THESIS FOR THE DEGREE OF LICENTIATE OF ENGINEERING
Increased Visual Quality through Robust Split-Line Design
ANDREAS DAGMAN
Department of Product and Production Development CHALMERS UNIVERSITY OF TECHNOLOGY Göteborg, Sweden 2005
Increased Visual Quality through Robust Split-Line Design Andreas Dagman Copyright @ Andreas Dagman Research Series from Chalmers University of Technology Department of Product and Production Development ISSN 1652-9243 Report No. 4 Published and Distributed by Chalmers University of Technology Department of Product and Production Development Division of Product Development SE – 412 96 Göteborg, Sweden Printed in Sweden by Chalmers Reproservice Göteborg, 2005
ABSTRACT
In the early phases of the product development process, the initial product geometry can be a non-divided shape, just a shell. This initial product geometry has to be divided into individual parts in order to manufacture and fulfil the functionality of the product. In the case of automobiles, these parts can be fenders, doors, hoods, etc. The design of the parts is influenced by several aspects, like design language, geometrical dimensioning, crash safety, wind resistance, etc. The decisions regarding how to divide and design the parts are made in multidisciplinary teams. The final appearance of the spatial relations between the parts, i.e. the split-lines, in an assembly affects the quality appearance of the product. This in turn influences the sales of the product. By supporting the product development process with methods, tools and guidelines, the quality of the final solution will improve. The focus of the presented research has been on how to split the initial product geometry in order to achieve a geometrically and visually robust solution. In other terms, it has been on how the split-line shall be designed and placed in order to achieve as geometrically and visually robust solution as possible. Geometrical robustness aims at the dimensional and functional aspects of the product. Visual robustness is more concerned with the aesthetic aspects of the product. A geometrically and visually robust solution suppresses part and assembly variation. This leads to a decreased geometrical output variation, the variation between parts the customer sees. A geometrically robust solution often results in a visually robust solution. A geometrically sensitive solution, on the other hand, can be made visually acceptable in some cases by controlling the position and the direction of the splitline. Different areas are more or less sensitive for the eye. The research results has contributed to an enhanced knowledge and a supporting tool in the area of geometrical and visual robustness. The tool supports the division of initial product geometry with respect to geometrical robustness. The enhanced knowledge enlightens how different assembly aspects influence the final result when assembling. There is also a contribution concerning customers’ and the industry’s attitude regarding the visual quality appearance of split-lines. It concerns more the aesthetic aspects of the split-line design. The conclusion of the presented research project is that the division of the initial product geometry, in order to achieve a geometrically and visually robust solution, is influenced by many aspects. These aspects can sometimes be antagonists and there have to be a compromise between them. Also, the tool presented does calculate and visualize robust areas between two geometries to be split and this result can hopefully be used as a support in the product development process. The customer’s consider high visual quality of splitlines as important and that narrow split-lines is of importance for the perceived quality but there is other factors that is influencing the quality appearance of split-lines. By using the presented tool and by using the knowledge, gained from this research project, the chances of creating products with high visual quality will increase with decreased cost and time for the product development process. Keywords: Tolerance Analysis, Split-Line, Quality, Visual Quality Appearance, Simulation.
i
ACKNOWLEDGEMENTS
The research and work behind this thesis has been carried out at the Department of Product and Production Development at Chalmers University of Technology, Sweden. In order to carry out something as comprehensive as a licentiate thesis, several people have to be involved, contributing in their own way. I would first of all like to direct a big thanks to my supervisor Professor Rikard Söderberg for his support and unique ability to encourage. He introduced me to the area of tolerancing and robust design and made it interesting. In addition, a special thanks goes to Associate Professor Lars Lindkvist, who has helped me with computer programming and thoughtful comments during the research process. I would also like to direct many thanks to my colleagues at the Department of Product and Production Development. You always have time for a question or two and have encouraged me when struggling with problems. The Swedish Foundation for Strategic Research has supported this work financially through the Swedish Engineering Design Research and Education Agenda (ENDREA) and the ProViking Graduate School. I gratefully acknowledge their support. A big hug goes to my family, who is always there for me and adds value to my life. Thanks. Last but not least I would like to thank my loving fiancée and colleague Jessica Isaksson for all her support and valuable critique. Keep on fighting with “skruttis”!
Andreas Dagman Göteborg 2005
iii
PUBLICATIONS
The following papers are included in this licentiate thesis. Paper A Dagman, A. and Söderberg, R., (2003), Virtual Verification of Split-Lines with Given References. Paper presented at the 14th International Conference in Engineering Design, ICED03, Stockholm, Sweden, 18-20 August 2003. Paper B Dagman, A., Wickman, C. and Söderberg, R., (2004), A Study of Customers' and the Automotive Industry's Attitude Regarding Visual Quality Appearance of Split-Lines. Presented at Advances in Engineering Design, AED2004, Glasgow, Scotland, 5-8 September 2004. Paper C Dagman, A., Söderberg R. and Lindkvist L., (2004), Split-Line Design for Given Geometry and Location Systems, submitted to the Journal of Engineering Design.
Distribution of work In paper A, Söderberg initiated the idea. Dagman performed the research and carried out the writing of the paper, with the support of Söderberg as a reviewer. In paper B, Dagman and Wickman made most of the planning with support from Söderberg. The interviews were conducted equally between the three researchers. Dagman and Wickman analysed the results and wrote the paper together. In paper C, Söderberg initiated the idea. Dagman performed the research in cooperation with Söderberg and Lindkvist. Dagman wrote the paper with the support of Söderberg and Lindkvist as reviewers.
v
TABLE OF CONTENTS 1
INTRODUCTION................................................................................................................................. 1 1.1 1.2 1.3 1.4 1.5
2
RESEARCH APPROACH ................................................................................................................... 5 2.1
3
THE RESEARCH SURROUNDINGS ..................................................................................................... 9 ENGINEERING DESIGN..................................................................................................................... 9 INDUSTRIAL DESIGN ..................................................................................................................... 12 TOLERANCE MANAGEMENT .......................................................................................................... 12 COGNITIVE PSYCHOLOGY............................................................................................................. 18 MARKETING ................................................................................................................................. 20
RESULTS ............................................................................................................................................ 21 4.1 4.2 4.3
5
DESIGN RESEARCH METHODOLOGY ................................................................................................ 6
FRAME OF REFERENCE .................................................................................................................. 9 3.1 3.2 3.3 3.4 3.5 3.6
4
GEOMETRICAL AND VISUAL ROBUSTNESS ...................................................................................... 2 RESEARCH FOCUS ........................................................................................................................... 3 RESEARCH HYPOTHESIS AND RESEARCH QUESTIONS ...................................................................... 3 DELIMITATIONS .............................................................................................................................. 4 OUTLINE OF THE THESIS ................................................................................................................. 4
PAPER A ....................................................................................................................................... 21 PAPER B ....................................................................................................................................... 21 PAPER C ....................................................................................................................................... 22
DISCUSSION ...................................................................................................................................... 27 5.1 5.2 5.3
RESEARCH APPROACH .................................................................................................................. 27 RESULTS IN TERMS OF RESEARCH QUESTIONS............................................................................... 27 CONTRIBUTION ............................................................................................................................. 28
6
CONCLUSIONS ................................................................................................................................. 29
7
FUTURE RESEARCH ....................................................................................................................... 31
8
REFERENCES.................................................................................................................................... 33
APPENDED PAPERS PAPER A PAPER B PAPER C
vii
INTRODUCTION
1
INTRODUCTION
This chapter gives an introduction to the research project, including the background and objectives. The goals and research questions are presented. Finally, the outline of the thesis is gone through at the end of the chapter. Product development always faces new problems to solve and new requirements to fulfil. The problems and requirements differ from time to time, between markets, the types of products, etc. Despite this, the lead-time and cost are always expected to decrease and the quality to increase. This raises a need for methods, tools and guidelines for the daily product development work in order to handle the requirements and solve the problems. This thesis presents results that support the area of tolerance management in order to meet functional, dimensional and aesthetic requirements. The area of tolerance management is trying to find ways to handle geometrical variation - derived from part variation and assembly variation - in order to meet the different types of requirements. In more concrete terms, the research has focused on how the initial product geometry shall be split into parts to achieve a geometrically and visually robust1 design. It is the spatial relation between the mating parts in the divided geometry that is critical: It is that characteristic that is in focus for the geometrical and visual robustness. The relations here are called split-lines (See Figure 1). Geometrical robustness is an assembled product’s ability to suppress part and assembly variation. Visual robustness is an assembled product’s ability to suppress a, by the customer or user2, visually perceived lacking quality appearance caused by part and assembly variation. The common opinion among automobile manufacturers is that splitlines shall be narrow, parallel and equal. It is a sales argument for a number of automobile manufacturers. Even the customers are of the opinion that the geometrical and visual quality of split-lines are important for the overall quality appearance of the automobile (Dagman et al., 2004). Both geometrical dimensioning aspects and aspects like the design language, crash safety, wind resistance, etc. affect the dividing of the initial product geometry. All these aspects have to be considered during the product development process.
Figure 1.
Examples of split-lines.
In order to delimit this project, only the aspects concerning design language and geometrical dimensioning have been treated. This research project has been conducted within the research group for Robust Design & Variation Simulation at Wingquist Laboratory, Chalmers University of Technology. 1 2
The term visually robust or visually sensitivity was first mentioned by Wickman and Söderberg (2001a). The user relation may involve, but need not involve, an owner relation, (Karlsson, 1996).
1
INTRODUCTION
1.1 Geometrical and visual robustness Assembling a product that meets dimensional and aesthetic requirements is difficult. All parts included can be of high quality. However, if they do not fit together geometrically, the product achieves a bad quality. In the automotive industry, split-lines between different parts in an assembly are often critical and are an important factor for the overall quality appearance (Söderberg and Lindkvist, 2002). The split-lines can be between fenders, doors, hoods and panels (See Figure 1). The split-lines in Figure 1 can be seen as functional split-lines since they fulfil functionality. There can also be split-lines that do not fulfil any functional requirements. In order to measure variation and judge split-lines, two main measuring directions in the split-lines have been of interest. They are gap and flush. These directions are commonly used in the automotive industry. Gap is defined as the distance perpendicular to the normal surface between two parts, while flush as the distance on the axis of the normal surface between two parts (See Figure 2).
Figure 2.
Two important measuring directions, gap and flush.
The possibility to fulfil the dimensional requirements is determined by the geometrical robustness of the design, the part variation and the assembly variation. The part variation has its origin from variation in the manufacturing process and the material characteristics. The assembly variation has its origin from variation in fixtures and tools in the assembly line. The simplest way to fulfil the dimensional requirements is to refine the tolerances for the parts and the assembly. But this increases the cost and might also increase the leadtime. Also, if the assembly process is unstable, this might still not solve the problem. By being designed geometrically robust, the variation is suppressed. This leads to a fulfilment of the dimensional requirements without refining the tolerances. It does not solve all problems with variation. Nonetheless, it sets the direction for how difficult it will be to fulfil the dimensional and aesthetic requirements. A geometrically robust solution often results in a visually robust solution. A geometrically sensitive solution, on the other hand, can be made visually acceptable in some cases by controlling the position and direction of the split-line. Different areas are more or less sensitive to the eye. Locating schemes are used in order to control and position components in assemblies. The locating schemes lock six degrees of freedom3 for each component in the assembly. The configuration of the locators in the locating schemes influences the output variation in the assembly. Research has been carried out finding the most geometrically robust placement of the locators (See (Söderberg and Carlson, 1999)). In automotive production of today the position of the locators are more or less fixed since the same production line is used on new models. This is a consequence of the rapidly increasing number of different models and decreasing development time. In this thesis the prerequisites with fixed locating schemes have been used in order to imitate the production climate of today.
3
The six degrees of freedom consist of three rotations and three translations.
2
INTRODUCTION
Variation between parts in an assembly does also has an influence on the aesthetic requirements of a product. The aesthetic requirements of the split-line are not only controlled by input variation and geometrical robustness but also by the visual robustness of the design. The aesthetic aspects of split-lines in this thesis are treated in terms of Visual Quality Appearance (VQA) of a split-line (Dagman et al., 2004). VQA is the quality impression that a product conveys visually to a customer when observing it. The measurement of the VQA is subjective by nature since perception differs between users.
1.2 Research focus The general research focus of this project has been to support tolerance management during the product development process of consumer products where VQA is of importance. A design solution can be both visually and geometrically sensitive. This leads to increased lead time, higher cost and reduced product quality. This addresses a need for increased knowledge in the area. It also directs itself to a need for tools and methods to support the design process in general and tolerance management specifically. The goal is to find out how to split the initial product geometry with respect to visual and geometrical sensitivity in order to achieve a product with high VQA. 1.2.1 Industrial and scientific relevance By using the tools and the knowledge created during the presented research project in industry, the chances to create a product with high VQA of split-lines will increase. In addition, verification of design solutions in the early stages of the product development process, with respect to geometrical and visual sensitivity, can decrease lead-time and cost. The expected contribution to science is to enhance the knowledge regarding how to split the initial product geometry, with given locating schemes, to achieve as geometrically and visually robust a solution as possible.
1.3 Research hypothesis and research questions The hypothesis for this research is formulated as: Enhanced knowledge and proper tools to support the splitting of the initial product geometry into individual parts with respect to visual and geometrical robustness will improve the VQA of products. The main research question is formulated as: How shall the initial product geometry, with given location schemes, be split into individual parts to achieve a geometrically and visually robust solution? Four aspects that influence VQA of split-lines form the basis for the four sub-research questions, treated in the appended papers.
3
INTRODUCTION
Sub-research questions
How shall a geometrically and visually robust splitline be placed and designed with respect to:
Gap?
Flush?
Geometry?
Design language? (Renault, 2005) The flush and gap aspects have been described in the early text. The geometry concerns, for example, the influence of the curvature of the geometry and where geometry the split-line is placed, etc. The design language aspects concern the aesthetic rules when designing spilt-lines and how split-lines affect the general language of the product.
aspect on the use of design
1.4 Delimitations The products treated so far in the projects have been delimited to automobiles, and the focus has been on the early stages in the product development process. The geometries used during the experiments have been shell models. As such, the exterior shape of the product has been in focus for the analysis and simulations. The location schemes for the parts used have been fixed, providing no possibility to move after defining the location. No consideration has been taken regarding the function of the product.
1.5 Outline of the thesis This thesis starts with an introduction to the research area. The goals, hypothesis and research questions are presented. The adopted research approach and the field of design research methods are described in Chapter 2. The frame of reference is presented in Chapter 3, in order to give a theoretical guide to the research area. The results are then presented in Chapter 4. These results and the research approach are discussed in Chapter 5. Conclusions regarding the research are presented in Chapter 6. Finally, in Chapter 7, future research is presented and analysed. At the end, three published papers that are the basis of this research are presented
4
RESEARCH APPROACH
2
RESEARCH APPROACH
This chapter introduces the area of research methodologies and methods applied in this research. The methods and methodologies are used in order to structure and focus the research process. This also increases the possibility to recreate the results. Design science4 is a multi-facetted research area involving many disciplines. There are influences such as natural, psychological, social, engineering and computer science that contribute to the design research field. Hence, the research approach has to be adopted so that it matches these research areas. A number of design research methodologies have been developed, (Duffy and Andreasen, 1995), (Blessing et al., 1995) which try to meet the nature of design science. The research methodologies consist of a number of methods that shall be used in order to increase the knowledge in a certain area or to solve a problem. Since this research project focuses on geometrical and visual robustness, different methods have been used in order to perform the research. The nature of the two aspects does not allow the same kind of research methods. Geometrically robustness is in the area of engineering science, while psychological science also influences visual robustness. The research presented has been generated by the use of empirical studies both on the field and in the laboratory environment, as well as existing theories from literature studies. According to Hubka and Eder (1988), design science can be divided into four main directions: prescriptive and descriptive research in the areas of technical systems and design process (See Figure 3).
Figure 3.
The dimensions of design science (Hubka and Eder, 1988) and the contribution from this thesis, marked with grey.
4
Design science is defined by Hubka and Eder (1996) as: ”a system of logically related knowledge, which should contain and organize the complete knowledge about and for designing”.
5
RESEARCH APPROACH
The prescriptive statements prescribe ways of working with either the technical system or the design process. The descriptive statements describe the current or future state of a technical system or the design process. The contribution of the research presented in this thesis is marked with a grey shape in Figure 3. The main contribution from the research project is positioned in the upper part, both right and left (See Figure 3). The contributions have been developed and implemented in a computerized environment. This leads to the contribution to the part called CAD (Expert systems). Design research is a special form of problem solving. A similar cycle of activities can be recognized in all forms of problem solving. De Groot (1969) has sharpened the cycle, and calls it ‘The cycle of empirical scientific inquiry’. This cycle is shown in Figure 4 and shows how empirical research can be carried out. It is used for solving problems about knowledge or theoretical problems. Empirical material is collected in the observation phase. It is then grouped and the (tentative) formation of hypothesis is carried out. The precise formulation of the hypothesis is carried through in the induction phase. Deduction leads to explicit and more accurate predictions. They are tested, mostly experimentally, on new empirical material. Evaluation is the interpretation of the results of the testing in a wider context. This is a general formulation of how research can be executed. The work flow in this thesis is connected to the cycle. Figure 4.
The cycle of empirical scientific inquiry, (Roozenburg and Eekels, 1995).
2.1 Design research methodology Design Research Methodology (DRM) was presented by Blessing et al. (1995) (See Figure 5). It can be seen as a refinement of the cycle of empirical inquiry presented by De Groot, specially adopted to design research. It is based on a measurable success criterion. To be able to investigate whether your research has had the intended effect, you need to have something measurable, qualitative or quantitative that indicates success. When the success criteria are described, a first descriptive study (descriptive study I) has to be made.
6
RESEARCH APPROACH
Figure 5.
The Design Research Method.
The aim with this study is to investigate the current status of the research area and to identify what factors influence the success criterion and in what way they do it. The descriptive study answers ‘what’ questions of the type ’how many’ and ‘how much’ (Yin, 1994). There is a wide range of methods to use when performing a descriptive study. Examples include interviews, observational notes, videos, etc. The next step is to perform a prescriptive study. In this study, the researcher makes assumptions and uses his or her experience to synthesize a solution to the problems identified in descriptive study I. After this has been achieved, a second descriptive study (Descriptive study II) shall be done. This study focuses on a validation of the solution that came out of the prescriptive study. It also seeks to ascertain whether the factors that were identified in descriptive study I had the intended influence. The model describes a research approach covering all stages in a research project, from the very beginning to the end. 2.1.1 Applied research approach The research approach used in this research project is based upon DRM. The research steps taken in this thesis are presented in Figure 6.
Figure 6.
The current status of the research.
Since the research is based upon two different but closely linked areas, the research is positioned at two different stages. It is marked in dark grey in Figure 6. The research
7
RESEARCH APPROACH
focusing on geometrically robustness is in the prescriptive study phase, where a tool is designed to solve that issue. Knowledge regarding what factors affect the design and make it geometrically sensitive has been gained through literature studies. The studies functioned as input information to the research project. The research focusing on visual sensitivity is still in the descriptive study I phase. It seeks to describe the current status. Since the main focus for the visual sensitivity part has been to learn about customer and industry attitudes, semi-structured interviews (Westlander, 2000) and (Kvale, 1996) and questionnaires (Trost, 2001) have been used. The first paper presents a prescriptive study in the area of geometrical sensitivity. It presents a computerized tool and a way of working to solve the problem in that area. The second paper is a descriptive study that attempts to discover customers/users’ and industry’s opinions regarding a specific area. Semi-structured interviews and questionnaires were used. The questionnaires were used as a guide for structured interviews, i.e. the same questions were asked in the same way and the interviewee had no possibility to influence the subject discussed. The analysis of the results was evaluated in a qualitative way, meaning that the results from the study were transformed into numbers for the analysis. The third paper is similar to paper one in that it presents a prescriptive study describing a computerized tool that solves geometrical sensitivity issues.
8
FRAME OF REFERENCE
3
FRAME OF REFERENCE
This chapter gives an overview of the areas that have contributed to and influenced the research project. The character of this research project is multidisciplinary: it embraces contributing aspects from a wide range of research areas. Areas of greater importance for the research are explained in more detail.
3.1 The research surroundings The main areas contributing to this research are illustrated in Figure 7. The areas are clustered together into five main groups: Cognitive Psychology, Engineering Design, Industrial Design, Marketing and Tolerance Management. These five main areas consist of several sub-areas, illustrated as bubbles in Figure 7. In this chapter, the areas will be described and explained.
Figure 7.
Illustration of areas contributing to the presented research.
3.2 Engineering design The area of engineering design is large, covering several aspects in the product development process. This frame of reference chapter only presents the parts of engineering design that have had a direct influence of this project. 3.2.1 The design process The design process has shifted in time and with the complexity of the product (Ullman, 2003). In the early days, the product could be designed and manufactured by one person. But in the middle of the 20th century, the complexity of the product increased in such a way that it was difficult for just one person to do all the steps to create a product. A product development project of today can involve thousands of people from several continents. This necessitates a different approach. Different suggestions for how to manage the work organization during the design process have been proposed. Examples of these design methodologies are concurrent engineering or integrated product development (Andreasen and Hein, 2000). A common understanding in the methodologies is that the
9
FRAME OF REFERENCE
development process shall be performed simultaneously and not sequentially to achieve the best result in the shortest time (Andreasen and Hein, 2000). The area of tolerance management involves many disciplines in a company. Hence, integrated product development could be a good way of structuring the work. The design process has been in focus for research for many years. By understanding the design process, suggestions for how to improve the process can be made. But the design process is complex, and the conditions for the process are unique every time. This makes it difficult to analyse. Models of the design process have been presented (Pahl and Beitz, 1996); (Ullman, 2003) and (Ulrich and Eppinger, 2004). The models describe a number of major activities and phases during the design process. According to Pahl and Beitz (1996), the four phases are: 1. Clarification of the task. The problem is analysed, information regarding the problem is collected, and the design specification is drawn up. Information regarding requirements and existing constraints is taken into account and documented. 2. Conceptual design. Principle solutions are generated and evaluated in accordance with the requirements. The result is the specification of principles. 3. Embodiment design. In this phase, the concept is refined into a preliminary design and then into a definitive layout. The result is the specification of the layout. 4. Detail design. All the characteristics of the concept are specified, resulting in product documents. It is intended that the results from this research will be used in the borderline between the conceptual design phase and the embodiment design phase. This is the stage in the product development process in which the concepts have been presented but the final design is not fixed. The design process is iterative, meaning that the mentioned phases can be treated several times during a design process (Roozenburg and Eekels, 1995). For example, if the evaluation of the principle solution indicates that specific requirements will not be fulfilled, the process has to step one phase back and try to clarify the task better. The number of iterations and the design process time can decrease if proper methods and tools are used. That is the starting-point of this thesis. Even enhanced knowledge gained from earlier projects and individual experience affects the lead-time of the design process. 3.2.2 Product quality Mørup (1993) defines product quality as “the customer’s experience (or perception) of how well the totality of quality properties of a product satisfies is stated or implied needs.” According to Ullman (2003), apprehension regarding product quality is very individual. The product quality term can comprise a wide range of aspects, from “a product works as it should” to “a product looks good,” for example. Ullman (2003) also states that “quality cannot be manufactured or inspected into a product, it must be designed into it.” According to Mørup (1993), there are different stakeholders of the product who should be considered during the product development. There are internal stakeholders, e.g. designers, marketing, production personnel, etc. There are also external stakeholders, e.g. customers, users, etc. Two quality concepts have been presented by Mørup (1993) based on the different stakeholders:
10
FRAME OF REFERENCE
- Q-quality (“big Q”). Q is the customer’s qualitative perception of the product. - q-quality (“little q”). q is the internal stakeholders qualitative perception of the product in relation to his product-related tasks. In this research project, the area of geometrical sensitivity is q-quality and the area of visual sensitivity is Q-quality where VQA is of high importance. There is a wide range of parameters that affect the VQA of a split-line. It could be quantitative parameters, like the actual distance of the gap or the flush or non-parallelism. It could also be qualitative parameters, like the curvature of the split-line, what is visual within the split-line, colour of surfaces, the design, etc. Research in the area of tolerancing aims at giving the large volume products the intended functionality, dimensions, etc. However, it is also important that the assembled product meets the quality appearance requirements that have been settled. The Quality Appearance (QA) index introduced by Söderberg and Lindkvist (2002) evaluates how variation and mean gap and flush values effect the QA of an automobile. An automobiles split-line QA is judged by the measure values in gap and flush direction between automobile body components like hoods, fenders and doors. The geometrical quality of doors have been treated in Gerth et al. (2002). Robust design is a central issue in quality research and development (Bergman and Klefsjö, 1995); (Phadke, 1989). The meaning of robustness in the area of product quality is a product’s ability to suppress external distortions, internal distortions and manufacturing distortions. Below is a brief description of these three terms: • External distortions are variations in temperature or stress or other environmental factors during use. • Internal distortions are wear or crack propagation as a result of external distortions. • Manufacturing variation is an individual unit’s deviation from the target value for the manufacturing. This way of thinking comprises the usage of the product. Robust design in the tolerance management context does not do as such. Taguchi (1986) has treated the correlation between tolerances and quality loss, called quality loss function. Taguchi means that all deviation from the target value results in a loss that increases exponentially with the deviation (See Figure 8). Before Taguchi came up with this exponential deviation, the traditional approach considered a loss when tolerance limits were crossed. A common loss that has been presented is cost vs. tolerances.
11
FRAME OF REFERENCE
Figure 8.
Differences between Taguchi’s and the traditional loss function.
Chase and Parkinson (1991) presented a summary of purposed cost vs. tolerance models. Söderberg (1993), (1994), (1995) developed the loss function and incorporated both the manufacturing (q-quality) and customer (Q-quality) value into the loss function. This created a total loss to customer function.
3.3 Industrial design In the last years, the importance of a product’s aesthetic appearance has increased. The customer buys a product not only for its functionality but also for its design. Hence, it is of importance that the area of industrial design is incorporated in the product development process in a way that their solutions and suggestions co-operate with the functional solutions and not inversely. In the automotive industry of today, industrial design concept suggestions are very important. It is one characteristic of the product among many others. However, since its importance has increased, much effort is put into achieving the aesthetic requirements. By supporting industrial designers with tools and methods that allow them to evaluate the suggested design solution at an early stage of their process, it is possible to discover problems that would have been time-consuming and costly to solve if discovered later in the process. The tools presented in this research have the intent to support both engineering design and industrial design competences and organisations.
3.4 Tolerance management Tolerance management has been an area of interest for both research and development for a long time. Ever since plus/minus limits appeared at engineering drawings in the early 20th century, the area of tolerances has been an important issue for engineers in the product realization process (Hong and Chang, 2002). The tolerance issue arose when parts from different manufacturers were to be assembled together. If two components are to be assembled together, a stiff shaft fitting into a hole, for example, it is important that the geometrical dimensions of them are right. If the shaft is too big or the hole is too small, it will be impossible to assemble them without damaging the parts, even if each component is well made. A pioneer in the area was the Swede C. E. Johansson, who invented the gauge block that made mass production possible. Henry Leland, President of Cadillac Motors, once said:
12
FRAME OF REFERENCE
“There are only two people I take off my hat to. One is the president of the United States and the other is Mr Johansson from Sweden” (Johansson, 2005) C. E. Johansson also established a partnership with Henry Ford that lasted for 17 years. So it can be said that C. E. Johansson laid the grounds for this research project, especially since it involves both car manufacturing and measuring. 3.4.1 Tolerance management theory The possibility to perform precise measuring is one of the keystones of tolerance management. Such measuring enables the designer to see how much the individual part varies and the variation of the whole assembly. By measuring several manufactured components, it is possible to calculate statistics, i.e. standard deviation, mean, range, etc. The statistics describe the distribution of the components as a result of the process variation. This knowledge is very valuable when defining the tolerances for the design. Without knowing the range and distribution of the parts in an assembly, the designer is more or less groping in the dark when specifying the tolerances. By specifying tolerances, the functionality of the design is guaranteed, assuming the part and assembly variation have not changed. The research field of tolerancing is wide and consists of many different areas. Hong and Chang (2002) have presented a compilation and structure of the area of tolerancing, and has been used in this thesis to describe the area of tolerance management (See Figure 9). Research in tolerancing
Tolerancing schemes
Figure 9.
Tolerance modelling and representation
Tolerance specification
Tolerance analysis
Tolerance allocation
Tolerance transfer
Tolerance evaluation
Compilation of the main branches in tolerancing research (Hong and Chang, 2002).
The information regarding tolerances has to be able to be transferred between two competences without losing any information. For example, the tolerances the designer specifies have to be transferred to the production department. Hence, there has to be a common nomenclature. Two main types of tolerance representations are used: conventional (plus and minus tolerances) and geometrical. In more theoretical language, these types of tolerances are called two types of tolerance schemes. The two types are parametric and geometric. Parametric tolerances aim at finding a set of parameters and assigning limitations to the parameters so that they define a range of values. Geometric tolerances assign values to specific attributes of parts, such as shape, position location, etc. The area of tolerance modelling and representation tries to find an efficient way of defining and specifying tolerance information mathematically or electronically. The development in solid modelling techniques has made it easier to integrate information
13
FRAME OF REFERENCE
in the product model as an intrinsic part of product definition. However, a totally unified system has not been established yet. The specification of tolerances is mainly based on the designers experience and/or empirical data and is called tolerance specification. The area is often closely treated with tolerance standards. Tolerance analysis focuses on two parameters: variation of and between parts. These two parameters are used to evaluate the functionality of the design. This area has been in focus for this research project and will thereby be explained in more detail later on. Tolerance allocation aims at distributing the tolerances to individual parts of an assembly in order to meet the tolerances requirements for the assembly. Tolerance transfer aims at determining how to convert design tolerances into a manufacturing plan. It is sometimes mentioned as tolerance analysis and allocation in process planning. In the area of tolerance evaluation, the research focuses on how to access geometrical variation by use of Coordinate Measuring Machines (CMM). Product key characteristics The area of tolerancing has the goal to control and handle the output variation in order to fulfil the requirements set on the product. To control the output variation, the designer is able to manipulate the input variation of the parts, the assembly variation, and the level of geometrical robustness in the design (See Figure 10). Lee and Thornton (1996) have introduced the Product Key Characteristics (PKC). The key characteristics refer to product features, manufacturing process parameters, and assembly features that significantly affect a product’s performance, function and form (Lee and Thornton, 1996). The PKC in this research project have been geometrical variation, which has been introduced in Figure 10. The PKC is influenced by the mentioned manipulation parameters for the designer. Component variation Machine precision Manufacturing process
Process variation
Assembly variation Assembly precision
Process variation
Assembly process PKC variation Robustness
Design concept
Figure 10.
The affection of the PKC, variation (Söderberg, 1998).
3.4.2 Robust design To achieve the functionality and the quality appearance that meets the predetermined requirements, it is necessary that the included parts and assembly achieve the specified tolerances. Tight tolerances are expensive and not desirable in all stages. Robust design is a method for tackling this issue (Taguchi et al., 1989), (Söderberg, 1998) and (Söderberg
14
FRAME OF REFERENCE
and Lindkvist, 1999). The robustness can be explained simply with a beam and a support (See Figure 11). If the support is placed in the middle of the beam, the output variation will be the same as the input variation. Output
Input
Robust
Sensitive
Figure 11.
A beam and support describing the relation between input and output variation.
If the support is moved to the left, the output variation will increase and the solution will be sensitive to variation. Conversely, if the support is moved to the right, the input variation will be suppressed. That leads to a lower output variation in relation to the input variation. 3.4.3 Tolerance analysis There are many ways of calculating the assembly tolerances. Kumar and Raman (1992) have made a summary of different methods, where a number of examples will be described in the following text. The worst-case model (1) describes a straight summation of all of the individual part variations. TA stands for assembly tolerance and tc for component tolerance. The number of components is represented by n. This model guarantees that the final assembly tolerances of the product will be in this area. However, this model is very sensitive to the number of integral parts. As the number of parts increases, the individual tolerances of the integral parts have to decrease in order to meet the assembly limit (Chase and Parkinson, 1991). n
TA = ∑ tc (1) c =1
While this method ‘only’ gives the worst-case scenario, it would be better to use a statistically-based method, since only a very small quantity of the manufactured parts diverge from the nominal value that much. In the Root Sum Square (RSS) model (2), the component tolerances are assumed to have a normal distribution. This allows the component tolerances to increase significantly since they add on as RSS. The RSS analysis generally predicts too few rejects compared to real assembly processes (Chase and Parkinson, 1991). This is due to the fact that the normal distribution is an approximation of the real distribution. The real distribution can be different (for example, flatter or skewed). TA =
n
∑t c =1
2 c
(2)
To be able to handle statistical models that do not represent the actual assembly distribution, a correction factor was introduced, C (3). The value of C has to be determined based on practical assembly results (Kumar and Raman, 1992). TA = C
n
∑t c =1
2 c
(3)
Spott (1978) suggested that the mean of the worst-case and statistical model, Spotts’ modified model (4), represented the assembly tolerance. Compared with the other
15
FRAME OF REFERENCE
models, this reduces the rejection rate of the assembly when the components have a skewed distribution (Kumar and Raman, 1992).
TA =
n
n
c =1
c =1
∑ t c + (∑ t c2 ) 0.5
(4) 2 New and more complex ways of calculating the assembly tolerances have been presented using different kinds of simulation models to represent the part variations. One of these simulation models is the Monte Carlo simulation, which will be described in the next chapter.
Monte Carlo simulation The meaning of simulation is to replace the real world with a mathematical or physical model. Monte Carlo simulation has been used to be able to investigate the characteristics of a complex system, typically a mechanical assembly. The Monte Carlo simulation is based on the use of a random number generator to simulate the complex system (Chase and Parkinson, 1991). In this case, the effects of manufacturing variations on assemblies have been simulated. The random number generator is used to apply variation to each component dimension according to a predefined distribution. The assembly variation is then calculated in accordance with the component variations and is, for example, plotted in a histogram. The mentioned procedure is iterated until a sufficient number of assemblies have been simulated to plot a histogram and estimate the standard deviation. The Monte Carlo simulation requires a large number of samples to achieve accuracy. This was a problem in the past. However, with today’s computer capacity, it is one no longer. Locating schemes In the area of assembling, the location scheme, i.e. in what way a part is positioned in a fixture, for example, affects the output variation. Variation in the assembling equipment will affect the final variation of the assembled product. There are a number of different location schemes that describe how a part is positioned in space. The 3-2-1 scheme is commonly used in the automobile industry (See Figure 12). Theoretically, the locators are points used to lock all degrees of freedom for a part to be able to position it. Physically, physical geometrical features like holes, planes and slots solve the locator points. A part has six degrees of freedom, three translations and three rotations. The first three points, which form a plane, lock one translation and two rotations. The next two points, which form a line, lock one translation and one rotation. The sixth point locks the remaining translation. The 3-point scheme is similar to the 3-2-1 scheme with the difference that only three points are used. The first point locates the part in three directions, the second in two and the third in one direction.
16
FRAME OF REFERENCE
Figure 12.
The 3-2-1 location scheme.
In Figure 13 the contribution to gap and flush of the locator points is illustrated. The first three points, A1-A3, influence mainly in the flush direction. The two points that form a line, B1-B2, influence in the gap direction. The direction of the split-line determines how much B1-B2 and C contribute to variation in gap direction. As long as the split-line is perpendicular to B1-B2, as in Figure 13, the contribution from the C locator point will be zero, given that there is orthogonality between the reference directions in the locating scheme. The configuration and orientation of the location schemes determine which locators have influence in the different directions. Thus, if the locator schemes had been placed differently the individual locators might have influenced the variation direction differently. A study of different locating scheme configurations has been presented by Brahmst et al. (2002).
Figure 13.
The movements of the parts are dependent on the orientation of the locating schemes.
In most cases, when calculating tolerances the integral components are considered rigid. Dahlström et al. (2002) have examined how non-rigid components that are part of a compliant assembly are affected by variation and how these should be treated. In this thesis the parts are considered rigid. 3.4.4 Computer-aided tolerancing The use of computerized tolerance management tools has increased. A number of different Computer Aided Tolerancing (CAT) tools have been developed for the analysis of assemblies regarding tolerances and geometrical robustness, for example RD&T, VSA, CE/TOL and 3DCS. Kumar and Raman (1992) and Prisco and Giorleo (2002) have presented overviews of CAT tools. In a CAT tool it is possible to analyze assemblies in
17
FRAME OF REFERENCE
different stages of the product development process. In the concept phase, when the real tolerances are not known, the focus should be on optimizing the geometrical robustness of the concept. In the later stages when the real tolerances are known, the goal is to find the appropriate tolerances to meet the design, manufacturing and cost requirements. The analysis can be performed in both 2D and 3D. By combining a CAT tool and a visualization software, the output variation can be visualized in a virtual reality environment (See (Wickman and Söderberg, 2001b) and (Maxfield et al., 2000)). By using non-nominal geometries, i.e. geometries that are affected by manufacturing variation, the actual variation can be visualized. When CAD (Computer Aided Design) softwares are used as a basis for decision regarding dimensional requirements, the models are nominal, which does not reflect the actual situation in production. When evaluating the non-nominal assemblies in the virtual environment, the visual and geometrical robustness can be judged by ocular inspection at the computer screen. Wickman and Söderberg (2003) have compared non-nominal geometry models represented in physical versus virtual environments. The result showed that it is harder to estimate variation in virtual assemblies compared to physical ones. The possibility to virtually evaluate geometrical robustness in the concept phase will increase the chances to improve the final product quality. The seam function is used in appended Paper A. It is a function that was introduced by Söderberg and Lindkvist (2001) as the relation between two parts over a specified distance, and describes one of the most frequently used quality characteristics for evaluations of geometrical variation in automotive body design (See Figure 14). Typically, seam variation is measured and evaluated in two directions, the gap and flush directions.
Figure 14.
Colour coding of seam variation, flush direction (Söderberg and Lindkvist, 2001).
3.5 Cognitive Psychology Cognitive psychology is the branch of psychology that studies how humans treat information. In this research project, the sense that is in focus is vision. When humans are perceiving stimuli with their senses there is, in most cases, a judgment regarding the sensation: it is good or bad. When perceiving an assembled product, the split-lines will not only be the result of two mating parts. They are, in most, cases a black line, dividing a shape into a number of individual parts.
18
FRAME OF REFERENCE
3.5.1 Product aesthetics The split-lines can be judged regarding aesthetic appearance etc. where aesthetics can be described by:
Empirical investigations of factors that effect aesthetical preferences and beauty experiences (Nationalencyklopedin, 2004). A product is in many ways closely connected with aesthetic preferences and experiences of beauty. The design and placement of split-lines affect these preferences and experiences (Dagman et al., 2004). An assumption based on this is therefore that split-lines should contribute to the aesthetic whole and not the contrary, in order to increase the VQA. If the shape of the split-lines does not follow the general design language, the VQA can be affected in a negative way. This assumption will be investigated in future research. To be able to distinguish between different brands, the design language is of importance. The design language of a brand is built up of a number of main design entities (Warell, 2001) that characterize the design language. Based on this research, Warell (2001) has presented a method for finding the entities in the visual product form. This knowledge can also be of importance when dividing the initial product geometry.
Gestalt theories Gestalt5 theories originated from a group of psychologists such as Kurt Koffka, Wolfgang Köhler and Max Wertheimer, who during the 19th century studied how individual elements can be organized into figures. A common topic was how individual stimuli are grouped under perception to form a whole or gestalt, where the whole is more than the sum of the parts included. This enables the human to perceive a whole and to interpret a message from what in many aspects are randomly placed parts. A number of gestalt laws have been developed with this knowledge as a foundation. One example of a gestalt rule is the common movement; factors that are moving in the same direction are seen as a whole, shown in Figure 15. Cars in the same line on a picture are grouped as a whole.
Figure 15.
The common movement (Monö, 1997).
5
Gestalt refers to an arrangement of parts which appears and functions as a whole that is more than the sum of its parts (Monö, 1997).
19
FRAME OF REFERENCE
3.6 Marketing Product development, production and marketing are the three activities that are the foundations for developing and producing a product (Andreasen and Hein, 2000). In order to develop a product, there has to be a need for the product established. There are two sources for a design project: the market or the development of a new product idea without market demands (Ullman, 2003). Most new product development is market-driven. Without having a customer for the product, the company will not last long, since the income that should cover the design and manufacture expenses of the product is missing. Marketing provides the company with information regarding what the customers want in their product (Ullman, 2003). Marketing can also identify stimuli from the market, such as the technical and economic position of the product, changes in the market requirements etc. (Pahl and Beitz, 1996). These stimuli are important for planning the new product. Marketing is also responsible for the sale of the product. High VQA of split-lines is a sales argument with a number of automobile brands, although far from all (Dagman et al., 2004), but it is increasing. The area of marketing is huge and, since this research project has had its focus on product development, that area will not be treated any further. The material presented in this chapter was presented to give a foundation for a deeper understanding of the research results that will be discussed in the next chapter.
20
RESULTS
4
RESULTS
A summary of the appended paper is presented to highlight the results from the research that is the basis for this thesis. Papers A and C concern the division of the initial product geometry with respect to gap and flush. Paper B investigates the customers' and the automotive industry's attitude regarding visual quality appearance of split-lines.
4.1 Paper A The goal with Paper A was to find a way to split two simple geometries (thin plates) with given locator schemes in order to achieve as geometrically robust split-line as possible. The locator schemes applied to the geometries should be fixed. The tool itself should not determine the actual split-line but show the least sensitive area. The results should be presented in two dimensions with the simple geometries. Two equally dimensioned and equally shaped plates were used as experimental geometries. The plates were modelled and placed on top of each other in a CAT software. Two different 3-point locating schemes were defined on the two geometries, one on each plate. In Figure 16 the locations of the locating schemes are marked with filled white circles. The three filled circles at the far right in both illustrations in Figure 16 belong to one part, and the other three at the left to the other part. A large number of seams were applied to the two geometries, creating a net of seams, represented as lines in different directions in Figure 16. Unit tolerances were applied, since no real tolerances were known, and an analysis of the general robustness was performed. The results were colour-coded (a feature in the CAT software; see Paper A for more information). The blue (dark) area in the middle of the geometries indicates low variation in the flush direction.
Figure 16.
The colour-coded result in the flush direction.
In the flush direction, distinct patterns describing the least geometrically sensitive area appeared. Different locating schemes resulted in different shapes on the most geometrically robust areas. The results in the gap direction pointed to similar patterns, as in flush. But these results were not as clear as in the flush direction.
4.2 Paper B This paper presented an empirical interview study of customers and the automotive industry’s attitude regarding VQA of split-lines. The aim of the paper was to find out whether the customer appreciates the effort that the automotive industry is putting into the area of geometrical and visual sensitivity. It was also investigated what factors affect the
21
RESULTS
exterior quality of the automobile, and whether there is a correlation between measurement data of gap and flush and the visually perceived VQA of split-lines of different automobile brands. Data were collected by performing structured and semistructured interviews with customers/users and persons from the industry. The interviews were carried out at a large motor show in Germany, in a large automobile manufacturer in Sweden, at automobile retailers in Sweden, and with the Swedish public. The total number of interviewees was 130. A selection of the results obtained during the interviews is presented together with
Visually, factors give the exterior appearance of a car high quality? the questionwhat asked. This question was posed as an initial question, without revealing the main purpose of the interview – to get an overall insight into what factors make an automobile express high quality. The two most often mentioned factors that affected the exterior appearance were the design and the colour, with high VQA of split-lines
general, do you think that good fit and flush is important when choosing a car? onInfourth place. Here the question was asked in order to investigate whether the area of fit and flush is something important for the customers and the industry’s. A majority of the
What car brand do you the best VQA regarding split-lines? interviewees thought that think it washas important. This question was posed only to the customers since an answer from the industry would most certainly contain bias. A majority of the interviewees could not mention which brand they thought had the best VQA regarding split-lines. Five brands stood out from the rest as most often mentioned. The results treating the brand which the customers thought possesses the best VQA regarding split-lines were compared with gap and flush measurement data compiled by the Ford Motor Company. They had measured a large number of split-lines at automobiles from different brands. The results showed that it was clear that the brands most often mentioned by the customers also had small gaps and flush. But there was not a one-to-one mapping between the measurement results and the interview results, indicating that there are some other factors that affect the VQA of split-lines than just the distance in the gap and flush directions.
4.3 Paper C Paper C is a refinement of Paper A. In this paper the analysis was made on 3D geometries, compared to Paper A where 2D geometries were used. Instead of using the seam function for the calculation of variation, measures were applied in the VRML nodes describing the geometries. This technique was possible to use in the flush direction but not in gap. In the gap direction, fictive split-lines were performed to create measuring points along them, similar to the seam function. The same analysis set-up was used as in Paper A, but this
22
RESULTS
time automobile geometries were used. The locator schemes were fixed. Three main results were presented.
Result 1 A CAT functionality to simulate and analyze where the most geometrically robust area between two parts, with given location schemes, is located and shaped. The colour coding shows a blue (dark) coloured area describing the geometrically most robust area in the flush direction between two geometries. The analysis area shown in Figure 17 describes the area between front door and front fender. By applying the colour-coded results to the geometries to be divided, the designer gives an opportunity to create the robust split-line with the result as a support.
Figure 17.
The analysis result of front fender and door on a shell geometry in the flush direction.
The results were clear in the flush direction but not in the gap direction. In Paper A the analysis results pointed toward the same shape of the robust area for both flush and gap. When the analysis method had been refined, it showed that this was not correct. The gap results are dependent on the direction of the split-line, since the measures are performed perpendicular to the gap direction. In this simulation setup no split-lines were carried through. Thus no useful results in the gap direction were achieved.
Result 2 Due to useless results in the gap direction, an algorithm describing how to find the design of the geometrically most robust split-line between two geometries with given location schemes was developed. In Figure 18 the most geometrical robust split-line in the gap direction is shown as a black line on top of the flush results. The functionality allows designers to evaluate split-lines with respect to gap and flush simultaneously. The generated split-line design can also be judged regarding how well the split-line is following the design language.
23
RESULTS
Figure 18.
The most robust split-line in the gap direction on an automobile geometry.
Result 3 During the simulation procedure, the influence on the split-line design and the robust area caused by the locating schemes’ configuration appeared. The use of the colour-coding feature made it more easily interpreted. As can be seen in Figure 19, the position of locator A1-A3 affects the shape of the least sensitive area in the flush direction.
Figure 19.
The enhanced colour coding of flush with six different locating schemes.
The position and configuration of the locators also influenced the shape of the most robust split-lines in the gap direction; see Figure 20. The placement of the locators was fixed but the configuration of the B1-B2 locators is changed. A number of the split-lines divide the geometries in such a way that one of the geometries achieves two locators (under-constrained) and the other four (over-constrained). This issue has not been treated and needs to be resolved before implementing it in an industrial context.
24
RESULTS
B2
B2
B1
B1
B2
B2
B1
B1
B1
B1
B2
B1
Figure 20.
B1
B2
B2
B2
The configuration of the B1-B2 locators influences the design of the split-line.
25
26
DISCUSSION
5
DISCUSSION
In this chapter the research questions, results and contribution of the research project are discussed.
5.1 Research approach DRM has been the research methodology used in this project. Most of the research methodologies presented have been exposed to criticism, and so has DRM (Eckert et al., 2004). The criticism is directed towards the measurable criteria. It is important to use the measurable criteria with caution. But as long as the researcher realizes and understands what is measured, the measurable success criteria may be a valuable way of investigating the fruitfulness of the research. The focus of this research project has been on understanding the nature of a designed product which emphasizes the use of DRM, since the measurable criteria have been clear. It would have been of interest to compare different research approaches, but since the DRM is a fairly time-consuming methodology it will not be possible to test different research methodologies. Hence it is of importance to take time in the beginning of the research project to choose a methodology that is adequate for the research project. Since the design and placement of split-lines are influenced by both the directly measurable values such as measures in gap and flush directions and the VQA perceived by the user, different research methods have to be used. Simulations in CAT softwares have been used in order to investigate the geometrical robustness issues. General tolerancing rules and methods have also been applied. The area of visual robustness is closely connected with the perception of the user. Hence interview studies and user experiments have to be used to investigate the area of visual robustness.
5.2 Results in terms of research questions The research questions presented in Chapter 1.4 will here be in focus for the discussion. The research questions have been revised and refined during the research project. The results have so far been tested and validated in a prototype software. Main research question How shall the initial product geometry, with given location schemes, be split to achieve a geometrically and visually robust solution? This is a broad research question and a total and final answer has not been delivered at the present stage of the research project. The final answer lies within the answers to the four research sub-questions. The visually robust solution aspects of the questions have not been treated to any great extent; this is part of the future work. The research sub-questions have their foci on four different characteristics that the design and placement of the split-line should be treated against. In the following text these characteristics will be discussed.
27
DISCUSSION
How shall the split-line be placed and designed with respect to: Flush and gap Flush and gap are the two most commonly treated characteristics of split-lines, and are dealt with in Papers A and C. In Paper A the results of the robust area in the gap and flush directions pointed towards a similar result. But in Paper C it emerged that this was not the case. From the results presented, it can be seen that an optimal solution in the gap direction is not always optimal in flush. In order to create a solution that is geometrically robust in both gap and flush directions, there has to be, in many cases, a compromise. One must decide whether gap or flush can achieve this and is most important in a particular case, and let that characteristic be the governing one – or design and place the split-line so that both gap and flush become as good as possible at the same time. The CAT functionality or tool that has been presented in this thesis simulates and analyzes the geometries in the gap and flush directions, and those only. The intention with the tool was to find the most geometrically robust placement of a split-line but without taking away the design and aesthetic freedom when designing. The possibility to manipulate with the shape and placement of the split-line within the robust areas, given by the CAT software, simplifies the compromise between flush and gap. The designer then has the possibility to decide which, gap or flush, should govern the decision. But the tool does not treat any of the other influencing characteristics.
How shall the split-line be placed and designed with respect to: Geometry The geometry aspects have not been treated to any great extent, but the number of locators on each geometry, after performing the most robust split-line in the gap direction, has become an issue. When performing the geometrically most robust split-line in the gap direction, the solution sometimes splits the parts in such a way that one of the geometries achieves two locators and the other four. This implies that one of the geometries becomes under-constrained and the other over-constrained. In future CAT functionality, this has to be resolved in order to implement it and test it in an industrial environment.
How shall the split-line be placed and designed with respect to: Design language The final characteristic, design language, has been treated shallowly. In Paper C, the design language is treated in terms of the shape of the geometrically most robust split-line in the gap direction. The shape of the split-line was subjectively judged in terms of whether or not it followed the design language of the automobile.
5.3 Contribution The presented research has contributed to an increased understanding of geometrical and visual sensitivity of split-lines. A tool has been provided for simulation and analysis of geometries regarding the design and placement of the split-lines, so that these become as geometrically robust as possible in the gap and flush directions. During the analysis of the geometries it emerged in what way the locating schemes controlled the gap and flush results, although these results have not been analyzed further in the presented research. The research has also contributed greater knowledge concerning customers’ and users opinions with respect to split-lines.
28
CONLUSION
6
CONCLUSIONS
This chapter gives an overview of the thesis and sums up the presented results. In the early phases of product development work, some products consist of an undivided shell that needs to be divided into individual parts in order to be manufactured or to fulfil the intended functionality. In the automotive industry, this shell may be the automobile body. In the present licentiate thesis, research results have been described that clarify how to split the initial product geometry with respect to geometrical and visual sensitivity. The prerequisites have been set to meet the production climate of today, where many new automobile models are developed and produced with the same production line. This entails fixed locating schemes for the parts to be assembled. Since the CAT functionality presented here had the purpose of supporting both industrial and engineering design competences, the simulation and analysis results had to be adapted in a way that fitted both competences. The use of colour coding facilitates the interpretation of the evaluation results for the robust areas and allows design and aesthetic freedom for the designer. The simulation results showed that there is a difficulty in presenting the results of the most robust area between two geometries in the gap and flush directions in the same way. The nature of the measurement direction in gap causes a difficulty since it is measured in the direction normal to the split-line direction. This means that the split has to be carried through in order to be able to create the measure. The results imply that the split-line solution in the flush-direction has to be verified in the gap direction. The importance of how different locating schemes affect the shape and placement of the geometrically most robust split-line and the robust area was illustrated during the simulations. The area of visual sensitivity has not been treated directly to any great extent in this thesis. But visual robustness is strongly coupled to geometrical robustness, and the results from the latter area will hopefully in the long run be used in the visual robustness area. However, an interview study treating the area of visual sensitivity has been carried out. The results from that study showed that the user/customer considers high VQA of splitlines as one of the four most important factors for the total quality appearance of the automobile. The results also showed that there is a correlation between narrow split-lines in the gap and flush directions and the automobile brands that the user considered to have the narrowest split-lines. The narrowness of split-lines is of importance for the perceived quality, but there are other factors that influence the VQA of split-lines. As mentioned, besides the design and placement of many aspects, only the dimensional influence of split-lines and partly the aesthetic aspects have been considered during this research project. In order to achieve a high-quality product, all aspects have to be taken into account. By using the presented results, the solution will hopefully be more geometrically and visually robust, which contributes to increased product quality. By increasing the knowledge and proper tools for the product development process, the development lead time will probably decrease, which is of value.
29
30
FUTURE RESEARCH
7
FUTURE RESEARCH
In this chapter the future research in the area of geometrical and visual robustness for split-lines is presented. The future research contains both planned studies for this research project and research areas that is of high importance for increasing the knowledge in the area. Since this research project is following the DRM the next step will be to test and evaluate the presented tool in a product development environment at an automobile manufacturer. The evaluation will include both the test of the functionality of the tool but also to examine if the tool is actually supporting the creation of more geometrically robust solutions. The results presented in paper C showed that the locating schemes influence the shape and placement of the robust area and the shape of the geometrically robust split-line in gap direction. These results should be refined and most likely structured in a way that they can be used as guidelines with or without computer support. Only two of the four sub-research questions have been treated to any bigger extent. The area of how the geometry itself is influencing the robustness of the split-line needs to be examined. A planned activity is a study where a number of subjects shall judge split-lines at a number of basic shapes i.e. spheres, cubes, cylinders etc regarding the geometrical dimension and their parallelism. The result from that study can hopefully move the research one step closer to an answer to how the split-line shall be designed and placed with respect to the geometry. In the DRM the research was divided into two areas; the geometrical and visual robustness. The area of visual robustness has not been explored to the same extent as geometrical robustness in research society and it has been the same in this research project. In the meantime the visual sensitivity is very central and important for product quality. It is therefore a need for a more deeply evaluation of visual robustness when dividing the initial product geometry. The area of geometrical and visual robustness invites to several different research questions and research aspects which makes the research area interesting in the future. This research project will continue as mentioned which implies both prescriptive studies and descriptive studies in the area of geometrical and visual robustness.
31
32
REFERENCES
8
REFERENCES
Andreasen, M. M. and Hein, L. (2000). Integrated Product Development, R. Sharp, translator, The Institute for Product Development, The Technical University of Denmark, Lyngby, Denmark. Bergman, B. and Klefsjö, B. (1995). Kvalitet från behov till användning, Studentlitteratur, Lund, Sweden. Blessing, L., Chakrabarti, A. and Wallace, K. (1995). "A Design Research Methodology." Proceedings of ICED 95, Praha, the Czech Republic, pp. 50-55. Brahmst, E., Gerth, R., Baron, J. and Geddes, S. (2002). "Comparative Datuming and Hanging Strategies of Doors: A Benchmark Study." Proceedings of International Body Engineering Conference & Exhibition and Automotive & Transportation Technology Conference, Paris, France. Chase, K. W. and Parkinson, A., R. (1991). "A Survey of Research in the Application of Tolerance Analysis to the Design of Mechanical Assemblies." Research in Engineering Design, 3, pp. 23-37. Dagman, A., Wickman, C. and Söderberg, R. (2004). "A Study of Customers' and the Automotive Industry's Attitude Regarding Visual Quality Appearance of Split-Lines." Proceedings of 4th International Conference on Advanced Engineering Design AED 2004, Glasgow, Scotland. Dahlström, S., Hu, S. J. and Söderberg, R. (2002). "Identifying Variable Effects on the Dimensional Quality of Compliant Assembly, Using Computer Experiments." Proceedings of 28th Design Automation Conference, Montreal, Que., Canada, pp. 23-32. De Groot, A. D. (1969). Methodology : Foundations of Inference and Research in the Behavioural Sciences, The Hague, the Netherlands. Duffy, A., H, B and Andreasen, M., Myrup. (1995). "Enhancing the evolution of design science." Proceedings of International Conference on Engineering Design (ICED 95), Praha, Czech Republic, pp. 29-35. Eckert, C., Clarkson, P. J. and Stacey, M. (2004). "The Lure of the Measurable in Deign Research." Proceedings of International Design Conference-DESIGN 2004, Dubrovnik, Croatia. Gerth, R. J., Baron, J., Brahmst, E. and Geddes, S. (2002). "Comparative Dimensional Quality of Doors: A Benchmarking Study." Proceedings of International Body Engineering Conference & Exhibition and Automotive & Transportation Technology Conference, Paris, France.
33
REFERENCES
Hong, Y. S. and Chang, T.-C. (2002). "A Comprehensive Review of Tolerancing Research." International Journal of Production Research, 40(11), pp. 2425-2459. Hubka, V. and Eder, W. E. (1988). Theory of Technical Systems : a Total Concept of Technical Systems, Springer-Verlag, Berlin, Germany. Hubka, V. and Eder, W. E. (1996). Design Science : Introduction to the Needs, Scope and Organization of Engineering Design Knowledge, Springer, London. Johansson, C. (2005). CE Johansson homepage. Retrieved January 4, 2005, from http://www.cej.se/index_org.htm. Karlsson, M. (1996). "User Requirements Elicitation - A Framework for the Study of the Relation between User and Artefact," Doctoral dissertation, Chalmers University of Technology, Göteborg, Sweden. Kumar, S. and Raman, S. (1992). "Computer-Aided Tolerancing: the Past, the Present and the Future." Journal of Design and Manufacturing, 2, pp. 29-41. Kvale, S. (1996). Interviews : an Introduction to Qualitative Research Interviewing, Sage, Thousand Oaks, California, USA. Lee, D. J. and Thornton, A. C. (1996). "The Identification and Use of Key Characteristics in the Product Development Process." Proceedings of The 1996 ASME Design Engineering Technical Conferences and Computers in Engineering Conference, Irvine, California, USA. Maxfield, J., Dew, P. M., Zhao, J., Juster, N. P., Taylor, S., Fitchie, M. and Ion, W. J. (2000). "Predicting Product Cosmetic Quality in the Automobile Industry." Proceedings of 33rd International Symposium on Automotive Technology and Automation (ISATA), Dublin, Ireland. Monö, R. G. (1997). Design for Product Understanding : the Aesthetics of Design from a Semiotic Approach, M. Knight, translator, Liber, Stockholm, Sweden. Mørup, M. (1993). "Design for Quality," Doctoral dissertation, Technical University of Denmark, Lyngby, Denmark. Nationalencyklopedin. (2004). Homepage of Nationalencyklopedin. Retrieved November 9, 2004, from http://www.ne.se/jsp/search/article.jsp?i_art_id=164689. Pahl, G. and Beitz, W. (1996). Engineering Design : A Systematic Approach, K. Wallace, translator, Springer, Berlin, Germany. Phadke, M. S. (1989). Quality Engineering Using Robust Design, Prentice Hall, Englewood Cliffs, N.J. USA.
34
REFERENCES
Prisco, U. and Giorleo, G. (2002). "Overview of current CAT systems." Integrated Computer-Aided Engineering, 9(4), pp. 373-387. Renault. (2005). Homepage of Renault. Retrieved January 13, 2005, from www.renault.se. Roozenburg, N. F. M. and Eekels, J. (1995). Product Design Fundamentals and Methods, Wiley, Chichester, West Sussex, England. Söderberg, R. (1993). "Tolerance Allocation Considering Customer and Manufacturer Objectives." Proceedings of Proceedings of the 19th Annual ASME Design Automation Conference. part 2 (of 2), Albuquerque, NM, USA, pp. 149-157. Söderberg, R. (1994). "Robust Design by Tolerance Allocation Considering Quality and Manufacturing Cost." Proceedings of Proceedings of the 1994 ASME Design Technical Conferences. Part 1 (of 2), Minneapolis, MN, USA, pp. 219-226. Söderberg, R. (1995). "Optimal Tolerance Band and Manufacturing Target for Monotonic Loss Functions with Functional Limits." Proceedings of Proceedings of the 1995 ASME Design Engineering Technical Conferences, Boston, MA, USA, pp. 345-352. Söderberg, R. (1998). "Robust Design by Support of Cat Tools." Proceedings of ASME Design Engineering Technical Conferences, Atlanta, Georgia, USA, p. 260. Söderberg, R. and Carlson, J. S. (1999). "Locating Scheme Analysis for Robust Assembly & Fixture Design." Proceedings of ASME Design Engineering Technical Conferences, Las Vegas, Nevada, USA, p. 251. Söderberg, R. and Lindkvist, L. (1999). "Computer Aided Aassembly Robustness Evaluation." Journal of Engineering Design, 10(2), pp. 165-181. Söderberg, R. and Lindkvist, L. (2001). "Automated Seam Variation and Stability Analysis for Automobile Body Design." Proceedings of 7th CIRP International Seminar on Computer Aided Tolerancing, ENS de Cachan, France, pp. 255-264. Söderberg, R. and Lindkvist, L. (2002). "Stability and Seam Variation Analysis for Automotive body design." Journal of Engineering Design, 13(2), pp. 173-187. Taguchi, G. (1986). Introduction to Quality Engineering: Designing Quality Into Products and Processes, Productivity Inc. Taguchi, G., Hsiang, T. C. and Elsayed, E. A. (1989). Quality Engineering in Production Systems, McGraw-Hill, New York, USA. Trost, J. (2001). Enkätboken, Studentlitteratur, Lund; Sweden.
35
REFERENCES
Ullman, D. G. (2003). The Mechanical Design process, McGraw-Hill, New York, USA. Ulrich, K. T. and Eppinger, S. D. (2004). Product design and development, McGraw-Hill, Boston, USA. Warell, A. (2001). "Design Syntactics: A Functional Approach to Visual Product form Theory, Todels and Methods," Doctoral, Chalmers Tekniska Högskola, Göteborg, Sweden. Westlander, G. (2000). "Data Collection Methods by Question-Asking. The Use of SemiStructured Interviews in Research." TRITA-MMK 2000:8, Integrated Product Development, KTH, Stockholm, Sweden. Wickman, C. and Söderberg, R. (2001a). "Defining Quality Appearance Index Weights by Combining VR and CAT Technologies." Proceedings of 2001 ASME Design Engineering Technical Conference and Computers and Information in Engineering Conference, Pittsburgh, PA, USA, pp. 1215-1224. Wickman, C. and Söderberg, R. (2001b). "Towards Non-Nominal Virtual Geometric Verification by Combining VR and CAT Technologies." Proceedings of 7th CIRP Seminar on Computer-Aided Tolerancing, Cachan, France. Wickman, C. G. and Söderberg, R. (2003). "Comparison of Non-Nominal Geometry Models Represented in Physical Versus Virtual Environments." Proceedings of 2003 ASME International Mechanical Engineering Congress, Washington, DC, USA, pp. 2937. Yin, R. K. (1994). Case Study Research : Design and Methods, Sage, Thousand Oaks, CA. USA.
36
APPENDED PAPERS
Paper A Dagman, A. and Söderberg, R., (2003), Virtual Verification of Split-Lines with Given References. Paper presented at the 14th International Conference in Engineering Design, ICED03, Stockholm, Sweden, 18-20 August 2003.
PAPER A
INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN ICED 03 STOCKHOLM, AUGUST 19-21, 2003
VIRTUAL VERIFICATION OF SPLIT LINES WITH GIVEN REFERENCES Andreas Dagman and Rikard Söderberg
ABSTRACT Product development of today is characterized of shortened lead time, increased expectations on the products and lack of time. The importance of the aesthetic design has increased and many brands are no longer competing only with performance but with the design of the products. This paper focus on the possibility to optimize relations between parts based on tolerance aspects and a fixed set of positioning schemes. A tool has been proposed to optimize split lines between parts based on tolerance aspects and a fixed set of locating schemes. The possibility to evaluate design solutions is of high importance. The exterior design of many products involves a number of parts who are in relation between one another. In the automotive industry a big share of the quality aspects is assessed by these relations. The effect of visual and geometrical sensitivity may be evaluated in early stages of the product development process using non-nominal models. Thus time and cost reduction will be met and the visual quality will increase. The tool can also be used as border crossing technical aid for both engineering design and industrial design to enable styling concept evaluation.
Keywords: evaluation of design, predictive design analysis, robustness.
1. Introduction The exterior design of many products involves a number of parts which are in relation to each other. The visual appearance of these relations has a big aesthetic and quality influence on the product. In the automotive industry a big share of the quality aspects are prescribed by these relations. A way to use virtual environments to verify and predict quality appearance has been suggested by [1], [2]. The possibilities to, in an early stage of the product development process, virtually verify a concept to get a predictive design analysis, has become an important factor to cut the expenses in the product development phase. The traditional approach of applying robust design in this area has been to find the optimal locator positions that minimize variation and make the assembly less affected by the tolerances of the included parts, for example [3] and [4]. This way of optimizing the robustness of the design assumes the freedom to place locators wherever the optimal geometrical location is. However, often in production environment of today, the assembly strategy, i.e. how parts are located and assembled with fixtures and/or robots, is more or less fixed to a number of principle layouts. Product development of today is characterized of shortened lead time, increased expectations on the products and lack of time. On top of this, persons in teams with different educational background and skills are trying to achieve the goal in their way. Concurrent engineering (CE) which has been described by [5] is one way to deal with this
A-1
PAPER A
complex system. The possibility to analyse and evaluate a solution at an early stage of the product development process is invaluable. The result from the evaluation must be possible to interpret regardless of educational background.
2. Robust design Many products of today are built up of several components. These are arranged in a way that makes them more or less geometrically dependent of other parts in the assembly. Geometrical variation is introduced to a product as component variation and assembly variation. The usual procedure at companies is that the suppliers control the component variation and the assembly variation is taken care of in-house by the assembly process. The final variation is dependent of the sensitivity of the concept. In a sensitive concept, component and assembly variation is amplified, whereas in a robust concept the variation is suppressed. Since the shape of the parts and the placement of the locators govern the robustness of the product, it is very important to achieve as high geometrical robustness as possible already in the concept phase. A CAT (Computer Aided Tolerancing) software, see [6] can be used for this purpose. Another alternative can be early robustness analysis. These tools are used in trying to foresee and avoid geometrical problems that are related to geometrical variation. They can also help out to create a robust design, insensitive to geometrical variation. The relation between robustness and variation can be explained with a beam and a support, see figure 1. Concept robustness is controlled by the relation between input and output relation. Output
Input
Sensitive
Robust
Figure 1. A beam support explaining robustness and variation.
Depending on the placement of the support the input variation will either increase or decrease the output variation. If the support is moved to the left the input variation will lead to an amplified output variation. And if the support is moved to the right the output variation will not be as affected as in the first example and the robustness will increase. This means that input and output are depending on each other and the placement of the support. This relationship can be expressed as following: robustness variation
= x 0 x x
locators tolerance
Since the position of the locator controls two important product characteristics, this should be treated first. Based on final requirements for the output variation and known sensitivity (relations between input and output), the tolerance for the input variation may then be determined.
A-2
PAPER A
In the automotive industry a quality aspect is brought in to this area. The relationship between the doors, hoods and panels are important from this point of view. The possibility to make the split lines tighter and more accurate gives the impression of a good craftsmanship. The splits are actually something measurable which is important to be able to give a grade of the car. This impression is very important in the competition between different brands and models. These relations can also be looked at from an aesthetic point of view. The importance of working with the visual product form is discussed in [7]. These relations are one of the parts that underlie the Quality Appearance (QA) Index, see [8]. The QA Index measures the over all quality appearance of a product using statistical simulations.
2.1. The seam function The seam function was introduced in [9] as the relation between two parts over a specified distance and describes the most frequently used quality characteristics for evaluations of geometrical variation in automotive body design. Typically, seam variation is measured and evaluated in two directions, the gap and flush directions. To be able to efficiently evaluate flush and gap along the seams of an automotive body, they must be generated more or less automatically in the CAT environment. For that purpose an algorithm for automatic seam generation is developed and used for seam variation evaluation. The algorithm uses points and lines to create seams, see figure 2. Several chaining options and criteria are then available. The measure used in this experiment has been point-topoint measures which mean that the seam function creates measures between two points, one on each part.
Figure 2. Seam variation, flush direction [9].
2.2. The 3-2-1 Locating scheme All parts in an assembly are positioned with a positioning scheme. A frequently used scheme in the automotive industry is the 3-2-1 system, where six theoretical points are used to lock six degrees of freedom for a part, three translations and three rotations, see figure 3. Three points forms a plane, A1, A2 and A3, which locks two rotations and one translation. Two points, B1 and B2, forms a line that locks one rotation and one translation. The final point, C1, locks the remaining translation. The locating points are represented in reality by physical locators such as planes, holes and slots. This is the foundation how to place components in an assembly structure.
A-3
PAPER A
Z B2 C1
Y
X
B1 A1
A3 A2 Target 3-2-1
B2 B1 A1
C1 A3
A2 Object 3-2-1
Figure 3. The 3-2-1 locating scheme.
3. Split line optimization This paper focus on the possibility to optimize relations between parts based on tolerance aspects and fixed sets of locating schemes. The overall question in this work is how the initial product geometry shall be split into parts in the most robust way, i.e. how to achieve the most robust product architecture with respect to placement of split lines? The split line, the parallel edges, between two plane plates, will be analysed, see figure 4.
z
y x
Figure 4. A split with measures in gap and flush directions.
The distance between all edges shall be equal for all relations in a product, to achieve an aesthetically well balanced product structure. The distances in foci are in the plane (gap) and in the normal to the plane (flush). A change from this is going to lower the total customer and producer impression of the product. An optimal relation between the two plates, from a geometrical and tolerance point of view, shall consist of two parallel edges with minimum variation in flush and gap direction, for a given set of locating schemes.
4. The experiment set up Optimizing the relation between surfaces/plates from a tolerance point of view has so far in most works presented implied finding the optimal location of locators. In this paper the
A-4
PAPER A
prerequisite has been changed. We are here going to fix the position of the locators and determine where the best suited area for a split line is located. This way of working can be seen as a complement to the traditional way of optimizing locator points. The possibility to find the best suited area between two parts for a split line with given locator position may be a good help in the early stages of the product development process, especially when evaluating and developing “styling” concept with respect to geometrical variation. The models being used in the evaluation phase are non-nominal. A CAT software, in this case RD&T, is here used as a workbench. Two parts, shaped as rectangles, were modelled and placed on top of each other, to be able to use the seam function. The locators were placed at the right side of one part and on the left side on the other. In RD&T the user has different possibilities to define positioning systems. In this case a 3-point system was used to position the parts.
4.1. The 3-point system The 3-point system is similar to the 3-2-1 system with the difference that only three points are used. The first locates the part in three directions, the second in two and the third in one direction. In figure 5 the position of the locators and their locating directions are shown. All the locators are able to move in the Z-direction and the arrows shows the additional displacement possibilities.
Figure 5. The grid of seams on the two plates. The arrows show the location and the freedom of movement of the locators.
4.2. The seam grid and colour coding In this experiment set up, a number of seams were created to be able to measure between the parts. The parts used in this research set up were 2D plates; however the method can be used on other types of more complex surfaces. The seams changed both in angle and direction see figure 5. The goal with the seam direction and angles were to create a closer grid in the center and a more spread grid at the edges. The reason behind this is the fact the two parts shall be divided in the Y-direction with three locators on each part and the most interesting area will be in the centre of the plates. A number of seams with an angle of 45 degrees were set up and number of seams with different angles running through the centre point was also added. The measurement of the area can also be done by using a number of points in pairs, one from each part. By adding a large number of point pairs on the two parts the same
A-5
PAPER A
result may be achieved. During this experiment, the seam function has been the only method used. When all the seams had been created a variation analysis, using Monte Carlo simulation technique were performed to analyze variation in the measures All the measures could then be simulated and visualized with colour coding of variation or amplification. By using the colour coding feature in RD&T, areas that had less measure variations were detected (coloured in blue, the darker area). Different measure interval creates different colours. The seam grid and the colour coding made it possible to find patterns on the parts. A big advantage by use of colour coding is that the function gives a simple overview so that different competences easier can understand the results and have opinions about them. During the tests, only the locators were affected by tolerance changes, the shape of the parts themselves did not change. The results were shown both in gap and flush directions. To be able to see the placement influence of the locators, the position of the points were altered. The location points were alternated between four positions on each plate. The number of points and placement locations made a number of set up possibilities to be solved. The results were analyzed to see if patterns were possible to find.
4.3. Evaluation of flush First the results were analyzed from a flush point of view. As can be seen in figure 6 we were able to see areas that were less affected by the measure variation.
Figure 6. The colour coded results for flush.
Depending on the placement of the locators different patterns were created when the colour coding were applied.
A-6
PAPER A
4.4. Evaluation of gap When observing the results for gap directions almost the same patterns as in flush were found, however not as clear. Because the gap measures are perpendicular to the side of the part the direction of the split lines also influenced the results. As can be seen in figure 7, the vertical seams have lower variation compared to the ones with an angle. When increasing the angel of the seam against the horizontal plane, the variation increases as well. High variation takes place in the outskirts of the parts.
Figure 7. The colour coded results for gap.
4.5. Discussion of flush and gap evaluation Some differences between flush and gap evaluation results can be noted. The main differences were the angle influence in the flush direction. The grid has been applied to a schematic picture of a car, see figure 8, to show a simple picture how the tool can be used.
A-7
PAPER A
Figure 8. The seam grid applied on a car structure.
By using the variation results from the grid of seams, the minimum measure variation in a number of levels in the Y direction were found. The minimum measure variation locations were then marked at the plates and linked together to create a line, as can be seen in figure 9. This line shows the optimal way, from a tolerance point of view, to split two parts with given locators position. The optimizing of the split line can be described by an algorithm for automatic split line optimization.
Figure 9. The ultimate split line orientation according with given locator positions.
A-8
PAPER A
This tool has the possibility to become a good help for designers in the early stages of the product development process. To be able, in early stages of the product development process, determine in what area the split line shall be placed, to make the product less affected by tolerances will decrease the cost and time consumption.
5. Conclusions In this paper we have proposed a method to optimize relations between surfaces/plates based on tolerance aspects and a fixed set of locating schemes. The possibility to, in an early stage of the product development process, be able to evaluate design is of high importance. With a tool as the purposed the effect of visual and geometrical sensitivity may be evaluated in early stages of the product development process, using non-nominal models. Thus time and cost reduction will be met and the visual quality will increase. The tool will give an insight of possible solution changes of the geometrical shape. By receiving proposal for the most tolerance independent solution, alternatives can be found quickly which will improve the product architecture. It will also be used as border crossing technical aid for both engineering design and industrial design to enable styling concept evaluation. The result may be implemented in a CAT system that can give proposals on both reference locations and geometry changes, when using locked reference locations. This will increase the total solution quality of the final outcome.
6. References [1] Maxfield, J., Dew, P. M., Zhao, J., Juster, N. P., Taylor, S., Fitchie, M., Ion, W. J. and Thompson M., “Predicting Product Cosmetic Quality using Virtual Environment” Proceedings Of 2000 ASME Design Engineering Technical Conferences and Computers and Information in Engineering Conference, Baltimore Maryland, September 10-13 2000. [2] Maxfield, J., Dew, P. M., Zhao, J., Juster, N. P., Taylor, S., Fitchie, M. and Ion, W. J., “Predicting Product Cosmetic Quality in the Automobile industry” Proceedings of the 33rd international Symposium an Automotive Technology and Automotation (ISATA 2000), Dublin, Ireland September 25-29. [3] Camelio, J. A., Hu, S. J. and Ceglarec, D. J., “Impact of Fixture Design on Sheet Metal Assembly Variation”, Proceedings of the 2002 ASME Design Engineering Technical Conferences and Computers and Information in Engineering Conference, Monteral, Canada, September 29 –october 2, 2002, DETC2002/DFM-34167. [4] Söderberg, R. and Lindkvist, L., “Computer aided assembly robustness evaluation”. Journal of Engineering Design, Vol. 10, No. 2, 1999 pp. 165-181. [5] Vos, R. G., “The emerging basis for multidisciplinary concurrent engineering, In E. J. Haug (Ed), Concurrent Engineering: Tool and Technologies for Mechanical System Design”, Springer Verlag. 1993, pp. 111-127.
A-9
PAPER A
[6] Söderberg, R. and Lindkvist, L., “Two-Step Procedure for Robust Design Using Cat Technology”, Proceedings of the 6th CIRP International Seminar on Computer Aided Tolerancing, Enschede, the Netherlands, Mars 22-24 1999. [7] Warell, A., “Design Syntactics: A functional Approach to Visual Product Form”, Doctoral Thesis, Chalmers University of Technology, ISBN 91-7291-101-8, Göteborg, Sweden, 2002. [8] Wickman, C. and Söderberg, R., “Increased Concurrency Between Industrial and Engineering Design using CAT Technology Combined with Virtual Reality”, Proceedings of Advances in Concurrent engineering, CA, USA, July29-August 1, 2001. [9] Söderberg, R. and Lindkvist, L., “Automated Seam Variation and stability Analysis for Automobile Body Design”. 7th CIRP International Seminar on Computer Aided Tolerancing, ENS de Cachan, France April 24-35 2001.
A-10
Paper B Dagman, A., Wickman, C. and Söderberg, R., (2004), A Study of Customers' and the Automotive Industry's Attitude Regarding Visual Quality Appearance of Split-Lines. Presented at Advances in Engineering Design, AED2004, Glasgow, Scotland, 5-8 September 2004.
PAPER B
A STUDY OF CUSTOMERS’ AND THE AUTOMOTIVE INDUSTRY’S ATTITUDE REGARDING VISUAL QUALITY APPEARANCE OF SPLIT LINES M.Sc. Andreas Dagman and Professor Rikard Söderberg Product and Production Development Chalmers University of Technology, SE-412 96, Göteborg, Sweden
Lic. Eng. Casper Wickman Volvo Car Corporation Body and Trim Engineering Dept. 93910, PVS 3,5 SE-405 31 Göteborg, Sweden
ABSTRACT The exterior design of an assembled car body consists of a number of components, which are in spatial relation to each other. These relationships have a significant impact on the overall perceived Visual Quality Appearance, (VQA) of a car body exteriorly. The automotive industry has the ambition to produce products with high VQA. For companies, which are producing products composed of several components, it is costly and time-consuming to consider and control the impact of geometric variation. This paper investigates the industry’s and the customers’ attitude regarding VQA of split lines, by use of an emperical study. It also investigates if there is a correrlation between the car brand that customers think possesses the best VQA regarding split lines and measurement data of the geometrical variation of split lines of those car brands. KEYWORDS: Empirical study, Visual Quality Appearance, Split Lines, Tolerance analysis
1.
INTRODUCTION
The exterior design of an assembled car body consists of a number of components, which are in spatial relation to each other. These relationships have a significant impact on the overall perceived Visual Quality Appearance, (VQA) of a car body exteriorly. In this paper, VQA is defined as the quality impression that a product conveys visually to a customer when observing it. These relationships can be between doors, bonnets, wings and other panels on the car, see figure 1, and are in the automotive industry called split lines. Little research has been published that investigates customers’ opinions about VQA of split lines. A split line is defined as the relation between two mating parts over a specified distance. Split lines can be the result of two mating parts or just a hollow in a part. There are a wide range of parameters that affect the VQA of a split line. It could be quantitative parameters like the actual distance of the gap or the flush or nonparallelism. It could also be qualitative parameters like the curvature of the split line, what is visual beneath the split line, colour of surfaces, the design, etc. A gap is defined as the distance perpendicular to the normal surface between two parts and flush is defined as the distance on the axis of the normal surface between two parts. The size of the gap and flush measures, distances of a split line, are affected by variation, either part variation or assembly variation.
Figure 1. A number of split lines. According to the automotive industry they has the ambition to produce products with high VQA. Since the automotive industry holds the belief that the distance of the flush and the gap has an impact on VQA the trend has been to minimise the size of gap and flush. For companies, which are producing products composed of several components, it is costly and time-consuming to consider and control the impact of geometric variation. In the automotive industry, substantial effort is put into developing methods and tools
B-1
PAPER B
for evaluation of VQA of split lines since a considerable share of the quality appearance is judged by these relationships, see [1].
1.1
THE FIELD OF TOLERANCE MANAGEMENT
The area of tolerance management originates from plus/minus tolerances in the early 1900's. It has been one of the most important issues for engineers involved in the product realization process. Tolerance management influences all phases of the design process from the early design phase all the way to the production line. Tolerance management is a large research area and it covers several different subdisciplines, see figure 2. Research in tolerancing
Tolerancing schemes
Tolerance modelling and representation
Tolerance specification
Tolerance Analysis
Tolerance synthesis or allocation
Tolerance evaluation
Tolerance transfer
Figure 2. The field of research in tolerancing [2]. Tolerance analysis is a research area and a method to verify the proper functionality of a design, taking into account the variation and tolerances of the individual parts. The area has been the topic of a large number of research publications during the years, see [3]. Variation in critical product dimensions has its roots in three areas, namely component variation, assembly variation and the robustness of the concept, see figure 3. C om ponent variation M achine precision
M anufacturing process
Process variation
A ssem bly variation A ssem bly precision
Process variation
A ssem bly process PK C variation R obustness
D esign concept
Figure 3. Geometrical Product Key Characteristics variation contributors [4]. The area of tolerance allocation is a method to allocate assembly function tolerances to the individual component tolerance. A branch from tolerance allocation is the quality engineering area, [5]. The branch foci on an overall quality control in which every activity involved in production is controlled, called robust design. Suh [6] described the uncoupled design which also affects the assembly variation. Hu [7] present the “stream-of-variation” theory for automotive body assembly. The possibility to evaluate quality appearance and variation of geometrical- and assembly variation in Virtual Reality has been discussed in [8] and [9; 10].
1.2
ROBUSTNESS AND VARIATION
The geometrical aspects of an assembly can be judged by two characteristics, its robustness and its variation in critical dimensions. The level of robustness describes the ability an assembly has to suppress input variation. For the producer, high robustness is importance since it reflects whether it will be easy or difficult to meet the product requirements. The variation in critical dimensions is what the customers see in the end and is controlled by the robustness of the concept and by the input variation.The relation between robustness and variation can be explained with a beam and a support, see figure 4. Depending on the placement of the support the input variation will either increase or decrease the output variation. If the support is moved to the left the input variation will lead to an amplified output variation. If the support is moved to the right the output variation will not be as affected as in the first example and the robustness will increase.
B-2
PAPER B
Output
Input
Sensitive
Robust
Figure 4. Figure explaining the relation between variation and robustness. The output variation is controlled by two parameters, the position of the support and the input variation. These relations can be described in matrix form as in figure 5. On the basis of this reasoning it is clear that, since the position of the locator controls two important product characteristics, this aspect should be treated first. Then, based on final requirements for the output variation and known sensitivity (relation between input and output), the tolerance for the input variation should be determined.
⎡robustness ⎤ ⎡ x 0 ⎤ ⎡locators ⎤ ⎢variation ⎥ = ⎢ x x ⎥ ⎢tolerances ⎥ ⎣ ⎦ ⎣ ⎦⎣ ⎦ Figure 5. Robustness and variation
1.3
LOCATING SCHEMES
There are a number of different ways to locate parts, i.e. locating schemes. One common way is the 3-2-1 system where 6 theoretical points are used to lock the six degrees of freedom, three translations and three rotations, see figure 6. The first three points form a plane that locks two rotations and one translation. The next two points form a line that locks one rotation and one translation and the last point locks the resulting translation. The positioning points are realised by holes, slots, screws for example. Likewise is there six locator points at a target locating scheme.
Z
B2 C1 Y
X
B1 A1
A3 A2 Mating 3-2-1
B2 B1
C1 A3
A1
A2 Object 3-2-1
Figure 6. The 3-2-1 locating scheme.
1.4
AIM OF THE PAPER
This paper investigates whether the achievements within the automotive industry to produce products with a high VQA of split lines is something that customers regard as important for the overall VQA of the exterior of a car body. Furthermore, an investigation to find whether a correlation between small gaps and flushes and customers' impression of the VQA exists.
B-3
PAPER B
2.
METHOD
A study has been carried out in order to collect empirical data from both industry and customers regarding VQA of split lines among premium cars. VQA is an important aspect, particularly in the premium car segment and was therefore chosen. Nineteen different brands were selected from the premium segment in order to be used as a basis for data collection. The industry is providing an expensive, complicated aspect to produce and also a non-functional aspect of a product that customers might not demand. The two interview groups; the Industry and Customer were divided into sub-groups, see figure 7. Empirical data was collected in both Germany and Sweden. All data was collected with structured interviews except the data that was collected from the automotive industry where semi-structured interviews were conducted. The questionnaires for the interviews were prepared in accordance with [11] and [12]. Since the results in a qualitative study are dependent on the source of data and the method for data collection, it is preferable to use multi-data sources, also referred to as triangulation [13]. The results were analysed in a quantitative way. The total number of interviewees was 130. The large number of interviewees increases the validity of the results. The numbers of interviewees from each category were distributed in accordance with figure 7. The questions were adjusted to the interviewees depending on what group the interviewee belonged to.
Industry Stand staff from different car brands at a large motor show Designers/industrial designers at a large automotive company Car dealers
Customers Visitors at a large motor show
The automotive industry; 14
Visitors at the motor show ; 50
Car dealers; 22
General public from Sweden Stand staff; 19 General public from Sw eden; 25
Figure 7. The distribution of the interviewees in the study.
3.
RESULTS
The results have been divided into two main categories: Firstly, whether VQA of split lines is an important aspect as concerns overall exterior VQA. Secondly, the relationships between small gaps and flushes and perceived high VQA.
3.1
IS THE VQA OF SPLIT LINES AN IMPORTANT ATTRIBUTE FOR CUSTOMERS?
An open question was posed to those interviewed in order to receive information of whether split lines were perceived as an influencing factor for VQA. The question was as follows; “Visually, what factors give the exterior appearance of a car high quality?” This was the first question asked and it is also an open type of question. The result from this question is presented in figure 8. The interviewees were allowed to give more than one example of influencing factors. This question was also posed to engineering designers and industrial designers in the automotive company. The question was part of a deeper and more extensive semistructured interview that was conducted in a study presented in [14]. Since that study focuses on VQA evaluated in virtual environments, the answers might be misleading. Hence, the results from this data source have been excluded in this question. The interviews in Germany were conducted in English. The lack of language skills from some of the interviewees forced us to compile and present some responses that had the same meaning but were expressed by different words. Three of the influencing factors are a collection of several similar answers. The factor “Design” is a compilation representing the terms, design, styling and form. The factor “Feeling” is a compilation of several subjective factors, such as, sporty, masculine, elegant, etc. “VQA of a split line” is a compilation of fit and finish, good fit, geometries integrated in the form, etc.
B-4
PAPER B
80
Percentage (%)
70 60
Customer
50
Industry
40 30 20 10
C hr om e
D et ai ls
R im s
Si ze
Br an d
lin e
M at er ia l
sp lit
Fe el in g VQ A
of a
C ol ou r
D es ig n
0
Figure 8. Response frequency to “Visually, what factors give the exterior appearance of a car high quality?”. In the automotive industry expressions, such as, fit and finish or gap and flush are used when discussing split lines. If, for instance, an interviewee answered both design and form, this was registered as one contribution to the factor “Design”. “VQA of a split line” turned out to be one of the main influencing factors for the overall VQA of a car. To find out if the VQA of a split line affects the buyer when purchasing a car, a closed question was posed: “In general, do you think that good fit and flush is important when choosing a car?”. The results can be seen in figure 9. The bars; Customer (total) and Industry (total) in the two graphs represent the total response frequency of the interviewees. During the interviews with the car dealers and the stand staff it emerged that VQA of a split line is not a common sales argument among the brands. It also emerged that there are not many customers asking the representatives about VQA of a split line. C u s to m e r s 80 Y es
70
No
Percentage (%)
60 50 40 30 20 10 0 Cus tomer (total)
V is itors at the motor s how
General public f rom Sw eden
In d u s t r y 100 90
Y es
80
No
Percentage (%)
70 60 50 40 30 20 10 0 In d u s tr y ( to ta l)
S ta n d s ta f f
Ca r d e a le r s
Th e a u to mo tiv e in d u s tr y
Figure 9. Response frequency to “In general, do you, think that good fit and flush is important when choosing a car?” 3.1.1 FINDINGS The first four factors were the same for Customers and the Industry. They were Industrial design, Colour, Feeling and VQA of a split line. As can be seen in Figure 4, there is quite a large difference between the values for the Industry and Customers. One part of the difference could be that the Customers mentioned a larger number of different factors, roughly 30, whereas the Industry mentioned roughly 15 factors. VQA of a split line was placed third among the Industry and fourth among Customers. This indicates that VQA of a split line is a factor that affects the overall VQA of car. More than 70% of Customers and the Industry find
B-5
PAPER B
that VQA of split lines is an important factor when choosing a car. This further emphasises the conclusion that VQA of split lines is an important factor overall.
3.2
IS THE DISTANCE OF GAP OR FLUSH IMPORTANT FOR HIGH VQA OF SPLIT LINES?
During the study, the brand that customers think possesses the best VQA regarding split lines has also been investigated. “What car brand do you think has the best VQA regarding split lines?” was posed to all customers. The purpose was to find what car brand or automotive company that has products that correspond to customer’s idea of high VQA of a split line. The results have been compared with measurement data of the geometrical variation of split lines for all car brands that were mentioned by the interviewees during the interviews. The results might indicate whether it is the quantitative parameters, such as, the distance of the gap or flush or if it is qualitative parameters, such as, how the split lines are integrated in the design that have the major impact on perceived VQA of a split line. 3.2.1 CHART INTERVIEW RESULTS WITH MEASUREMENT DATA In figure 10 below, the results from interviews with visitors from the automotive show and the general public in Sweden are shown. The figure shows the most frequently mentioned brands. Furthermore, five other brands where also mentioned but are not represented in the figure since they were only mentioned once. Car brands one to five and eight are all European and six and seven are Japanese. In figure 10, it can clearly be seen that a majority of interviewees have answered that they do not have any idea of what brand that has the best VQA of the split lines. It can also be seen in the figure that brands one to three are clearly the most mentioned. Automotive companies measure competitive brands in order to make competitive analysis of split line dimensions. The measurement data used in this study has been measured and compiled by the Ford Motor Company. The data has been gathered from two saloon models of each brand from model years between 1999 and 2003 except brands mentioned by less then 3% of the interviewees. For those brands measurement data from only one model has been used. In table 1 below, the results from the interviews has been compared with measurement data. In total, 13 different brands were mentioned during the interviews and for those brands measurement data has been analysed. 45% 40% 35%
Percentage (%)
30% 25% 20% 15% 10% 5% 0% No idea
1
2
3
4
5
6
7
8
Figure 10. Response frequency to “What car brand do you think has the best VQA regarding split lines?”
B-6
PAPER B
Table 1. Interview results charted against measurement data. R a t in g a c c o r d in g t o m e a su r e m e n t da t a
µµ
µµ + 6 σ µ
µ
µµ + 6 σ µ
1
5
2
7
7
2
2
5
8
3
3
7
7
1
1
4
4
O th er
4
4
5
O th er
4
2
2
6
8
1
O t h er
6
7
1
O th er
3
O th er
8
O th er
3
6
8
9
3
8
5
O th er
10
O th er
O th er
O t h er
5
11
O th er
O th er
O t h er
O th er
Gap
F lu sh
12
6
6
O t h er
O th er
13
O th er
O th er
O t h er
O th er
The mean value (µ) and six sigma (6σ) have been calculated for all split lines based on 30 replicates and up to 300 measurements points on each car measured. For all models the mean (µµ) of all mean values and the mean of all six sigma (6σµ) of the split lines have been calculated. Finally the sum (µµ+6σµ) has been calculated for both flush and gap. Hence, for flush and gap two good values based on all the exterior split lines for two models for each brand have been calculated. Car brands rated lower than eight by Customers has been marked as “Other” in table 1. The grey shaded boxes in table 1 represent brands that have received a higher rating by Customers than the actual variation of gap and flush measured. 3.2.2 FINDINGS In general, it is quite clear that many of the brands rated between one to seven by Customers also have small gaps and flushes. The distribution of the results also shows that the lowest rated brands also have wider gap and flush, i.e., the majority of “Others” are placed at the bottom half of table 1. This indicates that in most cases, the actual size of the gap and flush has an impact on the perceived VQA of split lines. Table 1 also shows that those brands that were rated in the range eight to thirteen actually have smaller geometric variation than many of those brands that were graded in the range one to seven. Furthermore, it can also be seen that brand one had a superior position among customers, actually did not reach any of the top positions regarding measurement data. This indicates that there are other factors that have a greater impact on the VQA of split lines than just the size of gap or flush. The grey shaded boxes indicate that half of the brands have been overrated as concerns actual variation. This also highlights that other aspects than just the size of the gap or flush have a significant impact on the VQA.
4.
DISCUSSION
Companies that are producing products composed of several components put financial and time-consuming efforts into governing geometric variation. Although high VQA of split lines do not have any functional influence on the product it is considered as one of the four most important aspects for the overall VQA of a car by customers. This finding is based on the first question posed. The interviewees had no knowledge about the background of the study other than that it investigated VQA of cars. This open question was chosen for the beginning of the interview with the intent of really finding the non-coloured opinion and to increase the reliability of this finding. The results show that high VQA of split lines is an important aspect for customers when comparing different brands. It seems like the industry's ambition to create cars with high VQA of split lines is good and motivated, since over 70% of the Customers interviewed considered it to be an important aspect. Four out of the eight highest rated brands were German. This could possibly be a biased result as two out of three of the interviews with Customers were conducted in Germany. However, in the meantime two out of three brands with lowest variation for both gap and flush were also German. It should also be mentioned that the automotive industry sometimes purposely creates under-flush in order to hide effects that occur due to variation. This fact might have had an influence on the flush measurement data and hence the results in table 1. In the study the overall opinion has been found to be that high VQA of split lines is something that the customer can expect from a number of specific brands. Customers take high VQA of
B-7
PAPER B
split lines for granted when buying premium cars and thus, this is something that not is emphasised although it is important for customers.
ACKNOWLEDGEMENTS The authors like to acknowledge Swedish Foundation for Strategic Research through the research program ProViking for financial support and the former research school ENDREA. Moreover, the authors would like to thank Volvo Car Corporation for sharing measurement data and empirical data in the form of interviewees.
REFERENCES [1] Söderberg, R. and Lindkvist, L.: Stability and Seam Variation Analysis for Automobile Body Design. Journal of Engineering Design, 13(2), 2002, p. 173-187. [2] Hong, Y. S. and Chang, T. C.: A comperehensive review of tolerancing research. Journal of Production Research, 40(1), 2002, p. 2425-2459. [3] Chase, K., W., and Parkinson, A., R: A Survey of Research in the application of Tolerance Analysis to the Design of Mechanical Assemblies. Research in Engineering Design, 3, 1991, p. 23-37. [4] Söderberg, R.: Robust Design by Support of Cat Tools. ASME Design Engineering Technical Conferences, Atlanta, Georgia USA, 1998. [5] Taguchi, G.: Introduction to quality engineering: designing quality into products and processes, Productivity Inc., 1986. [6] Suh, N., P.: The Principles of Design, Oxford, Oxford University Press, Inc., 1990. [7] Hu, S. J.: Stream-of-Variation Theory for Automotive Body Assembly. CIRP, 1997. [8] Wickman, C. and Söderberg, R.: Towards Non-Nominal Virtual Geometric Verification by Combining VR and CAT Technologies. 7th CIRP Seminar on Computer-Aided Tolerancing, Cachan, France, 2001. [9] Maxfield, J., Dew, P. M., Zhao, J., Juster, N. P., Taylor, S., Fitchie, M. and Ion, W. J.: Predicting Product Cosmetic Quality in the Automobile industry. 33rd international Symposium an Automotive Technology and Automotation (ISATA), Dublin, Ireland, 2000. [10] Maxfield, J., Dew, P. M., Zhao, J., Juster, N. P., Taylor, S., Fitchie, M., Ion, W. J. and M, T.: Predicting Product Cosmetic Quality using Virtual Environment. ASME Design Engineering Technical Conferences and Computers and Information in Engineering Conference, Baltimore Maryland, 2000. [11] Trost, J.: Enkätboken, Lund, Sweden, Studentlitteratur, 2001. [12] Kvale, S.: InterWiews. An introduction to Qualitative research interviewing, London, Sage Publications, 1996. [13] Merriam, S. B.: Case Study Research in Education: A Qualitative Approach, San Francisco, CA, Jossey-Bass Publishers, 1988. [14] Persson, S. and Wickman, C.: Effects of Industrial Design and Engineering Design Interplay: an Empirical Study on Tolerance Management in the Automotive Industry. Design Conference - Design 2004, Dubrovnik- Croatia, 2004.
B-8
Paper C Dagman, A., Söderberg R. and Lindkvist L., (2004), Split-Line Design for Given Geometry and Location Systems, submitted to the Journal of Engineering Design.
PAPER C
Split-line Design for Given Geometry and Location Schemes Andreas Dagman, Rikard Söderberg and Lars Lindkvist Department of Product and Production Development Chalmers University of Technology SE-412 96 Göteborg, Sweden tel. +46 (0)31 772 1472 Fax. +46 (0)31 772 1375. Email:
[email protected],
Keywords: Product Design, Industrial design, Design Process, Quality. ABSTRACT The spatial relations between parts in an assembly can be critical for the functional and aesthetic quality of a product. In the case of the automobile, these relations can be between doors, fenders, hood, panels etc. Variation in these relations, caused by part and assembly variation, influences the output variation which is what the customer sees and judges. This paper presents a computer-aided tolerancing tool that supports and improves splitline design with respect to geometrical variation. A split-line is the relation between two mating parts over a distance. The design and placement of a split-line in an automobile body are influenced by several aspects such as design language, geometrical dimensioning, crash safety etc. In this paper only the geometrical dimensioning aspects have been considered. The research has been carried out by using simulations and analyses in a computer-aided tolerancing software. The tool presented describes a way to calculate and visualize the geometrically most robust area and split-line between two parts. The findings from the research show that it is difficult to calculate and visualize the result in flush and gap direction in the same way. The tool gives insight into how the configuration of the locating schemes influences the geometrical robustness of the design.
1. Introduction A geometrically robust design is a design that fulfils its functional requirements and meets its constraints even when the geometry is afflicted with small manufacturing or operational variations (Söderberg and Lindkvist 1999). The exterior design of an assembled automobile body consists of a number of components, which are in spatial relation to each other. According to the automotive industry, the actual distance between geometries should be small, equal and parallel for all relations to achieve high geometrical quality (Wickman and Söderberg 2003). The relation between two mating parts over a distance is defined as a split-line; see Figure 1.
C-1
PAPER C
Figure 1. Examples of split-lines.
To achieve the intended dimensions and relations in split-lines, the use of tolerances is necessary. The area of tolerancing and tolerance management is multifaceted and includes several different research disciplines. A compilation of the research disciplines in tolerance management has been presented by Hong and Chang (2002). One of the areas in tolerance management is tolerance analysis, which aims at predicting the variation in one or several critical dimensions of a design with respect to expected manufacturing variation (Söderberg and Lindkvist 2002). There are many critical dimensions in an assembled product, among them being the dimensions of split-lines. The two most commonly used measuring directions of a split-line are gap and flush; see Figure 2. Flush is defined as the distance in the direction normal to a common plane between two parts. Gap is defined as the distance perpendicular to the normal of a common plane between two parts. Gap Flush
Figure 2. The measuring directions of flush and gap.
The variation in split-lines is the result of variation in the parts included in the assembly and of variation derived from the assembly process. When assembling products, it is of great importance to know the exact location of the individual parts and to be able to lock the parts in space, to prevent them from moving in an undesired way. To control the parts and the relations between parts in the assembly, locating schemes are used. A part has six degrees of freedom – three translations and three rotations – and these must be locked by locators in the locating schemes in order to control the part. Theoretically, the locators are points used to lock all degrees of freedom for a part in order to position it. Physically, the locator points are solved by physical geometrical features such as holes, planes and slots. In order to support the work with geometrical dimensioning and tolerancing, Computer Aided Tolerancing (CAT) tools have been developed. A evaluation of commercial CATsystems has been presented by Prisco and Giorleo (2002) and Kumar and Raman (1992). There are a number of different variants such as VSA-GTD, CETOL, 3DCS and RD&T. Their purpose is to support the three-dimensional tolerance analyses and the geometrical robustness analysis. To be able to fulfil the dimensional requirements of a solution, a common approach has been to find the optimal locator positions that minimize variation and make the assembly less affected by variation of the included parts (Söderberg and Lindkvist 1999) and (Camelio, et al. 2002). This way of optimizing assumes a freedom to place locators wherever the optimal geometrical location is. However, in production environments today, the assembly strategy – i.e. how parts are located and assembled with fixtures and/or
C-2
PAPER C
robots – is often more or less limited to a number of principal layouts. These conditions have been used as a prerequisite for this paper. Geometrical variation in the split-lines also has an effect on the aesthetic appearance of the assembly. A study presented by Dagman et al. (2004) showed that high Visual Quality Appearance (VQA) of split-lines is one of the main factors that influence the overall VQA of an automobile. The overall perceived VQA of the exterior of an automobile body is defined as the quality impression that a product conveys to a customer just by visually observing it (Dagman, et al. 2004). 1.1.
Scope of the paper
This paper presents a CAT-functionality that aims at dividing an initial product geometry, with given locating schemes, into parts in the most geometrically robust way. The geometries used in the work have been shell models – in this paper, the exterior parts describing the body of the automobile. The geometrical sensitivity of the split-lines has been in focus, and none of the other aspects that affect the split-lines such as crash safety, vibrations, air resistance etc. have been taken into consideration. Several different aspects of split-line design have been investigated in early research. The importance of locator schemes has been treated in Söderberg and Lindkvist (1999), and the quality appearance of the split-lines in Söderberg and Lindkvist (2002). The judgement in virtual reality has been discussed in Wickman and Söderberg (2001), and the importance of high split-line quality for industry and customers is investigated and confirmed by Dagman, et al (2004). This paper presents research based on the abovementioned aspects, but here the locating schemes are fixed and the design and location of the split-line are studied. The research focuses on the early phases of the product development process where no real tolerances are known. Therefore the analysis and simulation of the geometries are carried through with unit tolerances applied to the locating schemes, to examine the general robustness of the concept. The results are implemented and tested in the CAT software RD&T. The intended functionality will enable both synthesis and analysis which allows the user to make his/her own decisions on the basis of the results from the simulation.
2. Tolerance management The area of tolerance management tries to decrease and solve the problem of variation. Tolerance management is divided into several sub-areas treating specific aspects of the variation issue, tolerance analysis being one of the sub-areas. The methods of tolerance analysis can be either deterministic or statistical; the design models to be analyzed may be 1D, 2D or 3D. Tolerance analysis is an important link between design and production, and it has been the topic of a large number of research publications during the years. 2.1.
Geometrical robustness
Variation in critical product dimensions has its roots in three areas, namely component variation, assembly variation and the robustness of the concept; see Figure 3.
C-3
PAPER C
Component variation Machine precision
Process variation
Manufacturing process
Assembly variation Assembly precision
Process variation
Assembly process PKC variation Robustness
Design concept
Figure 3. Geometrical PKC variation contributors (Söderberg 1998).
These characteristics can be seen as Product Key Characteristics (PKC) (Lee and Thornton 1996). The PKC are a set of product features that are highly constrained or for which minute deviations from nominal specifications, regardless of manufacturing capability, have a significant impact on the product’s performance, function, and form at each product assembly level. PKCs are permanent for a given product design decomposition and set of requirements (Lee and Thornton 1996). The usual procedure is that the suppliers control the component variation and the assembly variation is taken care of inhouse by the assembly process. The final variation is dependent on the geometrical robustness of the concept. In a sensitive design, component and assembly variation is amplified, whereas in a robust design the variation is suppressed. Since the robustness of a concept is controlled by the placement of the locators, it is important to achieve geometrical robustness already in the concept phase. The relation between robustness and variation can be explained with a beam and a support; see Figure 4. Output
Input
Sensitive
Robust
Figure 4. Robustness and variation explained by a beam and support.
Depending on the placement of the support, the input variation will either increase or decrease the output variation. If the support is moved to the left, the input variation will lead to an amplified output variation. If the support is moved to the right, the output variation will not be as affected as in the first example and the robustness will increase. This indicates that input and output variations depend on each other and on the placement of the support. Design robustness is thus controlled by the relation between input and output. This relationship can be expressed as follows: robustness variation
= x 0 x x
locators tolerance
Since the position of the locator controls two important product characteristics, this should be treated first. Based on final requirements for the output variation and known sensitivity
C-4
PAPER C
(relations between input and output), the tolerance for the input variation may then be determined. 2.2.
Locating schemes
All parts in an assembly are positioned with a locating scheme. A frequently used scheme in the automotive industry is the 3-2-1 scheme, where six theoretical points are used to lock six degrees of freedom for a part, three translations and three rotations; see Figure 5. Three points, A1, A2 and A3, form a plane which locks two rotations and one translation. Two points, B1 and B2, form a line that locks one rotation and one translation. The final point, C1, locks the remaining translation. Z
B2 C1 Y
X
B1 A1
A3 A2 Target 3-2-1
B2 B1 A1
C1 A3
A2 Object 3-2-1
Figure 5. The 3-2-1 locating scheme.
The 3-point positioning scheme is a variant of the 3-2-1 positioning scheme where three points, instead of six, are used to lock the six degrees of freedom for the part. In conformity with the 3-2-1 scheme, the first point in the 3-point scheme locks two rotations and one translation. The second point locks one rotation and one translation, and the third point of the 3-point scheme locks the final translation. The locating points and their possibilities to move are illustrated with arrows in Figure 6. L2: A2, B2
L3: A3
L1: A1, B1, C1
Figure 6. The 3-point locating scheme with locator L1-L3.
2.3.
Quality appearance
The possibility to make the split-lines tighter and more accurate gives the impression of good craftsmanship. The geometrical robustness in relation to product quality has been discussed by Taguchi et al. (1989) and Phadke (1989). To be able to evaluate an existing split-line, the Quality Appearance (QA) index has been presented by Söderberg and Lindkvist (2002). The QA index measures the overall quality appearance of a product’s split-lines by using statistical simulations and pre-defined “rating lists”. The QA index is calculated for seams, the spatial relations of two parts over a specified distance in the gap and flush directions. The relations can also be considered from an aesthetic point of view. The importance of working with the visual product form is discussed by Warell (2002).
C-5
PAPER C
Dagman et al. (2004) investigated customers’ and industry’s attitude regarding split-lines. They found that relations between different parts, i.e. the split-lines, were one of the main factors that affected the VQA. There is a wide range of parameters that affect the VQA of a split-line. These could be quantitative parameters such as the distance of the gap or the flush, or non-parallelism etc. There might also be qualitative parameters like the curvature of the split-line, what is visible beneath the split-line, the colour of surfaces, the design, etc. For companies which are making products composed of several components, it is costly and time-consuming to consider and control the impact of geometric variation. The achievement within the automotive industry of making products with a high VQA of split-lines is something that customers regard as important for the overall VQA of the exterior of an automobile body (Dagman, et al. 2004). The possibility at an early stage of the product development process to virtually verify a concept, so as to get a predictive design analysis, has become an important factor to cut the expenses in the product development phase. A number of ways to use virtual environments to verify and predict quality appearance has been suggested (Maxfield, et al. 2000), (Maxfield, et al. 2000), and (Wickman and Söderberg 2003). In Part 3 the analysis setup is presented and explained. In Part 4 the results are presented, and in Part 5 the conclusions and contributions are presented.
3. Split-line design In the early phases of an automobile product development process, the computer models are mainly shell models and not solid. In these phases the designers have laid down the general design for the automobile and it is possible to change the design without having a huge increase in cost. The shape of the doors, fenders, hood etc. is a result of several different requirements like functionality, design language, geometrical dimensioning, crash safety etc. that affect the shape. In this paper, the geometrical dimensioning requirements are treated deeply and the area of aesthetics and design language are treated shallowly. The other requirement areas have not been treated at all. Different design concepts vary in their geometrical robustness for assembly, depending on the location schemes and the design of the parts. As noted above, the assembly strategy – i.e. how parts are located and assembled with fixtures and/or robots – is often more or less limited to a number of principal layouts. The prerequisites for the split-line design in this paper have been fixed locating schemes, evaluation in gap and flush directions, and shell models. The geometries used in the evaluation phase were non-nominal and the parts were considered as rigid. The simulation software RD&T, which was used as a workbench in this research, allows the user to define different locating schemes. In this case, the 3-2-1 scheme and the 3-point scheme have been used to position the parts. The general robustness of the design is simulated and analyzed by using unit tolerances, with the same tolerance range in each locating point. The simulation and analysis process will be described in the following. In the explanatory example, the split-line is to be placed and designed in a geometrically robust way between the front door and the front fender; see Figure 7.
C-6
PAPER C
Figure 7. The area of interest is marked with a circle on a shell model of an automobile.
3.1.
The setup for simulation and analysis in the flush direction
The simulation and analysis have been carried out as follows. 1. The first step was to create a computerized shell geometry including both the geometries to be divided. This model was converted to VRML. In this case the geometries are front fender and front door. The final design of the fender and door is not fixed in this stage of the product development process, which allows a design change. The geometry was then copied; see Figure 8. The 3-point location schemes were used to lock the two geometries.
Figure 8. Two identical geometries, front door and front fender, to be analyzed.
One of the geometries got the locating scheme configuration for the front fender, and the other got the location for the front door; see Figure 8. The locating point is shown by hourglass shapes with arrows. In Figure 9, the analysis principle is shown with simple geometries. The placement of the locating schemes is shown by arrows pointing in the influencing variation’s direction. Only four of the six locating points are shown in Figure 9, the other two being hidden.
C-7
PAPER C
Figure 9. The principle of analysis where the input and output variation is shown, exaggerated.
2. Unit tolerances were next applied to the locating schemes. The first geometry was then placed on top of the second geometry. This gives two identical geometries superimposed but with different locating schemes; see Figure 10. To obtain properly constrained geometries, the split-line must divide the initial geometry in such a way that the final result is two geometries with three locating points each. This is a constraint when working with placement and design of split-lines.
Figure 10. The position of the locating scheme and the two geometries on top of each other.
3. A large number of measuring points were then generated on both parts; see Figure 11. The measuring points were created in the vertices of the VRML triangles describing the geometry. Since it is the same geometry used twice, there will be pairs of measuring points at the same locations. Measures in the direction normal to the surface of the geometries were then created between the pairs of measuring points. The measures used in this paper were point-to-point measures, which means a measure from one point on a part to another point on the other part.
C-8
PAPER C
Figure 11. Measuring points on the geometries.
4. A stability analysis was performed to examine the design. The stability analysis evaluates the general robustness of a design by using a unit disturbance. The result from the analysis is presented in Figure 12, using the colour-coding feature in the CAT-software. The colour-coding feature applies different colours to the analysis results in the measuring points according to the level of variation. Since the colour is applied to the vertices of the VRML triangles, an interpolation is performed between them in order to apply the right colour to the entire triangle.
Figure 12. The analysis result at the shell geometry in the flush direction.
Measures with low variation are coloured in blue (dark grey) and measures with high variation in red (light grey). In Figure 12 the analysis result shows a delimited blue area, on the automobile geometry, that is less affected by variation. If the main part of the splitline goes through the robust area, it will not be as affected by input variation as it would have been if placed differently. Flush measures between the two outputs are controlled by the position of the locators and by the position of the split-line. Since the locators are
C-9
PAPER C
fixed, variation is controlled by the position of the split-line alone. This way of simulating and analyzing is suitable only in the flush direction. 4.1.
The setup for simulation and analysis in the gap direction
Since measurement in the gap direction is performed perpendicular to the split-line direction, the simulation and analysis setup had to be changed. The direction of the splitline must be fixed before the measuring can take place. Steps 1 and 2 of the analysis procedure in 3.1 were identical for the gap analysis too. In step 3 the measuring points were applied along fictitious split-lines to the geometries, instead of applying them in the vertices of the VRML triangles; see Figure 13.
Figure 13. Fictitious split-lines on the geometries.
By adding a large number of these fictitious split-lines on the geometries, and by using the colour-coding feature, gap could be analyzed in a manner similar to flush. 4.2.
Split-lines design in the most robust way in the gap direction
An algorithm to find the most geometrically robust split-line in the gap direction was developed, and works as follows. A starting point is defined by the user in the lower or upper part, in the z-direction, of the geometry to be analyzed; see Figure 14. The starting areas for this example are also shown in Figure 14. The split-line will start in this point. An incremental search for the most geometrically robust segment of the final split-line is performed, as illustrated in Figure 14. Fictitious segments of the split-line are created from the starting point, and measurement in the gap direction is performed. The incremental scan is performed in such a way that it ensures movement only in the positive or negative z-direction, depending on the placement of the starting point. Variation in each of the fictitious split-line segments is calculated, and the one with lowest variation in the gap direction is chosen.
C-10
PAPER C
Figure 14. Principles of split-line design in the gap direction, simplified.
The starting point is moved to the end of the chosen fictitious split-line segment with lowest variation, and a new incremental search is performed with the new starting point; see Figure 15. This is iterated until the geometries are completely divided and a split-line is created.
Z X
Figure 15. The creation of the split-line.
5. Results Three main results will be presented in this paper: I. A CAT-functionality to simulate, analyze and visualize where the most robust area in the flush direction, between two parts with given location schemes, is located and shaped. This is presented in sub-section 4.1. II. A CAT-functionality to simulate, analyze and visualize the geometrically most robust split-line in the gap direction, between two geometries with given location schemes, is presented in sub-section 4.2.
C-11
PAPER C
III. Insight into how different configurations of the locating schemes affect the location and shape of the robust area, and the most robust split-line, is presented in sub-section 4.3. I. The most robust area in the flush direction As mentioned, the geometries used were the front fender and the front door, with individual locating schemes. When performing the stability analysis, and applying the colour coding, the results shown in Figure 16 were obtained. The colour coding gives the designer a description of where the least sensitive area between the two parts is located and what its shape is. If it is not possible to move the location scheme, the main part of the split-line should go through the blue-coloured (dark) area to achieve a robust solution. The configuration of the locating schemes influences the results, as can be seen in Figure 16. Four different locating-scheme configurations have been used to show their influence on the appearance of the most robust area in the flush direction.
Figure 16. Colour coding of flush results with four different locating schemes.
II. Geometrically most robust split-line in the gap direction In Figure 17 the optimal split-line, illustrated as a black line, divides the automobile geometry. The starting point of the split-line was in the bottom part of the geometry. Depending on the placement of the starting point and the configuration of the locating points, the shape of the split-line will be different.
C-12
PAPER C
Figure 17. Split-line optimized in the measuring direction of gap.
The algorithm describing the optimal split-line does not take into account the number of locator points on each geometry when dividing the geometry. The starting point has to be in a location which ensures that there are three locators on each side. This is a prerequisite for achieving three locators on each geometry after the division is carried out. When applying the results in the gap and flush directions at the same time on the geometry to be divided, the results were not always congruent. In Figure 18 both a unified and a non-unified solution are shown. In the left corner of both Figures 18a and 18b, the analysis result is shown by thin plates. The same configuration of the locating schemes has been used in the automobile geometries as in the plate geometries. The plates are used to elucidate the interpretation of the simulation results. By changing the configuration of the location schemes in Figure 18a in accordance with 18b, the gap and flush solutions will be congruent. Since the results show that there is not always congruence between the results in the gap and flush directions, there must be a compromise between them. The tool can also be seen as a way of verifying the optimal split-line in the gap direction against the robust area in the flush direction.
a
b
Figure 18. The results in the flush and gap directions applied to the same geometry.
C-13
PAPER C
The predicted split-line may either contribute to the design language of the automobile or not. In Figure 18, the shapes of the two split-lines are different and their contributions to the design language differ. These concerns have not been treated more thoroughly than to form a subjective opinion regarding the split-line solutions. For example, the split-line presented in Figure 18b can be interpreted as corresponding more to the design language than 18a. This view is based on the fact that the shape of the split-line has the same slope as the A-Pillar. But the design of the split-lines has not been focused upon, and it was not intended to follow the design language of the product. III. The influences on variation depending on locator placement During the analysis, the importance of the configuration of the locator points was accentuated. According to the locator scheme used, the 3-2-1 scheme, the first three locators A1-A3 have the main effect on variation in the flush direction. Locator B1-B2 influences variation in the gap direction. The configuration possibilities of the locators were restricted to four positions on each geometry. Thin plates were used as example geometries during the simulation and analysis performed. In Figure 19, the colour coding of the analysis results in the flush direction are shown with six different configurations of the A1-A3 locators. As can be seen, different patterns appeared after the simulation. The blue areas (dark grey) indicate low variation. By knowing how the configuration and location of the locators influence the robust area, it is possible to evaluate the design solution.
Figure 19. Colour coding of the results in the flush direction with six different configurations of the locating schemes.
The location and configuration of the locators also influenced the design of the most robust split-line in the gap direction. In Figure 20, the same configuration of locating scheme has been used but the positions of the B1-B2 locators are changed.
C-14
PAPER C
B2
B2
B1
B1
B2
B2
B1
B1
B1
B1
B2
B1
B1
B2
B2
B2
Figure 20. The configuration of the B1-B2 locators influences the design of the split-line.
It can be seen that the design of the split-line changes depending on where the locators are placed.
6. Conclusions and future work Several different aspects of split-line design have been investigated in early research. The importance of locator schemes has been treated in (Söderberg and Lindkvist 1999) and the quality appearance of the split-lines in (Söderberg and Lindkvist 2002). The judgement in virtual reality has been discussed by (Wickman and Söderberg 2001), and the importance of high split-line quality for industry and customers is investigated and confirmed by (Dagman, et al. 2004). In this paper we have proposed CAT-functionalities for simulating and analyzing geometries with given locating schemes in order to find the geometrically most robust design and placement of a split-line between two parts. The following results have been presented. 1. A CAT-functionality to simulate, analyze and visualize where the most robust area in the flush direction, between two parts with given location schemes, is located and shaped. 2. A CAT-functionality to simulate, analyze and visualize the geometrically most robust split-line in the gap direction, between two geometries with given location schemes. 3. Insight into how different configurations of the locating schemes affect the location and shape of the robust area and the most robust split-line. The CAT-functionalities in this paper are adapted to the automotive industry, and a geometry consisting of front door and front fender has been used to exemplify the functionalities. But the same procedure can be used on different products. In the automotive industry, the design and placement of split-lines influence the VQA of the vehicle and are therefore of importance. The results presented should give the designer information regarding how to design and place the split-line while preserving the design’s aesthetic freedom. By using both the geometrically robust area in the flush direction and
C-15
PAPER C
the most robust split-line in the gap direction, an opportunity arises to verify a design solution from a viewpoint of geometrical dimensioning and tolerancing. The presented functionalities are to support the geometry design process in the early stages of the product development process when the final design is not fixed. The functionalities can help to increase the VQA, but this is just one of several quality aspects in the automotive industry. The results also showed, in an easily interpreted way, how different configurations of the locating schemes affect the robustness of a design solution. New methods and tools for supporting the daily product development work are of high value. The next step in this research will be to implement the tool at an automobile company, so as to determine its functionality in an industrial project. A refinement of the findings regarding how different locating schemes affect the location and shape of the most robust area will be investigated. 6.1.
Acknowledgements
The authors would like to acknowledge the Swedish Foundation for Strategic Research through the research program ProViking for financial support, and the former research school ENDREA.
7. References Camelio, J. A., Hu, S. J. and Ceglarek, D. J., 2002, Impact of fixture design on sheet metal assembly variation. 7th Design for Manufacturing Conference (Montreal, Canada: American Society of Mechanical Engineers), pp.133-140. Dagman, A., Wickman, C. and Söderberg, R., 2004, A Study of Customers' and the Automotive Industry's Attitude Regarding Visual Quality Appearance of Split-Lines. 4th International Conference on Advanced Engineering Design AED 2004 (Glasgow: AED2004), p.CD. Hong, Y. S. and Chang, T.-C., 2002, A comprehensive review of tolerancing research. international journal of production research, 11, 2425-2459. Kumar, S. and Raman, S., 1992, Computer-aided tolerancing: the past, the present and the future. Journal of Design and Manufacturing, 29-41. Lee, D., J. and Thornton, A. C., 1996, The identification and use of key characteristics in the product development process. The 1996 ASME Design Engineering Technical Conferences and Computers in Engineering Conference (Irvine, California: American Society of Mechanical Engineers), Maxfield, J., Dew, P. M., Zhao, J., Juster, N. P., Taylor, S., Fitchie, M. and Ion, W. J., 2000, Predicting Product Cosmetic Quality in the Automobile industry. 33rd international Symposium an Automotive Technology and Automotation (ISATA) (Dublin, Ireland: ISATA), Maxfield, J., Dew, P. M., Zhao, J., Juster, N. P., Taylor, S., Fitchie, M., Ion, W. J. and M, T., 2000, Predicting Product Cosmetic Quality using Virtual Environment. ASME Design Engineering Technical Conferences and Computers and Information in Engineering Conference (Baltimore Maryland: American Society of Mechanical Engineers), p.19.
C-16
PAPER C
Phadke, M. S., 1989, Quality engineering using robust design (Englewood Cliffs, N.J.: Prentice Hall). Prisco, U. and Giorleo, G., 2002, Overview of current CAT systems. Integrated Computer-Aided Engineering, 4, 373-387. Söderberg, R., 1998, Robust Design by Support of Cat Tools. ASME Design Engineering Technical Conferences (Atlanta, Georgia USA: American Society of Mechanical Engineering), p. 260. Söderberg, R. and Lindkvist, L., 1999, Computer aided assembly robustness evaluation. Journal of Engineering Design, 2, 165-181. Söderberg, R. and Lindkvist, L., 2002, Stability and seam variation analysis for automotive body design. Journal of Engineering Design, 2, 173-187. Taguchi, G., Hsiang, T. C. and Elsayed, E. A., 1989, Quality engineering in production systems (New York: McGraw-Hill). Warell, A., 2002, Design Syntactics: A functional Approach to Visual Product Form, Chalmers University of Technology, Göteborg. Wickman, C. and Söderberg, R., 2001, Defining quality appearance index weights by combining VR and CAT technologies. 2001 ASME Design Engineering Technical Conference and Computers and Information in Engineering Conference, Sep 9-12 2001 (Pittsburgh, PA, United States: American Society of Mechanical Engineers), pp.12151224. Wickman, C. and Söderberg, R., 2003, Increased concurrency between industrial and engineering design using CAT technology combined with virtual reality. Concurrent Engineering Research and Applications, 1, 7-16. Wickman, C. G. and Söderberg, R., 2003, Comparison of non-nominal geometry models represented in physical versus virtual environments. 2003 ASME International Mechanical Engineering Congress, Nov 15-21 2003 (Washington, DC, United States: American Society of Mechanical Engineers), pp.29-37.
C-17
