A THREE-STAGE METHODOLOGY FOR DESIGN EVALUATION IN PRODUCT DEVELOPMENT FAIZ

In order to remain competitive in today’s technologically driven world, the faster and more efficient development of innovative products has become the focus for manufacturing companies. In tandem with this, design evaluation plays a critical role in the early phases of product development, because it has significant impact on the downstream development processes as well as on the success of the product being developed. Owing to the pressure of primary factors, such as customer expectations, technical specifications and cost and time constraints, designers have to adopt various techniques for evaluating design alternatives in order to make the right decisions as early as possible. In this work, a novel three-stage methodology for design evaluation has been developed. The preliminary stage screens all the criteria from different viewpoints using House of Quality (HoQ). The second stage uses a Fuzzy-Analytical Hierarchy Process (Fuzzy-AHP) to obtain the alternatives weighting and the final stage verifies the ranking of the alternatives by a Rough-Grey Analysis. This method will enable designers to make better-informed decisions before finalising their choice. Case examples from industry are presented to demonstrate the efficacy of the proposed methodology. The result of the examples shows that the integration of Fuzzy-AHP with HoQ and Rough-Grey Analysis provides a novel alternative to existing methods of design evaluation.


Background
The product development process is one of transformation from customer requirements to a physical structure with consideration of the various design constraints (Li et al., 2010). For a long time, new product development has been considered an essential element for organisational competitiveness and success (Edwards et al., 2005). Product development also plays a critical role in the survival and success of manufacturing enterprises and many researchers have improved their understanding of the need for its strategic management (Brown & Eisenhardt, 1995;Griffin & Hauser, 1996;Krishnan & Ulrich, 2001;Chesbrough & Teece, 2002;Ayag & Odzemir, 2008). However, truly effective product development remains difficult (Lee & Santiago, 2008). A study by Minderhoud & Fraser (2005) indicates that product development practices have evolved over recent years as product cost; quality and time-to-market have each become progressively important. In parallel, the rapid pace of technological development has led to shorter product life cycles for many product categories, most notably in consumer electronics.
Following the identification of a market (user need), a total design system, as espoused by Pugh (1996), is a systematic activity that is necessary to produce and sell a successful product to satisfy that need; the activity encompasses product, process, people and organisation. In accordance with this, Ebuomwan et al. (1996) proposed that the total design activity model consists principally of a central design core, which in turn comprises a market (user need), product design specification, conceptual design, detailed design, manufacture and sales. Pahl et al. (2007) classify the activities of designers into conceptualising, embodying, detailing and computing, drawing and collecting information. Wallace (1989) points out that "the engineering design process cannot be carried out efficiently if it is left entirely to chance..." (p.35). Furthermore, Finger & Dixon (1989b) mentioned that the mapping between the requirements of a design and the attributes of the artefact is not fully understood.
Because the goal of design is to create artefacts that meet functional requirements, further fundamental research is needed on relating the attributes of designs to those functional requirements, that is, on prescribing the artefact. In addition, Chandrasegaran et al., (2013) stated that product design is a highly involved, often ill-defined, complex and iterative process and that the needs and specifications of the required artefact become more refined only as the design process moves towards its goal.
In today's industries, product design has become the main focus in a highly competitive environment and fast-growing global market (Turan & Omar, 2012;. The benchmarks used to determine the competitive advantage of a manufacturing company are customer satisfaction, shorter product development time, higher quality and lower product cost (Hsu & Woon, 1998;Subrahmanian et al., 2005;Shai et al., 2007). Today's product designer is being asked to develop highquality products at an ever increasing pace (Ye et al., 2008). To meet this challenge, new and novel design methodologies that facilitate the acquisition of design knowledge and creative ideas for later reuse are much sought after. In the same context, Liu & Boyle (2009) highlighted that the challenges currently faced by the engineering design industry are the need to attract and retain customers, the need to maintain and increase market share and profitability and the need to meet the requirements of diverse communities. Tools, techniques and methods are being developed that can support engineering design with an emphasis on the customer, the designer and the community (Chandrasegaran et al., 2013). Thus, a good design process should take into account the aforementioned criteria as early as possible in order to ensure the success of a product (Turan & Omar, 2012;. One important step in designing new products is generating conceptual designs (Turan & Omar, 2013). The conceptual design process includes a set of technical activities, which are the refinement of customer requirements into design functions, new concept development and the embodiment engineering of a new product (Li et al., 2010). A study by Lotter (1986) indicates that as much as 75% of the cost of a product is being committed during the design phase. In the same context, Nevins & Whitney (1989) surmise that up to 70% of the overall product development cost is committed during the early design phases. Furthermore, Ullman (2009) points out that 75% of the manufacturing cost is committed early in the design process.
Under such circumstances, the design concept evaluation in the early phase of product development plays a critical role because it has a significant impact on downstream processes (Zhai et al., 2009). Similarly, Geng et al. (2010) point out that design concept evaluation, which is at the end of the conceptual design process, is one of the most critical decision points during product development. It relates to the ultimate success of product development, because a poor design concept can rarely be compensated in the latter stages.
Design concept evaluation is a complex multi-criteria decision-making (MCDM) process, which involves many factors ranging from initial customer needs to the resources and constraints of the manufacturing company. Concept design selection is the process of evaluation and selection from a range of competing design options with respect to customer needs and other criteria, comparing the relative strengths and weaknesses of the concept design and selecting one or more concept designs for further investigation, testing, or development (Green, 2000). However, how to evaluate effectively and objectively design concepts at the early stage of product development has not been well addressed, because the information available is usually incomplete, imprecise, and subjective or even inconsistent (Rosenman, 1993). As such, the quest for more effective and objective approaches to evaluate systematically design concepts in the early stage of the design process has invoked much research interest.
The success of the completed design depends on the selection of the appropriate concept design alternative (Green, 1997;Ulrich & Eppinger, 2005;Zhai et al., 2009). A mismatch between the customer's need and the product and manufacturing process causes loss of quality, delay to market and increased costs (Millson et al., 2004). Changes made early in the design process are less costly than those made during detailed design and later stages (Childs, 2004). Any design defect in the conceptual design is very difficult to correct in the detailed design stage and will incur further costs in the future (Francis et al., 2002). The process of choosing the concept design is frequently iterative and may not produce immediately a dominant concept design (Liu et al., 2003). An initially large set of concept design alternatives should be screened down to a smaller set, because some would clearly not be feasible for reasons, such as infeasibility of manufacturing or the cost of production concept d  (Scott, 2002;Ayag & Odzemir, 2007b). In accordance with this, an ideal design evaluation method, as espoused by Ayag & Odzemir (2007b), Zhai et al. (2009) and Turan & Omar (2013), needs to use fewer numbers of design criteria, fewer numbers of pair-wise comparisons and have a support tool to verify and validate the ranking of the alternatives obtained.
The conventional Fuzzy-AHP method aims to use an optimum number of pair-wise comparisons. In AHP, pair-wise comparisons are often preferred by the decision makers, because they facilitate the weighting of criteria and scores of alternatives from comparison matrices, rather than quantifying the weights or scores directly (Javanbarg et al., 2012). In many practical situations, the human preference model is uncertain and decision makers might be reluctant or unable to assign exact numerical values to the comparison judgements. Although the use of the discrete scale for performing pair-wise comparative analysis has the advantage of simplicity, a decision maker might find it extremely difficult to express the strength of his preferences and to provide exact pair-wise comparison judgements in relation to the design criteria (Triantaphyllou & Lin, 1996;Duran & Aguilo, 2007). Consequently, the decision makers will need a process of reconsideration of design alternatives in relation to the design criteria, which might not help them reduce the number of design criteria. In addition, the final weight of design alternatives might not produce significant differences, which will affect the designers or decision makers when making a judgement. Thus, a sole conventional Fuzzy-AHP is insufficient when applied to ambiguous problems.
With the Fuzzy-AHP method, designers also face the same issues in design evaluation for new product development. A study by Zhai et al. (2009) indicates that although the Fuzzy-AHP method offers many advantages for design concept evaluation, it can be a time-consuming process due to the increase in the number of design criteria and design concepts. This might result in a huge evaluation matrix and the need to conduct a large number of pair-wise comparisons, which might lead to low consistency (Ayag & Ozdemir, 2007b).

Objective
The following defines in more detail what this work intends to achieve. Thus, it will be possible to evaluate later on, whether the steps chosen in the proposed methodology have led to successful results.
The overall aim of the research is formulated as follows: To develop a novel methodology for design evaluation that enables designers to make better-informed decisions than conventional method when finalising their choice.
This research proposes a novel three-stage method of design evaluation using the integration of Fuzzy-AHP with House of Quality (HoQ) and the Rough-Grey Analysis approach.
As the overall aim is broad, it has been divided into single objective in order to support its achievement. The objective of this research, as depicted in Figures 1.3 and 1.4 is to develop a method of interfacing Fuzzy-AHP with HoQ and Rough-Grey Analysis as the following steps: (i) Introduce the scale of "Weighting criteria" for survey process prior to the first stage of design evaluation, which is a screening process using the HoQ method. HoQ will reduce the number of design criteria.
(ii) Introduce the method of computing the priority of element for constructing the pair-wise comparison matrix to execute the second stage of design evaluation, which is Fuzzy-AHP method with fewer numbers of pair-wise comparisons using the results from the first stage.
(iii) Introduce the method of quantifying the attribute ratings ⊗v to carry out the third stage of design evaluation, which is verification and validation stage using the Rough-Grey Analysis method. This stage will reduce the unnecessary iteration process.
The final target of the proposed approach is to help the design community become better-informed than conventional method before making final judgements and consequently, reduce development time and cost.
where, As the scope of product development is too broad, this research will focus on prescriptive models of engineering design because this provides a systematic procedure for facilitating the design operations. Design operations will be limited to the conceptual design and embodiment design process and will focus entirely on the design evaluation, which is the integration of the Fuzzy-AHP method. Huang et al. (2006) mentioned that researchers had integrated fuzzy sets with other generic algorithms and neural networks to formulate an integrated approach for design concept generation and evaluation. In the same context, many researchers have successfully used fuzzy sets in engineering design evaluation (Carnahan et al., 1994;Khoo & Ho, 1996;Sun et al., 2000;Wang, 2001;Tsai & Hsiao, 2004). Furthermore, Fuzzy-AHP as one of the most commonly used MCDM techniques, has been adapted to evaluate alternatives of conceptual design (Zhai et al., 2009).
In summary, the proposed method of design evaluation process is expected to strengthen or improve the product being evaluated, or to maintain the product at an optimal level of specification and improve the operational time and cost.

Organisation of thesis
The thesis structure, as indicated in Figure 1.6, is as follows. Chapter 1 presents the introduction of the research. The first part of Chapter 1 describes the background of the research, followed by a presentation of the specific problem to be addressed. The third and fourth parts describe the objectives and scope of the research, respectively and the final part describes the organisation of the thesis.
Chapter 2 comprises ten parts that discuss the design model, prescriptive

Design model
According to Cross (2000) and Darlington & Culley (2002), Engineering Design Research (EDR) has customarily been partitioned into prescriptive and descriptive work and design support tools will be allocated under the two main headings as seems appropriate to their provenance. An additional partition of 'design automation' has been added, because this appears to the present authors to be a quite separate research focus (Darlington & Culley, 2002). A taxonomy of the categorisations is shown in Figure 2.1.  Ebuomwan et al. (1996) highlight that design models are the representations of philosophies or strategies that propose to show how design is. Often, they are drawn as flow diagrams, showing the iterative nature of the design process via a feedback link. Generally, from various philosophical viewpoints, design models can be divided into two main classes: prescriptive and descriptive models. However, another class can be added, known here as computational models, which emphasise the use of quantitative and qualitative computational techniques and artificial intelligence techniques, combined with modern computing technologies (Ebuomwan et al., 1996;Cross, 2000).
The prescriptive models tend to look at the design process from a global perspective, covering the procedural steps. They prescribe how the design process ought to proceed and sometimes suggest how best to carry it out. On the other hand, the descriptive models are concerned with designers' actions and activities during the design process. This comes from both experience of individual designers and from studies carried out.  In short, prescriptive design models suggest the best way for how something should be done, whereas descriptive models give details on what is involved in designing and/or how it is done (Ebuomwan et al., 1996).

Prescriptive design process model
Prescriptive models of design process are concerned with trying to persuade or encourage designers to adopt improved ways of working. They usually offer a more algorithmic, systematic procedure to follow and they are often regarded as providing a particular design methodology. These models emphasise the need for further analytical work to develop the generation of solution concepts. The intention is to try to ensure that the design problem is fully understood, that no important elements of it are overlooked and that the real problem is identified.
The prescriptive models of both Taguchi and Suh are applied in practice and they have resulted in less expensive and more robust designs (Finger & Dixon, 1989a). In accordance with this, Pahl et al. (2007) introduced their model of the design process with the following stages: (i) Clarification of the task: Collect information about the requirements to be embodied in the solution and about the constraints.
(ii) Conceptual design: Establish function structures; search for the suitable solution principles; combine into concept variants.
(iii) Embodiment design: Starting from the concept, the designer determines the layout and forms and develops a technical product or system in accordance with technical and economic considerations.  In short, the prescriptive approach to design is concerned with the formalisation of process by means of encouraging better or more efficient performance by practicing engineers (Pugh, 1996;Shaw et al., 2001;Ulrich & Eppinger, 2005;Pahl et al., 2007;Ullman, 2009).

Design concept evaluation
Design concept evaluation is a complex MCDM process involving large amounts of data and expert knowledge, which are usually imprecise and subjective (Zhai et al., 2009). Finger & Dixon (1989a) mentioned that in conceptual design, functional requirements are transformed into a physical embodiment or configuration. In the same manner, Ulrich & Seering (1987a, 1987b, 1987c defined conceptual design as the transformation from functional and behavioural requirements to structural descriptions; that is, to configurations. Design concept evaluation can be classified into two categories: non-numerical methods and numerical methods (Ayag & Odzemir, 2007a). Generally, non-numerical methods are relatively simple, fast and are more suitable for quick screening of design concepts for simple applications. In contrast, numerical methods are more systematic and can assist designers in achieving evaluations that are more accurate, especially for complex design concepts.  A study by Zhang & Chu (2009) indicates that design concept evaluation is a complex MCDM problem, which involves many factors ranging from task-related factors (e.g., product complexity, initial customer requirements impreciseness and information scarcity) to decision related factors (e.g., the expertise and diversity of decision makers (DMs) and the method of aggregating judgements). Data and information involved in this problem come from design knowledge and experiences at earlier design stages and subjective judgements of DMs. At the earlier design stages, design information is deficient and imprecise. DMs' judgements often lack precision and the confidence levels in them contribute to various degrees of uncertainty (Lo et al., 2006). Therefore, coping with uncertainty and the vague characteristics of information is critical to the effectiveness of the process of decision making. Furthermore, the aggregation method of individual judgements in group decision making and the alternatives ranking method in the evaluation model, are critical to the accuracy and effectiveness of design concept evaluation (Geng et al., 2010).
In short, design concept evaluation in the early phase of product development plays a critical role because it has a significant impact on downstream processes (Zhai et al., 2009). In addition, early design concept evaluation can save both cost and time in product development.

Classical AHP
In situations where DMs might have difficulties in determining accurately the various factor weights and evaluations, the Analytical Hierarchy Process (AHP) method can be used (Chatterjee & Mukherjee, 2010). In AHP, the DM starts by laying out the overall hierarchy of the decision. This hierarchy reveals the factors to be considered as well as the various alternatives in the decision. Here, both qualitative and quantitative criteria can be compared using a number of pair-wise comparisons, which result in the determination of factor weights. Finally, the alternative with the highest total weighted score is selected as the best option (Saaty, 1980).
The basic principle of AHP is to construct a matrix expressing the relative values of a set of attributes.  (Nepal et al., 2010). For computing the priorities of the elements, a judgemental matrix (also known as a pair-wise comparison matrix) is constructed, as shown below (Saaty, 1977). In short, the classical AHP method is incapable of handling the uncertainty and vagueness involved in the mapping of one's preference to an exact number or ratio (Chatterjee & Mukherjee, 2010). The major difficulty with classical AHP is its inability to map human judgements. It has been observed that because of confusion in the DM's mind, probable deviations should be integrated into the decision-making process (Askin & Guzin, 2007).

2.5
Other existing tools

Fuzzy-TOPSIS
The technique for order preference by similarity to ideal solution (TOPSIS) is a useful technique in dealing with multi-attribute or multi-criteria problems of decision making (MADM/MCDM) in the real world (Hwang & Yoon, 1981). The positive ideal solution (PIS) is a solution that maximises the benefit criteria/attributes and minimises the cost criteria/attributes. The negative ideal solution (NIS) maximises the cost criteria/attributes and minimises the benefit criteria/attributes (Chen, 2000).
The best alternative is the one that is closest to the PIS and furthest from the NIS (Herrera et al., 1996;Herrera & Herrera-Viedma, 2000).
The use of numerical values in the rating of alternatives might have limitations when dealing with uncertainties. Therefore, extensions of TOPSIS were developed to solve problems of decision making with uncertain data, which resulted in Fuzzy-TOPSIS (Krohling & Campanharo, 2011). The general steps of the Fuzzy-TOPSIS approach can be summarised as in Figure 2.5.  Table 2.2. The table shows that the major weaknesses of TOPSIS are in not providing for weight elicitation and consistency checking for judgements. However, the use of AHP has been restrained significantly by the human capacity for information processing and thus, the number seven plus or minus two would be the ceiling in the comparison (Saaty & Odzemir, 2003). From this viewpoint, TOPSIS alleviates the requirement of paired comparisons and the capacity limitation might not be as dominant in the process (Shih et al., 2007). Hence, it would be suitable for cases with a large number of attributes and alternatives and especially handy for objective or quantitative data. In short, the disadvantages of the Fuzzy-TOPSIS method are not providing the weight elicitation and consistency checking, which are very useful for the DMs in making judgements.

TRIZ
TRIZ, an acronym for the Theory of Inventive Problem Solving, began in 1946 when Altshuller, a mechanical engineer, began to study patents in the Russian Navy. This approach has been widely taught in Russia but it did not emerge in the West until the late 1980s. Several different solution systems have been derived by abstracting inventive principles from the ongoing analysis of patent data. Several of these solutions focus on contradictions or trade-offs in identifying innovative solutions (Li The basic constituents of TRIZ are the contradictions, 40 inventive principles, the contradiction matrix (Domb, 1997;Zoyzen, 1997), the laws of evolution (Petrov, 2002), the substance-field analysis modelling (Terninko, 2000), the ideal final result (Domb, 1997), substance field resources and scientific effects (Frenklach, 1998). The core of TRIZ consists of 40 contradiction principles and the matrix; other tools are auxiliary in assisting design engineers to construct the problem model and analyse it.
Altshuller's early work on patents resulted in classifying inventive solutions into five levels, ranging from trivial to new scientific breakthroughs (Altshuller, 1999). Figure 2.6 illustrates this abstraction process, which classifies problems and solutions in seeking a correlation that enables a set of generic problem solving operators or principles to be identified.
Figure 2.6: The general case for abstracting a solution system (Lee & Huang, 2009) Over time, Altshuller identified a further level of abstraction from the technical contradictions (Li & Huang, 2009). He found that by defining the contradiction around one parameter with mutually exclusive states, the correlation operators used to detect a solution could be more generic and there are four separation principles used to help resolve this type of contradiction. The separation principles can be summarised as separation of opposite requirements in space, separation of opposite requirements in time, separation within a whole and its parts, and separation upon condition. Figure 2.7 illustrates the relationship between these two levels of abstraction.
Figure 2.7: The first and second levels of abstraction (Li & Huang, 2009) In short, the contradiction matrix table of 40 innovative principles and 39 engineering parameters is used to ascertain the trade-off between design contradictions and engineering parameters. The design engineers can acquire more feasible solutions and inspiration through this method (Li & Huang, 2009). However, owing to vagueness and uncertainty in the DM's judgement, a decision support tool that can represent adequately qualitative and subjective assessments under the multiple criteria decision-making environment is required.

House of Quality (HoQ)
Quality Function Deployment (QFD) was developed in Japan by Mitsubishi in 1972.
This is a structured format used to integrate informational needs (Hauser & Clausing, 1988;Bounds et al., 1994). Applications begin with the HoQ, which is used to understand customer requirements and to translate these requirements into the voice of the engineer (Hauser, 1993). Posterior houses will deploy the requirements up to production requirements.
QFD is an iterative process performed by a multifunctional team (Hauser, 1993).
QFD employs four sets of matrices based on the "what-how" matrix, the so-called HoQ and is used to relate the voice of the customer to a product's technical requirements, component requirements, manufacturing operations and quality control plans (Vairaktarakis, 1999). Figure 2.8 shows the data needed by each of the four