Introduction
To satisfy customers, product performance should be consistently on target. This requirement for competitiveness is confirmed by anti-variation initiatives launched during the past two decades, many of them under the heading of Six Sigma. While there was a focus on the manufacturing stage at the beginning of this trend, the emphasis of these anti-variation efforts has now moved upstream resulting in recent initiatives such as Design for Six Sigma (Creveling et al., 2003), Variation Risk Management (Thornton, 2004) or Axiomatic Quality (El-Haik, 2005).
Chowdhury (2003) remarks that Design for Six Sigma (DFSS) has evolved from a major weakness of the Six Sigma initiative – the development of entirely new products, processes or services, not just the improvement of existing ones. Further, very often there is a lack of any existing process to measure, analyze and then incrementally improve through a Six Sigma approach, or incremental improvement to a process will not bring the necessary benefits. In such cases a complete redesign is needed, which, according to Tennant (2002), can be achieved with the help of DFSS. Berryman (2002, p.24) notes that DFSS is “not well-documented or understood” and there are different types of definitions available in literature: De Feo (2002, p. 62), for example, defines DFSS as “an established, data-driven methodology based on analytical tools that provide users with the ability to prevent and predict defects in the design of a product, service or process.” Mader (2002, p. 82) describes it more generally as “an enhancement to an existing new product development process that provides more structure and a better way to manage the deliverables, resources and trade-offs.”
Tennant (2002, p.12) emphasizes the customer by defining it as “a rigorous approach to the design of a new product or service to reduce delivery time and development cost and to increase the effectiveness of the product or service and hence customer satisfaction.” However, all different views do agree in one aspect: DFSS constitutes an improvement of the current product design process (PDP). Robust Design Methodology (RDM) including efforts towards insensitivity of product performance to sources of variation is commonly advocated as an important tool within DFSS (Creveling et al., 2003; Yang & El-Haik, 2003). Sources of variation are also referred to as noise factors and to simplify matters, in the following ‘product’ refers to ‘product or process’.
In the automotive industry, complete new product development projects usually takes some years; considerably longer than the 3 – 6 months common for Six Sigma projects. Further, for Six Sigma projects there is a well-established Define-Measure-Analyze-Improve-Control (DMAIC) procedure whereas there is no such established procedure for DFSS. We have e.g. Define-Customer-Concept-Design-Implement (Tennant, 2002), Define-Characterize-Optimize-Verify (Soderborg, 2004), Concept-Design-Optimize-Verify (Creveling et al., 2003) or Identify-Characterize-Optimize-Verify (Yang & El-Haik, 2003). Gremyr (2006) even finds “fuzziness” about the concept of DFSS. All these aspects entail difficulties for companies trying to implement DFSS. Often companies simply do not know where and how to start and how to continue their efforts. This is a pity given that the PDP in general is well-documented and understood and this paper will make a contribution to this issue.
The purpose of this paper is to clarify the actual connection between DFSS and the different stages of a PDP. As the different design phases all have different deliverables, different tools are required, respectively. We outline available tools and methods associated with DFSS and give recommendations for where these tools can be utilized in a PDP to increase product quality and customer satisfaction.
The remainder of the paper is organized as follows: the following section contains a model of a general PDP that will be used as a reference for the subsequent discussions. The third section presents different DFSS tools and discusses where and how these tools can be utilized in a DFSS context. The paper ends with conclusions.
The Product Design Process
This section describes a general scheme of a PDP that will be used as a reference in the discussion of DFSS tools and their application. Figure 1 shows four basic phases related to a generic PDP (Pahl & Beitz, 1996): Planning and Clarifying the Task, Concept Design, Embodiment Design and Detail Design. The Concept Design phase can be divided into Concept Generation and Concept Screening & Improvement, as will be further explained below.
The major deliverables connected to these four phases can be found in e.g. Roozenburg and Eekels (1995), Ulrich and Eppinger (1995), Pahl and Beitz (1996) and Ullman (1997). Separating the phases by gates allows for check-ups of the ongoing PDP.
The Planning and Clarifying the Task phase deals with gathering and interpreting information about the market situation. Ullman (1997, p. 60) characterizes this by stating that “the most important part in understanding the design problem lies in assessing the market, i.e., establishing what the customer wants in the product.”
According to Pahl and Beitz (1996), Planning and Clarifying the Task should lead to a product idea that is needed and that looks promising given the current market situation, company needs and economic outlook. Such an idea must be at hand before a product development project can be initiated. Ulrich and Eppinger (1995) further state that this proposal must be technically feasible and that customer needs should be identified and taken into account. Once this is achieved, Roozenburg and Eekels (1995) suggest the determination of design specifications defining the required functions and properties of the new product.
Ullman (1997, p. 120) defines a concept as “an idea that is sufficiently developed to evaluate the physical principles that govern its behavior.” Cross (1994) refers to the generation of design solutions as an essential and central aspect of design. The Concept Design phase resulting in a description of the form, function and features of a product (Ulrich & Eppinger, 1995), essentially breaks down into two consecutive components (Pugh, 1991):
- Concept Generation, and
- Concept Screening.
According to Thornton (2004) the Concept Generation phase results in rough design layouts, e.g. drawings and simple prototypes with key technical choices. It is important that many concepts with different solutions are generated. All concepts are evaluated in a screening process. Ullman (1997) states that concepts not fulfilling customer requirements are screened out while the remaining concepts are further developed. Techniques for generating and evaluating concepts are used iteratively until a winning solution proceeds for Embodiment.
The Embodiment Design defines the arrangement of assemblies, components and parts, as well as their geometrical shape, dimensions and materials. The outcome is the specification of layout (Roozenburg & Eekels, 1995; Pahl & Beitz, 1996). Pahl and Beitz (1996) find that several embodiment designs are often needed before a definite design can emerge. Roozenburg and Eekels (1995) write that this implies that Embodiment involves corrective cycles in which analysis, synthesis, simulation and evaluation constantly alternate and complement each other. The preliminary design developed in this phase should be continuously improved in the subsequent phase.
Pugh (1991) characterizes Detail Design as being concerned with the design of the sub-systems and components that together make up the whole design. It can be seen as an optimization phase in which the design should be close to finalized. This entails that the arrangements, forms, dimensions and surface properties of all the individual parts are finally determined, the materials are specified, production possibilities assessed, costs estimated and all the drawings and other production documents produced. In summary, according to Pahl and Beitz (1996), the outcome of the Detail Design phase is a specification of production. Ulrich and Eppinger (1995) remark that design of tooling and provision of assembly instructions are also deliverables in line with this phase.
Design for Six Sigma in the Product Design Process
The discussion of DFSS contributions is structured according to the previously described phases of the general PDP. Different tools addressing the respective phases will be outlined and their appropriateness clarified against the background of DFSS.
Planning and Clarifying the Task
From a DFSS perspective it can be particularly interesting to investigate warranty claims as they provide an insight into concerns and needs of current customers and they can further be used as input for working with Quality Function Deployment (QFD) (Kogure & Akao, 1983). Creveling et al. (2003) argue that warranty issues are not handled satisfactorily since the cause of failure is not determined or, in many cases, what is reported is symptom rather than root cause.
Companies should make sure that warranty claims are fully utilized and Fundin and Cronemyr (2003) present one method as an example in this respect. It is not likely that all customers are equally suited to provide qualified feedback. Consequently, Creveling et al. (2003) suggest the establishment of partnerships with a few key customers and repair depots and to train them to provide useful field failure information.
It is also desirable that companies do not ignore the time after the warranty period has expired. The long-term reliability of products strongly affects products’ second-hand value, which in turn affects how much customers are willing to pay in the first place. Nevertheless, Creveling et al. (2003) hold that most companies have poor routines concerning the gathering of information from beyond the warranty period. The solution can be of the same type as suggested above; establish partnerships with key customers and repair shops and train them to provide useful information.
Target values are particularly important from a DFSS perspective. In this early phase of design, QFD can facilitate the identification of preliminary target values taking the voice of the customer into account. Further, the work of gathering customer needs is hardly worthwhile unless they are properly translated into product characteristics (PCs). Pugh (1991) considers deriving PCs as a crucial part of the PDP since they constitute the objectives and boundaries of all the subsequent design phases. Sullivan (1986) further states that QFD is a suitable system for assuring that customer needs drive the product design and production process. Xie et al. (2003) provide advanced QFD applications and furthermore show how the Kano model (Kano et al., 1984) categorizing customer needs into attractive, expected and must-be quality can be integrated into QFD applications.
Concept Design – Concept Generation
The Concept Design phase requires creativity, experience and skill of the designer (Phadke, 1989; Suh, 1990), and merely applying appropriate tools and methods does not guarantee e.g. more reliable products. Designers should be encouraged to take the customer’s perspective and to think in terms of robustness. This is particularly true in the Concept Generation phase where a designer – enlightened from a DFSS point of view – can possibly generate concepts resulting in increased customer satisfaction and robustness intuitively. Supporting this, Pugh (1991, p. 71) notes that “concepts are often best generated by individuals”, whereas the “concept selection and enhancement is best performed in groups.” However, to initiate and organize individual thinking processes, brainstorming sessions have proven to be a useful approach (Priest & Sánchez, 2001; Chowdhury, 2002).
The Theory of Inventive Problem Solving (TIPS) by Altshuller serves with another important source of inspiration for generating but also improving already existing concepts. For example, when there are two coupled functional requirements in any existing design, Yang and El-Haik (2003) suggest using TIPS for finding design parameters resulting in a decoupled design.
Pahl and Beitz (1996) see another important source of ideas for generating new or improved solutions in the analysis of existing technical systems including those of competitors. In a DFSS context benchmarking seems to be a natural move if it turns out that e.g. competitors’ products are superior somehow. Customer satisfaction polls indicate customers’ perceptions of different product characteristics and can therefore provide relevant information.
To more accurately determine and fathom e.g. competitive products’ robustness, however, they must be subjected to testing. In this context, Bandurek et al. (1990) argue that design of experiments (DoE) can be applied to set up an efficient experimental plan to identify the combination of environmental factors that results in the maximum stress level for each product. The result can be compared with the anticipated environment that the product under development is likely to be exposed to. If any of the competitive products is found to be insensitive to some interesting combinations of noise factors, an understanding should be sought for why this is so. Aiding this understanding, Taguchi and Clausing (1990) point out that the basic principles for designing for robustness are often indistinguishable from the principles of designing for manufacture – reduce the number of parts, consolidate sub-systems and integrate the electronics.
The interplay of different parameters affecting a product and its performance can be illustrated with the help of a P-diagram (e.g. Phadke, 1989), see Figure 2. It describes a product as an input-output model and illustrates the different categories of parameters affecting it. The system response (y) can be described as a function f, which is determined by some input signal (M), design/control factors (X) and unwanted noise factors (Z): f (M, X, Z). In the ideal situation a given input signal constantly generates identical outputs. However, in reality, noise factors will distort the system response and cause variation in performance. The visualization of this context helps engineers to structure the design problem and to understand the product as a system with its influential parameters.
Concept Design – Concept Screening
The concept generation phase should result in many concepts with different solutions – “single solutions are usually a disaster” (Pugh, 1991, p. 69). All solutions are probably not equally well suited, which calls for a screening procedure. It is important to view the selection of concepts as a screening process where concepts are successively improved (Ullman, 1997).
Brainstorming sessions for Concept Screening should focus on the identification of potential noise factors. Since noise factors can originate from various stages of the product life cycle the brainstorming team should ideally be a cross-functional mix of representatives from product development, manufacturing, assembly, suppliers, retailers, customers etc. Ishikawa diagrams can facilitate the process of structuring and documenting the identification of potential noise factors as illustrated in Figure 3, where PC refers to product characteristic and NF refers to noise factor.
As commonly agreed, there are three different stages in a product’s life cycle where variation in product performance can arise (Taguchi 1986; Phadke, 1989; Park, 1996): product development, production, and the customer environment. This should be taken into account when potential noise factors are identified.
Pugh (1991) proposes a structured method for concept selection based on an evaluation matrix. The concepts are visualized with e.g. sketches and evaluated against a list of criteria that is based on the PCs. If a number of strong concepts emerges from this first procedure (Phase I), Pugh (1991) suggests further developing them to a higher level and going into a second evaluation phase. Phase II is similar to Phase I in its procedure, but a revised list of evaluation criteria is used as the designs are more detailed. Consequently, the concept selection technique provided by Pugh supports the idea of developing several concepts concurrently. From a DFSS perspective it seems natural to incorporate questions related to customer satisfaction and robustness into these criteria.
However, the scarceness of detailed and quantitative information at this early stage of the design process calls for qualitative approaches for concept evaluation. Failure Mode and Effect Analysis (FMEA), for example, is in fact a widespread method for preventive failure avoidance. Nevertheless, FMEA is a failure-oriented approach and thus might not be ideal for evaluating different concepts’ sensitivity to noise factors.
In Johansson et al. (2006) Variation Mode and Effect Analysis (VMEA) is presented as a variation-oriented tool to systematically look for noise factors affecting key product characteristics (KPCs). According to Thornton (2004, p. 35) “a key [product] characteristic is a quantifiable feature of a product or its assemblies, parts, or processes whose expected variation from targets has an unacceptable impact on the cost, performance, or safety of the product.” Each KPC can be further divided into a number of sub-KPCs, which are defined as the key varying elements that significantly contribute to the variability of the KPCs. A VMEA results in Variation Risk Priority Numbers (VRPNs) that give an estimate of the portion of variation contributed by each sub-KPC to the KPC. Furthermore, identified noise factors are also attributed a VRPN ranking them in terms of severity with respect to variation. VMEA can be used to evaluate and compare different concepts in their early stages with respect to robustness.
Andersson (1997) suggests the evaluation of concepts against design principles in order to explore their robustness in the Concept Design phase, while Pahl and Beitz (1996) see the benefits of design principles in the Embodiment Design phase. There is, however, no contradiction in utilizing design principles in both the Concept and the Embodiment Design phases. In this paper we present a selection of design principles in the Embodiment Design phase (see next section), but at the same time we encourage designers to evaluate concepts’ robustness against these principles as early as possible – already in the Concept Design phase if feasible.
Another approach related to the one presented by Andersson (1997) is to evaluate and develop concepts according to the principles of axiomatic design. Axiomatic design consists of two axioms and a number of corollaries (Suh, 1990). According to the first axiom, the independency axiom, design factors and PCs are related such that specific design factors can be adjusted to satisfy their corresponding PCs without affecting other PCs (Suh, 1990). Following the independency axiom does not necessarily lead to less failures. However, it lays the foundation for a successful setting of design factors. Naturally, it seems more straightforward to optimize the settings of design factors if they do not affect several product characteristics simultaneously.
Embodiment Design
In Embodiment Design the design is more detailed, which facilitates more accurate analyses of design alternatives. Regardless of whether the evaluation process is based on mathematically derived transfer functions, computer simulations or physical prototypes – from a DFSS perspective, the essence should be to understand the interactions between noise factors and corresponding design factors. For this reason noise factors need to be identified and KPCs must be defined in order not to loose the customer focus.
This identification process can – as described in the previous design phase – be supported by VMEA, which indicates where further investigations are most appropriate. VMEA is, however, merely a problem identifier and not a problem solver. To mitigate KPCs’ sensitivity to noise, consideration of certain design principles can be of guidance (e.g. Suh, 1990; French, 1994; Pahl & Beitz, 1996; Andersson, 1997). Pahl and Beitz (1996) emphasize that problems are introduced and breakdowns or accidents may occur if design principles are ignored. They further hold that all principles and guidelines can be condensed to the fundamentals of clarity, simplicity and safety.
Clarity entails unambiguous relationships between sub-functions, and appropriate inputs and outputs must be guaranteed. In Andersson (1997) simplicity of a solution implies fewer possible ways for noise to enter the system. Hence, a less complex solution has inherent robust qualities that facilitate the identification of potential noise factors in advance. The safety principle addresses issues related to strength, reliability, accident prevention and protection of the environment (Andersson, 1997). In Pahl and Beitz (1996) the safety principle involves clear specifications of operating conditions and environmental factors and analysis of components or systems to determine their durability when they are overloaded or subjected to adverse environmental impacts. All these objectives contribute to an approach to a less fault-prone PDP.
As the design is further developed it is possible to use prototypes to evaluate product performance. Bergman and Klefsjö (2003) note that well planned experiments can provide rapid knowledge of the values that have to be chosen for design and process parameters. By applying DoE to prototypes, interactions between design factors and noise factors can be investigated and the robustness of different design solutions can be evaluated. However, prototypes are expensive, which implies that other evaluation methods should be utilized as much as possible, such as computer simulation, unless the cost of alternative methods exceeds the cost of designed experiments.
Detail Design
Parameter design can be used to increase the product’s insensitivity to noise factors by identifying optimal settings of design factors (Taguchi et al., 2005). It is based on the functional relationship between the design factors (X), the noise factors (Z), the input signal (M) and the response variable (y). Optimal settings lead to low sensitivity of the product’s performance to noise factors. Kackar (in Nair, 1992) summarizes the prerequisites of parameter design with the existence of interactions between noise factors and design factors and the ability of engineers to identify the factors involved in such interactions.
The ambition is to find parameter settings that minimize the effects that noise factors (Z) have on the response variable (y), i.e. to prevent that variation propagates through the technical system and causes undesired variation in product performance. This can be achieved by exploiting possible non-linear relations between the design factors and the response variables (e.g. Box & Fung, 1994). Parameter design can be carried out using a number of methods that are all based on the transfer function f (M X Z) .
Theoretical or mathematical models that can describe the transfer function are often not available (e.g. Toutenburg et al., 1998). Consequently, empirical models based on designed experiments are often employed to explain the cause-and-effect relationships. On the other hand, designed experiments can dramatically improve the statistical accuracy of simulation results and facilitate the statistical analysis, as shown by e.g. Wild and Pignatiello (1991).
In the stage after parameter design, tolerance specifications must be determined for all parts and components. If the result achieved after parameter design is not satisfactory and the performance variation is still too high, tolerance design can be applied. According to Clausing (1994) tolerance design always involves a trade-off: if more precision is desired, it must be paid for. In fact, tolerance design is a trade-off between reduction in performance variation and increased manufacturing cost. Taguchi (1993) suggests that a quadratic loss function can be used to specify tolerance limits. The approach aims to choose the tolerance limits such that the overall cost, i.e. the sum of the customer loss caused by deviation from target and the cost to the manufacturer associated with variation reduction, is minimized. More information about how the quadratic loss function can be deployed in tolerance design can be found in e.g. Taguchi (1993), Clausing (1994) and Taguchi et al. (2005).
Conclusions
The contribution of this paper is a clarification of the link between the arisen DFSS concept and the well-established PDP. While the concept of DFSS is not so well-defined with definitions mostly on a conceptual level, the traditional PDP has established over time and its different stages with their deliverables are clear. Here it is shown where and how DFSS can contribute to the respective stages of a general PDP.
Design for Six Sigma efforts should be initiated early in the PDP since this is where the foundation of the design is laid, as e.g. Ford (1996) argues. Also at that time customer needs can be taken into account systematically by e.g. the application of QFD. Tools and methods are necessary aids in DFSS but cannot automatically generate six sigma quality levels and increased customer satisfaction. Important aspects are taking the customer’s perspective, an awareness of variation and thinking in terms of robustness.
The identification of the customer needs and wants and potential noise factors including their impact on product performance are particularly important early in design, where the use of quantitative methods is limited. Later on when more detailed design information is available the transfer function should be estimated and, on the basis of this estimation, optimal parameter settings should be determined. Depending on the situation different tools and methods are available to achieve this. When no mathematical model can be built, designed experiments on prototypes can deliver an empirical model and design factors should be set in a way that makes the product robust, i.e. insensitive to noise factors.
_____________________________________________
About the author:
Torben Hasenkamp is Ph.D. Candidate in Quality Sciences at Chalmers University of Technology, Sweden. He is actively involved in DFSS implementation programs at SKF and Volvo. Before taking up his Ph.D. studies he earned his M.Sc. from the same University and a Diploma in industrial engineering from the Technical University of Hamburg-Harburg.
_____________________________________________
References
Andersson, P. (1997). “On Robust Design in the Conceptual Design Phase – A Qualitative Approach.” Journal of Engineering Design 8(1): 75-90.
Antony, J. (2002). “Design for Six Sigma: A Breakthrough Business Improvement Strategy for Achieving Competitive Advantage.” Work Study 51(1): 6-8.
Bandurek, G. R., H. L. Hughes, et al. (1990). “The Use of Taguchi Methods in Performance Demonstrations.” Quality and Reliability Engineering International 6: 121-131.
Bergman, B. and B. Klefsjö (2003). Quality from Customer Needs to Customers Satisfaction. Lund, Studentlitteratur.
Berryman, M. L. (2002). “DFSS and Big Payoffs.” Six Sigma Forum Magazine 2(1): 23-28.
Box, G. and C. A. Fung (1994). “Quality Quandaries – Is Your Robust Design Procedure Robust?” Quality Engineering 6(3): 503-514.
Chowdhury, S. (2002). Design For Six Sigma. Chicago, Dearborn Trade Publishing.
Chowdhury, S. (2003). The Power of Design for Six Sigma. Chicago, Dearborn Trade.
Clausing, D. (1994). Total Quality Development – A Step-By-Step Guide to World-Class Concurrent Engineering. New York, ASME Press.
Creveling, C. M., J. L. Slutsky, et al. (2003). Design for Six Sigma – In Technology and Product Development. New Jersey, Prentice Hall PTR.
Cross, N. (1994). Engineering Design Methods – Strategies for Product Design. New York, John Wiley & Sons.
De Feo, J. A. (2002). “Viewpoint: Surviving the present and the future.” Measuring Business Excellence 6(3): 62-64.
El-Haik, B.S. (2005). Axiomatic Quality: Integrating Axiomatic Design with Six Sigma, Reliability, and Quality Engineering. NJ, John Wiley & Sons.
Ford, R. B. (1996). Process for the Conceptual Design of Robust Mechanical Systems – Going Beyond Parameter Design to Achieve World-Class Quality. Mechanical Engineering. Stanford, Stanford University: 329.
French, M. J. (1994). “Annotated list of design principles.” Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture 208(B4): 229-234.
Fundin, A. and P. Cronemyr (2003). “Use Customer Feedback to Choose Six Sigma Projects” Six Sigma Forum Magazine. 3(1): 17-21
Gremyr, I. (2005). “Exploring Design for Six Sigma from the viewpoint of Robust Design Methodology” International Journal of Six Sigma and Competitive Advantage 1(3): 295-306
Johansson, P., A. Chakhunashvili, S. Barone and B. Bergman (2006). “Variation Mode and Effect Analysis: a Practical Tool for Quality Improvement” Quality and Reliability Engineering International 22(8): 865-876.
Kano, N., N. Seraku, F. Takahashi, and S. Tsuji (1984). Attractive quality and must-be quality, Hinshitsu (Quality, The Journal of Japanese Society for Quality Control), 14: 39-48.
Kogure, M. and Y. Akao (1983). “Quality Function Deployment and CWQC in Japan” Quality Progress 16(10): 25-29.
Mader, D. P. (2002). “Design for Six Sigma.” Quality Progress July: 82-86.
Nair, V. N. (1992). “Taguchi’s Parameter Design: A Panel Discussion.” Technometrics 34(2): 127-161.
Pahl, G. and W. Beitz (1996). “Engineering Design.” Glasgow, Bell & Bain Limited.
Park, S. H. (1996). “Robust Design and Analysis for Quality Engineering.” London, Chapman & Hall.
Phadke, M. S. (1989). “Quality Engineering using Robust Design.” Englewood Cliffs, New Jersey, Prentice Hall.
Priest, J. W. and J. M. Sánchez (2001). “Product Development and Design for Manufacturing.” New York, Marcel Dekker.
Pugh, S. (1991). “Total Design, Integrated Methods for Succesful Product Engineering.” Harlow, Addison-Wesley.
Roozenburg, N. F. M. and J. Eekels (1995). “Product Design: Fundamentals and Methods.” New York, Wiley & Sons.
Soderborg, N. (2004) “Design for Six Sigma at Ford” Six Sigma Forum Magazine, November: 15-22
Suh, N. P. (1990). “The Principles of Design.” New York, Oxford University Press.
Sullivan, L. P. (1986). “Quality Function Deployment.” Quality Progress.
Taguchi, G. (1986). “Introduction to Quality Engineering – Designing Quality into Products and Processes”. Tokyo, Asian Productivity Organization.
Taguchi, G. (1993). “Taguchi on Robust Technology Development – Bringing Quality Engineering Upstream”. New York, ASME Press.
Taguchi, G., S. Chowdhury, et al. (2005). “Taguchi’s Quality Engineering Handbook”. Hoboken, New Jersey, John Wiley & Sons.
Taguchi, G. and D. Clausing (1990). “Robust Quality.” Harvard Business Review 68(1): 65-75.
Tennant, G. (2002). “Design for Six Sigma: Launching New Products and Services Without Failure”. Hampshire, Gower Publishing Limited.
Thornton, A. (2004). “Variation Risk Management, Focusing Quality Improvements in Product Development and Production.” Hokboken, New Jersey, John Wiley & Sons, Inc.
Toutenburg, H., R. Gössl, et al. (1998). “Quality Engineering – Eine Einfuhrung in Taguchi-Methoden”. Munich, Prentice Hall.
Ullman, D. G. (1997). “The Mechanical Design Process”. Singapore, McGraw-Hill.
Ulrich, K. T. and S. D. Eppinger (1995). Product Design and Development. Singapore, McGraw-Hill.
Wild, R. H. and J. J. Pignatiello (1991). “An experimental design strategy for designing robust systems using discrete-event simulation.” Simulation December: 358-368.
Xie, M., K.C. Tan and T.N. Goh (2003). “Advanced QFD Applications”. Milwaukee. ASQ Quality Press
Yang, K. and B. El-Haik (2003). “Design for Six Sigma – A Roadmap for Product Development”. New York, McGraw-Hill.
