117 Optimum allocation, 75 Optimum solution frontier, 187 Outward appearance, 61

P Pair comparison, 172 Pair comparison importance, 173 Pair comparison matrix, 173 Pair comparison method, 173, 182 Parameter design method, 93 Pareto, 31 Pareto optimum design solutions, 154 Pareto optimum solution line, 31, 100 Pareto optimum solution set, 31, 124, 154, 157, 161 Pareto optimum solutions, 66, 101, 125, 154 Part classification methods, 96 Part transportation operations, 128 Partial optimization, 29 Penalty function method, 119, 120 Penalty term, 120 Personality, 179, 190 Phenotypes, 129

198 Physical abilities, 58 Physical effort, 71 Physiological energy, 71 Plant maintenance, 45 Platform, 102 Popularity, 66 Positive synergy effects, 181 Potential customers, 59 Powder metallurgy, 101 Preferences, 23 Primary company, 52 Principal component analysis, 64 Priority relationships, 134, 148 Priority rules, 149 Probability analysis, 93 Problem-solving approaches, 85 Process capability, 17 Process planning, 98 Processing accuracy, 10 Processing order, 135 Processing time, 11 Product cost, 10 Product Data Management (PDM), 90 Product delivery times, 11 Product design, 4 Product design criteria, 23 Product design optimization, 185, 186 Product design requirements, 186 Product development, 4 Product development competition, 22 Product diversification, 103 Product Ideas, 85 Product lead time, 104 Product lifecycles, 12, 21 Product manufacturing, 4 Product manufacturing cost, 10, 30, 93, 146, 148 Product manufacturing innovation, 57 Product manufacturing paradigm, 5, 7, 47 Product performance, 10, 11, 30, 97, 130 Product quality, 10, 108 Product supply curve, 12 Product usability, 69 Product’s lifecycle, 41 Production mistakes, 11 production paradigm, 55 Production scheduling, 104 Production time, 104 Production to order, 13, 96 Production to stock, 13 Productivity, 96 Profit, 186

Index Program (Project) Evaluation and Review Technique (PERT), 106 Project, 74 Prototype parts, 82 Pull manufacturing, 14 Pull manufacturing methods, 14 Punch press, 146 Push manufacturing methods, 14

Q Quadratic optimization problems, 119 Qualitative evaluations, 75 Quality engineering, 18, 92 Quality function deployment (QFD), 108 Quality loss factors, 92 Quantitative attributes, 59 Quantitative expressions, 75 Quantitatively evaluating subjective, 58 Questionnaires, 107

R Rapid prototyping, 82 Receptance frequency response, 150 Recycle, 21 Recycling of products, 20 Reduce, 21 Reference data, 181 Refinement process, 85 Regret level, 180 Regret minimax criterion, 180 Relative displacement, 149 Reliability, 19, 20 Reliability evaluations, 19 Response action, 86 Response parameter, 120 Response surface method, 118 Reuse, 21 Reward, 188 Robust design optimization method, 93 Robust designs, 93

S Safety, 19 Safety evaluations, 19 Sale of the product, 4 Satellite, 141 Satisfaction function, 67, 182 Satisfaction levels, 11, 21, 23, 58, 67, 107, 130, 188

Index Scheduling problems, 106 Scheduling technology, 104, 127 Semantic differential (SD) method, 63 Sensibility, 58 Sequential Linear Programming (SLP) method, 119 Sequential Quadratic Programming (SQP), 119 Sequential Unconstrained Minimization Technique (SUMT), 120 Session, 86 Session fork, 87 Session method, 88 Shapes of the joints, 152 Side-constraints, 27 Simplex method, 118 Simplification, 131, 132, 133, 134, 146, 148, 162 Simplification of characteristics, 146 Simplified characteristics, 152 Simplified model, 132, 157 Simplified shapes, 132 Simultaneous engineering, 40 Single point searching procedures, 130 Single-person standing-style production, 17 Six sigma campaign, 18 Skill abilities, 75 SN ratios, 92 Socially useful benefits, 55 Software agents, 113 Solid models (3-D models), 81 Specific rules, 149 Spring stiffnesses, 148 Stability of the satellite, 143 Standard deviation, 18 Standards of living, 1 State variables, 27 Static displacements, 147 Static rigidity, 121, 147 Strategic Information System (SIS), 111 Stress distributions, 149 Structural analysis, 83 Structural design division, 142 Structural joint rigidity, 152 Structural joint rigidity optimization, 157 Structural member rigidity, 152, 157 Structural member rigidity optimization, 157 Structural member weight, 153 Structural members, 132, 147, 152 Structural weight, 121, 157 Sub-Optimization Problem (SOP), 153

199 Substructures, 163 Sub-system design variables, 139 Super finishing, 100, 157 Super polishing, 100 Supply chain management (SCM), 14, 91 Supply curve, 11 Supporting agent systems, 113 Surface characteristics, 186, 187 Surface level, 186 Surface level characteristics, 186, 187 Surface roughness, 10, 157 Surface stress distributions, 150 Synergy effects, 85 System design optimization strategies, 145 System division, 141 System division optimization, 141 System level design variables, 139 System optimization procedures, 162

T Taguchi method, 92 Taylor paradigm, 5 Terminate, 127 Theory of constraints (TOC), 91 Three-Dimensional CAD Technology, 80 Tolerance, 17 Top hierarchical level, 154 Total manufacturing cost, 148, 157 Total processing time (makespan), 104 Toyota manufacturing method, 14 Traction method, 83 Tree structure, 87 Trial experiments, 19 Trial manufacturing, 82 Trigger action, 86, 87 TRIZ(Theory of Innovative Problem Solving), 90 Two stage optimization, 138 Two-level optimizations, 140 Type I quantification theory, 68

U Ubiquitous, 113 Unbalanced collaborations, 53 Uncertain business conditions, 177 Uncertain environmental conditions, 180 Uncertain factors, 172, 177 Unreliability, 19 Unsold inventory, 12 Upper level optimization, 154

200 Upper level substructures, 165 Utility, 26, 177 Utility analysis, 23 Utility function, 182 Utopia point, 124

V Value analysis; VA, 32 Value criterion, 32 Value engineering, 32 Variant approach, 102 Variations, 17, 92 Virtual corporations, 47

Index Virtual enterprises, 50 Virtual reality (VR), 79, 111

W Weighted sum method, 124 Weighting coefficient, 124, 176, 177, 180 Weighting method, 124 Welded structure, 43 Welding, 100 Welding cost, 43, 44 Word pair, 63 Worker discomfort, 70

System Design Optimization for Product Manufacturing

Multi-objective Evolutionary Optimisation for Product Design and Manufacturing

System Synthesis: Product and Process Design

Introduction to Product Service-System Design

Engineering Design Methods: Strategies for Product Design

Engineering Design Methods: Strategies for Product Design

System Verilog For Design

Flexible Manufacturing System

Manufacturing Execution System - MES

Flexible Manufacturing System

Radio System Design for Telecommunications

System Design for Telecommunication Gateways

System Design for Telecommunication Gateways

Collaborative design and planning for digital manufacturing

Standards for Engineering Design and Manufacturing

Materials, Design and Manufacturing for Lightweight Vehicles

Design Reuse in Product Development Modeling, Analysis and Optimization

Product Design for Manufacture and Assembly

Understanding Product Design For Injection Molding

Standards for Engineering Design and Manufacturing

Plastics Engineered Product Design

Plastics Engineered Product Design

Product Design for Manufacture and Assembly

Integrating Design and Manufacturing for Competitive Advantage

Collaborative Product Design and Manufacturing Methodologies and Applications (Springer Series in Advanced Manufacturing)

Collaborative Product Design and Manufacturing Methodologies and Applications (Springer Series in Advanced Manufacturing)

Composites Manufacturing - Materials, Product, and Process Engineering

Composites manufacturing : materials, product, and process engineering

Sound Systems: Design and Optimization: Modern Techniques and Tools for Sound System Design and Alignment

Geometric programming for design and cost optimization

Sound Systems: Design and Optimization: Modern Techniques and Tools for Sound System Design and Alignment

System Design Optimization for Product Manufacturing

Multi-objective Evolutionary Optimisation for Product Design and Manufacturing

System Synthesis: Product and Process Design

Introduction to Product Service-System Design

Engineering Design Methods: Strategies for Product Design

Engineering Design Methods: Strategies for Product Design

System Verilog For Design

Flexible Manufacturing System

Manufacturing Execution System - MES

Flexible Manufacturing System

Radio System Design for Telecommunications

System Design for Telecommunication Gateways

System Design for Telecommunication Gateways

Collaborative design and planning for digital manufacturing

Standards for Engineering Design and Manufacturing

Materials, Design and Manufacturing for Lightweight Vehicles

Design Reuse in Product Development Modeling, Analysis and Optimization

Product Design for Manufacture and Assembly

Understanding Product Design For Injection Molding

Standards for Engineering Design and Manufacturing

Plastics Engineered Product Design

Plastics Engineered Product Design

Product Design for Manufacture and Assembly

Integrating Design and Manufacturing for Competitive Advantage

Collaborative Product Design and Manufacturing Methodologies and Applications (Springer Series in Advanced Manufacturing)

Collaborative Product Design and Manufacturing Methodologies and Applications (Springer Series in Advanced Manufacturing)

Composites Manufacturing - Materials, Product, and Process Engineering

Composites manufacturing : materials, product, and process engineering

Sound Systems: Design and Optimization: Modern Techniques and Tools for Sound System Design and Alignment

Geometric programming for design and cost optimization

Sound Systems: Design and Optimization: Modern Techniques and Tools for Sound System Design and Alignment

Recommend Documents