A methodology of integrating affective design with defining engineering specifications for product design

Abstract

Affective design and the determination of engineering specifications are commonly conducted separately in early product design stage. Generally, designers and engineers are required to determine the settings of design attributes (for affective design) and engineering requirements (for engineering design), respectively, for new products. Some design attributes and some engineering requirements could be common. However, the settings of the design attributes and engineering requirements could be different because of the separation of the two processes. In previous studies, a methodology that considers the determination of the settings of the design attributes and engineering requirements simultaneously was not found. To bridge this gap, a methodology for considering affective design and the determination of engineering specifications of a new product simultaneously is proposed. The proposed methodology mainly involves generation of customer satisfaction models, formulation of a multi-objective optimisation model and its solving using a chaos-based NSGA-II. To illustrate and validate the proposed methodology, a case study of mobile phone design was conducted. A validation test was conducted and the test results showed that the customer satisfaction values obtained based on the proposed methodology were higher than those obtained based on the combined standalone quality function deployment and standalone affective design approach.

Full text

This article was downloaded by: [Hong Kong Polytechnic University]
On: 03 March 2015, At: 22:57
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Click for updates
International Journal of Production Research
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tprs20
A methodology of integrating affective design with
defining engineering specifications for product design
Huimin Jianga, C.K. Kwonga, Y. Liub & W.H. Ipa
a Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University,
Hong Kong, China
b Institute of Mechanical and Manufacturing, School of Engineering, Cardiff University,
Cardiff, UK
Published online: 30 Oct 2014.
To cite this article: Huimin Jiang, C.K. Kwong, Y. Liu & W.H. Ip (2015) A methodology of integrating affective design with
defining engineering specifications for product design, International Journal of Production Research, 53:8, 2472-2488, DOI:
10.1080/00207543.2014.975372
To link to this article: http://dx.doi.org/10.1080/00207543.2014.975372
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

A methodology of integrating affective design with defining engineering specifications for
product design
Huimin Jiang
a
, C.K. Kwong
a
*,Y.Liu
b
and W.H. Ip
a
a
Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, China;
b
Institute of
Mechanical and Manufacturing, School of Engineering, Cardiff University, Cardiff, UK
(Received 30 October 2013; accepted 30 September 2014)
Affective design and the determination of engineering specifications are commonly conducted separately in early product
design stage. Generally, designers and engineers are required to determine the settings of design attributes (for affective
design) and engineering requirements (for engineering design), respectively, for new products. Some design attributes
and some engineering requirements could be common. However, the settings of the design attributes and engineering
requirements could be different because of the separation of the two processes. In previous studies, a methodology that
considers the determination of the settings of the design attributes and engineering requirements simultaneously was not
found. To bridge this gap, a methodology for considering affective design and the determination of engineering specifica-
tions of a new product simultaneously is proposed. The proposed methodology mainly involves generation of customer
satisfaction models, formulation of a multi-objective optimisation model and its solving using a chaos-based NSGA-II.
To illustrate and validate the proposed methodology, a case study of mobile phone design was conducted. A validation
test was conducted and the test results showed that the customer satisfaction values obtained based on the proposed
methodology were higher than those obtained based on the combined standalone quality function deployment and stand-
alone affective design approach.
Keywords: product design; affective design; determination of engineering specifications; chaos optimisation algorithm;
NSGA-II
1. Introduction
In the early product design stage, the two processes, affective design and the determination of engineering specifications,
are always involved, especially for consumer product design. Affective design has been shown to excite customers’psy-
chological feelings and can help improve customer satisfaction in terms of emotional aspects. Affective design involves
the processes of identifying, measuring, analysing and understanding the relationship between the affective needs of the
customer domain and the perceptual design attributes in the design domain (Lai, Chang, and Chang 2005). Design attri-
butes, such as shape and colour, evoke the affective responses of customers to products. Products with good affective
design can attract customers and influence their choices and preferences, such as loyalty to the company and joy of use
(Noble and Kumar 2008). A product usually is associated with a number of engineering requirements, such as weight,
size, processing speed, and power consumption that could affect customer satisfaction. Therefore, it is crucial for prod-
uct development teams to identify appropriate or even optimal engineering specifications for improving or maximising
customer satisfaction (Deng and Pei 2009). To determine engineering specifications for new products, quality function
deployment (QFD) is commonly used to translate the collected customer requirements to various engineering require-
ments for product design. New product design with QFD can enhance organisational learning and improve customer sat-
isfaction. It can also enable a company to reduce product costs, simplify manufacturing processes and shorten
development time of new products (Vonderembse and Raghunathan 1997).
In new product development projects, it is noted that some design attributes considered in affective design could be
identical to some engineering requirements considered in the determination of engineering specifications. For example,
setting the thickness of a new notebook computer may need to be considered in both affective design and the determina-
tion of engineering specifications. However, affective design and the determination of engineering specifications are
always conducted separately. Thus, the settings of engineering requirements and design attributes of a new product
based on existing practice could be different and may not lead to the maximum customer satisfaction to be obtained for
*Corresponding author. Email: c.k.kwong@polyu.edu.hk
© 2014 Taylor & Francis
International Journal of Production Research, 2015
Vol. 53, No. 8, 2472–2488, http://dx.doi.org/10.1080/00207543.2014.975372
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

the new product. Therefore, it is desirable to have a framework or a methodology that considers affective design and
the determination of engineering specifications simultaneously in determining the settings of engineering requirements
and design attributes. However, no such framework or methodology has been found thus far in previous studies.
To fill the existing research gap, a methodology of simultaneous consideration of affective design and the determina-
tion of engineering specifications is proposed to determine design attribute settings and engineering requirement settings
for a new product. In the proposed methodology, a chaos-based fuzzy regression (FR) approach is proposed to develop
customer satisfaction models based on QFD. The generated fuzzy polynomial models can contain second- and/or
higher-order terms and interaction terms such that nonlinearity of the modelling can be better captured. For affective
design, rough set and particle swarm optimisation (PSO)-based adaptive neural fuzzy inference system (ANFIS)
approaches are proposed to model the affective relationships in order to make up the deficiency of ANFIS and further
improve the modelling accuracy. The two types of customer satisfaction models are then used to formulate an optimisa-
tion model. The model is solved using a chaos-based non-dominated sorting genetic algorithm-II (NSGA-II) and the
optimal settings of the engineering requirements and design attributes of a new product can be determined. The pro-
posed methodology can yield higher customer satisfaction values to be obtained compared with the combined standalone
QFD and standalone affective design method. We organise the rest of this paper as follows: first, a review of the related
research is presented in Section 2. The proposed methodology is described in Section 3. Section 4describes the applica-
tion of the proposed methodology in the design of mobile phone products. The validation of the proposed methodology
and its results is presented in Section 5. Finally, discussion and conclusion are given in Section 6and 7, respectively.
2. Literature review
In the following, literature review of affective design, modelling customer satisfaction and defining engineering specifi-
cations, as well as combining Kansei engineering with QFD is provided in Sections 2.1–2.3 respectively.
2.1 Previous studies on affective design
Affective design is a systematic approach to the analysis of customer reactions to candidate designs (Barnes and Lillford
2009). It aims to quantify such reactions and integrate them into physical product design parameters to maximise cus-
tomer affective satisfaction with a product. Nagamachi (1995) proposed Kansei engineering or named affective engineer-
ing (Nagamachi 2008), which is a product development methodology for acquiring and transforming customer
affections into design attribute settings using quantitative methods. Surveys are always required in Kansei engineering
which study the affective meanings related to a product domain based on the semantic differential method (Chuang and
Ma 2001). Kansei engineering has been applied in various affective product designs, such as interior design of automo-
biles (Jindo and Hirasago 1997) and drink bottles (Barnes and Lillford 2009). The framework of Kansei engineering
encompasses four tasks (Nagamachi 1995; Barnes and Lillford 2007) which are the definition of the product domain,
determination of dimensions of customer affections, determination of design attributes and attribute options and evalua-
tion of relationships between customer affections and design attributes. One important task of the Kansei engineering
framework is the evaluation of relationships between the defined affective dimensions and the design attributes. Various
approaches have been attempted in previous studies on modelling the affective relationships such as quantification the-
ory I (Chang 2008), ordinal logistic regression (Barone, Lombardo, and Tarantino 2007), partial least-squares analysis
(Nagamachi 2008), artificial neural network (Lai, Lin, and Yeh 2005; Chen, Khoo, and Yan 2006), fuzzy logic approach
(Lau et al.2006; Lin, Lai, and Yeh 2007), FR (Sekkeli et al.2010), genetic programming-based FR (Chan, Kwong,
et al. 2011), support vector regression model (Yang and Shieh 2010) and ANFIS (Kwong, Wong, and Chan 2009).
Orsborn, Cagan, and Boatwright (2009) quantified aesthetic form preference using utility functions. In addition, some
previous studies attempted to discover the interactions between customer affections and design attributes. Park and Han
(2004) developed fuzzy rule-based models for explaining the relationship between affective user satisfaction and product
design attributes. Jiao, Zhang, and Helander (2006) developed a Kansei mining system for generating Kansei mapping
patterns using associated rule mining. Zhai, Khoo, and Zhong (2009) proposed a rough set-based decision support
approach to study the interactions between customer affective needs and product attributes. Fung et al. (2012) employed
a multi-objective genetic algorithm approach to generate approximate rules that can be used to determine the lower and
upper limits of the affective effect of design patterns. The mined rules were then introduced to guide genetic algorithm
for searching optimal affective design (Fung et al. 2014).
In the early product design stage, one of the key tasks of undertaking affective design is to determine the optimal set-
tings of the design attributes for affective aspects of products to achieve maximum customer satisfaction. Some studies
have been attempted to determine the optimal settings. Conjoint analysis was introduced to determine the optimal setting
International Journal of Production Research 2473
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

of design attributes in product design (Shi, Olafsson, and Chen 2001). Hong, Han, and Kim (2008) proposed a variant of
multiple response surfaces methodology for optimally balancing affective dimensions. A setting of design attributes which
optimally balance the luxuriousness, attractiveness and overall satisfaction was obtained. Aktar Demirtas, Anagun, and
Koksal (2009) adopted an ordinal logistical regression to determine an optimal design attribute settings by maximising the
overall preference scores. Hsiao and Tsai (2005) employed genetic algorithms to search for a near optimal design which
would satisfy the required product image using a trained neural network as a fitness function.
2.2 Previous studies on modelling customer satisfaction and defining engineering specifications
Numerous studies of modelling customer satisfaction based on QFD have been conducted. To address the fuzziness of
the modelling, quite a few previous studies have adopted fuzzy set theory to model customer satisfaction in QFD such
as fuzzy rule-based systems (Fung, Popplewell, and Xie 1998), FR-based mathematical programming approach (Chen
et al. 2004) and generalised fuzzy least-squares regression approach (Kwong et al. 2010). Some other artificial intelli-
gence techniques have been introduced to address the nonlinearities of the modelling. Zhang, Bode, and Ren (1996)
proposed a neural network approach to model the functional relationships between customer satisfaction and engineering
requirements. Fung et al. (2002) proposed a parametric optimisation method to develop nonlinear fuzzy models in QFD.
Chan, Kwong, and Wong (2011) proposed a method based on genetic programming (GP) to generate models for relating
customer satisfaction to engineering requirements. The projections to latent structures technique and the partial least-
squares path modelling algorithm were introduced to develop revealed value models that relate customer satisfaction to
engineering requirements (Withanage, Park, and Choi 2010; Withanage et al. 2012).
Some previous research has attempted to develop systematic procedures and methods for setting optimal target val-
ues of engineering requirements in QFD. Wasserman (1993) formulated the QFD planning process as a linear program-
ming model to select a mix of engineering requirements under a limitation of a given target cost. Kim and Park (1998)
developed an integer programming model to determine the target value setting of engineering requirements in order to
maximise customer satisfaction. A mixed-integer linear programming model was proposed by Zhou (1998) to optimise
the improvement of the target values setting and to prioritise engineering requirements through a fuzzy ranking method.
Fung, Law, and Ip (1999) proposed a fuzzy customer requirements inference system to determine design targets by
amalgamating the principles and techniques of the analytic hierarchy process, fuzzy sets theory and bisection method.
Kusiak (1999) presented an approach which allows one to analyse the impact of changing the value of a variable on the
values of engineering requirements by deriving quantitative relationships. Dawson and Askin (1999) developed a nonlin-
ear mathematical programme to determine optimal value settings of the engineering requirements and formulated a cus-
tomer value function which took into account development time constraints and production costs. Kim et al. (2000)
developed a prescriptive fuzzy optimisation model to determine the optimal target values of engineering requirements
by defining parameters, objectives and constraints in a crisp or fuzzy way. Bai and Kwong (2003) introduced an inexact
genetic algorithm approach to solve a fuzzy optimisation model for the determination of target values for engineering
requirements in QFD. Instead of obtaining one set of exact optimal target values, the approach can generate a family of
inexact optimal target values setting within an acceptable satisfaction degree. Kahraman, Ertay, and Buyukozkan (2006)
proposed an integrated framework based on fuzzy-QFD and a fuzzy optimisation model to determine target value set-
tings of the engineering requirements. Chen and Ko (2009) proposed a fuzzy nonlinear programming model based on
Kano’s concept to determine the fulfilment levels of the engineering requirements. Sun, Mei, and Zhang (2009) applied
a fuzzy conversion matrix in a simplified systematic method to acquire a tight limited target range for engineering
requirements. A mathematical programming model was developed by Sener and Karsak (2010) to determine target lev-
els of technical attributes using the customer satisfaction models developed based on FR. Sener and Karsak (2011) later
proposed a combined fuzzy linear regression and fuzzy multiple objective programming approach for setting target
values of engineering requirements.
2.3 Combining Kansei engineering with QFD
Some previous studies have attempted to combine Kansei engineering with QFD for service and product design. Schütte
(2002) discussed three possible ways to combine Kansei engineering with QFD. Hartono et al. (2012) employed Kano
model, Markov chain modelling, QFD and Kansei engineering in service design. In their study, the Kano model and
Kansei engineering were used to determine the significance of Kansei responses. Then, Markov chain modelling was
introduced to determine the future importance of service attributes. Finally, a house of quality (HoQ) was developed
using the future importance weights for hotel service design. Previous studies have shown that the concept of the
2474 H. Jiang et al.
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

combination of QFD and Kansei engineering is possible. However, research on simultaneous consideration of Kansei
engineering and QFD in determining design attribute settings and engineering specifications has not been found.
3. A methodology for integrating affective design with defining engineering specifications
In this paper, a methodology of simultaneous consideration of affective design and the determination of engineering
specifications is proposed to determine design attribute settings and engineering requirement settings for a new product.
In the proposed methodology, a chaos-based FR approach is introduced to model customer satisfaction based on QFD,
while PSO-based ANFIS approach is proposed to model the relationships between affective responses of customers and
design attributes. The generated customer satisfaction models are used to formulate a multi-objective optimisation
model. A chaos-based NSGA-II is proposed in this research to solve the optimisation model from which optimal settings
of the engineering requirements and design attributes for new products can be determined. Figure 1shows a flowchart
of the proposed methodology.
3.1 Modelling customer satisfaction based on QFD
To determine engineering specifications for new products, QFD is introduced in this research. After developing a HoQ,
modelling the relationships between customer satisfaction and engineering requirements based on the HoQ can be per-
formed. Normally, only a small number of data-sets can be obtained from HoQ for the modelling. On the other hand,
development of HoQ always involves human subjective judgement that would lead to the existence of a high degree of
fuzziness in modelling the relationships. Of the approaches used in previous studies for modelling the relationships, FR
was found to be the most suitable one for the modelling, as FR is the only one of those previously attempted
approaches which is able to capture the fuzziness of modelling data and requires only a small number of data-sets for
developing explicit customer satisfaction models. However, FR can only yield a linear type model but the relationships
Figure 1. A proposed methodology.
International Journal of Production Research 2475
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

can be highly nonlinear. To address both the fuzziness and the nonlinear issues in modelling the relationships, in this
research, a chaos-based FR approach is proposed to develop models for relating customer satisfaction and engineering
requirements based on QFD. In the proposed approach, a chaos optimisation algorithm (COA) is introduced to generate
the nonlinear polynomial structures of customer satisfaction models which could contain second- and/or higher order
terms and interaction terms. The FR method is employed to determine the fuzzy coefficients for all the terms of the cus-
tomer satisfaction models. Details of the chaos-based FR approach to modelling customer satisfaction based on QFD
can be referred to the authors’publication (Jiang et al. 2013).
3.2 Conducting a survey for affective design
In this research, product samples need to be identified and collected first for conducting a survey. Then, affective dimen-
sions and design attributes are defined. Design attributes commonly involve categorical and quantitative types. For cate-
gorical attributes, morphological analysis is adopted to create numerical data by emulating the possible options of
attributes. For example, the side shape can be trapezoidal, a parallelogram or polygonal. A market survey is then con-
ducted for gathering affective responses of customers on the product samples. In the survey, images of the product sam-
ples are shown to respondents and they are required to rate various affective dimensions of the product samples. The
semantic differential (SD) method (Chuang and Ma 2001) is adopted in this research to design a SD questionnaire for
collecting affective responses of customers on products. The SD scale can be a five- or seven-point scale with a pair of
affective words.
3.3 Modelling customer satisfaction for affective design
Based on the survey data, PSO-based ANFIS approach is proposed to model the relationships between the affective
responses of customers and design attributes. Since the number of fuzzy rules increases exponentially when the number
of ANFIS inputs increases, rough set theory is first employed to determine the indispensable design attributes and
reduce the number of fuzzy rules for simplifying the structure of ANFIS. Then, PSO-based ANFIS approach is intro-
duced to generate explicit non-linear customer satisfaction models for affective design in which PSO is used to deter-
mine the optimal values of antecedent parameters in membership functions, such that the errors between the predictive
customer satisfaction values and the actual customer satisfaction values can be minimised. Details of the PSO-based
ANFIS approach to modelling customer satisfaction for affective design can be referred to the authors’publication
(Jiang et al. 2012).
3.4 Formulation of an optimisation model for determining design specifications
Once the customer satisfaction models are developed, they can be used to formulate a multi-objective optimisation
model for determining optimal settings of the design attributes and engineering requirements. The two types of gener-
ated customer satisfaction models respectively based on chaos-based FR and PSO-based ANFIS approaches are used to
formulate objective functions of the optimisation model. Constraints of the optimisation model are mainly about the
ranges of design variables and the correlations among the design variables. In this research, chaos-based NSGA-II is
proposed to solve the optimisation problem for the determination of optimal settings of the engineering requirements
and design attributes for new products. The optimisation model can be expressed as follows.
Objectives:
Maximise customer satisfaction value of affective dimension 1;y1
^
.
.
.
Maximise customer satisfaction value of affective dimension k;y
^
k
Maximise customer satisfaction value of customer requirement;ym
^
Subject to:
yk
^=fkðxÞ(1)
ym
^=fmðxÞ(2)
2476 H. Jiang et al.
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

yk
^
;ym
^k(3)
xjmin xjxjmax (4)
xj¼gðxiÞ;where i 6¼ j(5)
where xdenotes design variables of new products, which involves design attributes in affective design and engineering
requirements in QFD; (1) is the nonlinear customer satisfaction model of the kth affective dimension for affective
design, which is generated based on PSO-based ANFIS approach; (2) is the fuzzy polynomial model generated based
on chaos-based FR approach in QFD. kis the minimum value of customer satisfaction, which is set as 3 when a five-
point scale is used in the market survey. (4) are the ranges of settings of the design variables. (5) is the correlation
model generated based on the technical correlation matrix of HoQ for relating the jth design variable and some other
design variables. For example, in mobile phone design, screen size has technical correlation with the weight and width.
3.5 Solving the optimisation model using a chaos-based NSGA-II approach
NSGA-II is an elitist genetic algorithm and commonly used to solve multi-objective optimisation problems. The major fea-
tures of NSGA-II include low computational complexity, parameter-less diversity preservation, elitism and real-valued rep-
resentation (Deb et al. 2002). NSGA-II uses a real-coded simulated binary crossover (SBX) operator and a real-coded
polynomial mutation operator to support crossover and mutation operations directly for real-valued decision variables.
Deb et al. (2002) found that NSGA-II was able to maintain better spread of solutions and had better convergence than
other multi-objective genetic algorithms, such as Pareto-archived evolution strategy and strength Pareto evolutionary
algorithm. A flowchart of NSGA-II is shown in Appendix 1. However, NSGA-II may not provide a good diversity of
Pareto optimal solutions and its searching solutions are easy to be trapped into a local optimum. In this paper, a
chaos-based NSGA-II approach is proposed to solve the multi-objective optimisation problem by which optimal settings
of the engineering requirements and design attributes for product design can be determined. In the proposed approach, a
COA is incorporated into the process of NSGA-II to refine the search range in order to achieve a good diversity of its
Pareto optimal solutions and avoid trapping into a local optimum. Details of the chaos-based NSGA-II are shown in
Appendix 2.
4. Case study
A case study of mobile phone design was conducted based on the proposed methodology to evaluate the effectiveness
of the methodology. With reference to the HoQ for mobile phones developed by Abu-Assab and Baier (2010), six repre-
sentative customer satisfaction dimensions and the related 16 design variables for mobile phone design were identified.
A HoQ for mobile phone design was developed as shown in Figure 2. Ten major competitive mobile phones were iden-
tified and denoted as brands A-J. Four lead users were invited to assess the 10 mobile phones in terms of the six dimen-
sions of customer satisfaction: ‘easy to use’,‘easy to read display’,‘good sound quality’,‘long operation time’,‘good
functionality’and ‘comfortable to hold’.
Customer satisfaction models based on the HoQ were generated using a chaos-based FR approach (Jiang et al.
2013). The following shows the generated customer satisfaction models, respectively, for ‘comfortable to hold’and
‘easy to use’.
~
y1¼14:9991;2:8422 1014

þ1:5036;2:8422 1014

x3þ0:2890;0ðÞx4þ0:3634;5:6843 1014

x1
þð0:0045;0:0077Þx1x2(6)
~
y2¼9:2351;0ðÞþ1:1582;0:2120ðÞx2
6þ0:1594;0:0437ðÞx5x6þ8:3566;0ðÞx6(7)
where ~
y1and ~
y2are the predicted customer satisfaction values of ‘comfortable to hold’and ‘easy to use’, respectively;
and x
1
,x
2
,x
3
,x
4
,x
5
and x
6
are the engineering requirements, weight, height, width, thickness, menu layers and screen
size, respectively.
For affective design of mobile phones, a total of 32 mobile phones of various brands were selected. By consulting
four product designers, nine representative design attributes for mobile phones: top shape, bottom shape, side shape,
function button shape, number buttons style, screen size, thickness, layout and weight, were identified. Four affective
International Journal of Production Research 2477
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

dimensions were used to evaluate the affective design of the mobile phones. They are simple–complex (S–C), unique–
general (U-G), high-tech–classic (H–C) and handy–bulky (H–B). A morphological matrix based on the nine design attri-
butes of the 32 mobile phones was generated as shown in Table 1.
A survey was conducted using a questionnaire, in which a five-point scale was used. A total of 34 design students
and designers were involved in the survey for the assessment of the mobile phone appearance corresponding to the four
affective dimensions. A part of the questionnaire is shown in Figure 3.
A PSO-based ANFIS approach was introduced to generate customer satisfaction models for the affective dimensions
(Jiang et al. 2012). The following shows the generated customer satisfaction models, respectively, for H–B and S–C.
F2¼0:6798ðx4Þ2x10:0891x4ðx1Þ258:9181ðx4Þ2þ1:0502ðx1Þ20:4287x4x1þ659:2281x484:5350x1þ40:2534
0:0362x4x12:8100x40:3523x1þ27:3809 (8)
F3¼40:1216ðx7Þ2x62:6635x7ðx6Þ283:3900ðx7Þ239:3286ðx6Þ2150:5709x7x6þ345:1479x7þ444:8509x6797:9387
6:3155x7x612:2878x75:4294x6þ10:5638 (9)
where F
2
and F
3
are the predicted customer satisfaction values of H–BandS–C, respectively; and x7is the side shape.
For illustrative purpose, two affective dimensions, H–B and S–C, and two customer satisfaction dimensions,
‘comfortable to hold’and ‘easy to use’, are considered in the optimisation. Thus, four customer satisfaction models,
respectively, for H–B, S–C, ‘comfortable to hold’and ‘easy to use’are used to formulate the objective functions of the
Figure 2. The HoQ for mobile phone products (Jiang et al. 2013).
2478 H. Jiang et al.
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

optimisation model. The first objective function is to maximise customer satisfaction in QFD. Considering the impor-
tance of customer requirements as shown in Figure 2, the objective function is formulated by linearly combining the
two customer satisfaction models (6) and (7), respectively, for ‘comfortable to hold’and ‘easy to use’. Since the
importance weights of ‘comfortable to hold’and ‘easy to use’are 3 and 5, respectively, the first objective function can
be formulated as follows:
F1¼314:9991;2:8422 1014

þ1:5036;2:8422 1014

x3þ0:2890;0ðÞx4

þ0:3634;5:6843 1014

x1þð0:0045;0:0077Þx1x2gþ5
9:2351;0ðÞþ1:1582;0:2120ðÞx2
6þ0:1594;0:0437ðÞx5x6þ8:3566;0ðÞx6

(10)
The second and the third objective functions are the generated customer satisfaction models of H–B and S–C,
respectively, which are expressed as (8) and (9), respectively.
Three constraints are involved in the formulation of the multi-objective optimisation model. The first constraint is
the data type of the design variables. Among the seven design variables, x
1
,x
2
,x
3
,x
4
and x
6
are quantitative variables,
which are real numbers. x
5
and x
7
are categorical variables, which are integers. The second constraint is the ranges of
the design variables, which are [78, 165], [90, 120], [43.5, 62.1], [9.9, 16], [2, 5], [1.8, 3.5], and [1, 6] for x
1
,x
2
,x
3
,x
4
,
Table 1. Morphological matrix of the 32 mobile phone samples.
International Journal of Production Research 2479
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

x
5
,x
6
and x
7
, respectively. The third constraint is the technical correlation among the design variables. As shown in the
HoQ of Figure 2, screen size x
6
has technical correlations with the weight x
1
and width x
3
. Statistical regression (SR)
was applied to model the relationship between x
6
and x
1
,x
3
. The confidence interval in SR was set as 95%. The
generated correlation model is shown below.
x6¼1:5090 þ0:0048x1þ0:0670x3(11)
After formulating the multi-objective optimisation model, a chaos-based NSGA-II approach was introduced to determine
the optimal settings of the engineering requirements and design attributes for mobile phone design. In this study, the
population size Nwas set as 100. The distribution index for crossover η
c
and mutation η
m
was both set as 20. The
crossover probability and mutation probability were set as 0.9 and 0.1, respectively. The above parameter settings, sug-
gested by Deb et al. (2002), have been adopted widely by other researchers. The tournament size and the size of the
mating pool are commonly set as 2 and one half of the population size, respectively. The number of generations was set
as 100 and the iteration number of COA was set as 200 through the repeated operations to ensure that the least running
time and optimal solutions are obtained. The proposed chaos-based NSGA-II approach was implemented using Matlab
software. Figure 4shows the Pareto optimal solutions of the multi-objective optimisation problem solved by the
chaos-based NSGA-II approach.
The Pareto front contains a total number of 100 non-dominated solutions, namely all individuals in the population.
Each optimal solution contains settings of the engineering requirements and design attributes. Decision-makers can
choose their optimal solutions according to various scenarios. In this research, the solution with the largest sum of the
values of the four dimensions of customer satisfaction, namely the largest total customer satisfaction value, is recom-
mended to be the optimal solution. The optimal settings of the design attributes and engineering requirements of a new
mobile phone, as well as the customer satisfaction values, can be obtained as shown in Table 2.
5. Validation
In order to evaluate the effectiveness of the proposed methodology, a validation test was conducted. In the test, two
optimisation models were formulated based on the standalone QFD and the standalone affective design, respectively.
Then, the two optimisation models were solved using chaos-based NSGA-II approach, from which optimal settings of
the engineering requirements and design attributes can be obtained. The parameter settings of the chaos-based NSGA-II
used in the three optimisation models, respectively, based on the proposed methodology, the standalone QFD and the
Figure 3. A part of survey questionnaire.
2480 H. Jiang et al.
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

standalone affective design are all the same. The customer satisfaction values obtained based on the proposed methodol-
ogy were then compared with those based on the standalone QFD and the standalone affective design.
5.1 Design optimisation based on the standalone QFD and the standalone affective design
To formulate an optimisation model based on the standalone QFD, Equations (10) and (11) were set as the objective
function and a constraint, respectively. The optimisation problem is a single objective one. Figure 5shows the searching
values of the objective function based on 100 generations.
From the figure, it can be observed that the value of the objective function reaches a maximum value at the 54th
generation and remains the same in the following generations. After solving, an optimal setting of the engineering
requirements for maximising total customer satisfaction value based on the standalone QFD is obtained. The setting and
the customer satisfaction values of ‘comfortable to hold’and ‘easy to use’are shown in the second column of Table 3.
To formulate an optimisation model for the standalone affective design, Equations (8) and (9) were adopted as
objective functions. The optimisation model is a multi-objective one. The third column of Table 3shows the optimal
setting of the design attributes for maximising the total customer satisfaction value.
From Table 3, it can be observed that the customer satisfaction value of ‘easy to use’based on the standalone QFD
is higher than that based on the proposed methodology, while the value of ‘comfortable to hold’based on the standalone
QFD is smaller. The sum of the customer satisfaction values of ‘comfortable to hold’and ‘easy to use’based on the
proposed methodology is slightly higher than that based on the standalone QFD, which are 7.72 and 7.55, respectively.
By comparing the optimisation results based on the standalone affective design with those based on the proposed meth-
odology, it can be seen that the customer satisfaction value of H–B based on the standalone affective design is higher
Figure 4. Pareto solutions based on chaos-based NSGA-II.
Table 2. Optimisation results based on the proposed methodology.
Design variables (units) Settings
Weight x
1
(g) 78
Height x
2
(mm) 119.99
Width x
3
(mm) 57.62
Thickness x
4
(mm) 9.9
Menu layers x
5
2
Screen size x
6
(in) 2.73
Side shape x
7
3rd alternative
Customer satisfaction
Comfortable to hold 3.64
Easy to use 4.08
Handy–bulky (H–B) 3.25
Simple–complex (S–C) 4.99
International Journal of Production Research 2481
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

than that based on the proposed methodology, while the customer satisfaction value of S–C based on the standalone
affective design is smaller than that based on the proposed methodology. The sum of the customer satisfaction values of
H–BandS–C based on the standalone affective design is higher than that based on the proposed methodology, which
are 8.47 and 8.24, respectively.
5.2 Validation test
From Table 3, it can be noted that there are two different settings of x
1
,x
4
and x
6
based on the standalone QFD and the
standalone affective design. Engineers and designers, who are responsible for defining engineering specifications and
affective design, respectively, need to compromise the differences of the settings. In this case study, different scenarios
of the compromise in terms of ratios of the settings based on the standalone QFD and the settings based on the stand-
alone affective design were studied, which are 0.3:0.7, 0.7:0.3, 0.5:0.5, 0.4:0.6, and 0.6:0.4. Taking the ratio 0.3:0.7 as
an example, the compromised settings of x
1
,x
4
and x
6
are calculated as follows:
x0
1¼0:378:33 þ0:779:2¼78:94g(12)
Figure 5. The searching results for the standalone QFD.
Table 3. Comparison of optimisation results based on the standalone QFD, the standalone affective design and the proposed
methodology.
The standalone
QFD
The standalone
affective design
The proposed
methodology
Design variables (units) Weight x
1
(g) 78.33 79.2 78
Height x
2
(mm) 119.64 N/A 119.99
Width x
3
(mm) 60.48 N/A 57.62
Thickness x
4
(mm) 9.91 15.44 9.9
Menu layers x
5
2 N/A 2
Screen size x
6
(in) 2.92 2.13 2.73
Side shape x
7
N/A 5th alternative 3rd alternative
Customer satisfaction Comfortable to hold 3.19 N/A 3.64
Easy to use 4.36 N/A 4.08
Handy–bulky (H–B) N/A 4.97 3.25
Simple–complex (S–C) N/A 3.5 4.99
2482 H. Jiang et al.
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

x0
4¼0:39:91 þ0:715:44 ¼13:78mm (13)
x0
6¼0:32:92 þ0:72:13 ¼2:37in (14)
where x0
1,x0
4, and x0
6refer to the compromised settings of x
1
,x
4
, and x
6
respectively.
Similarly, the compromised settings of x0
1,x0
4, and x0
6for the other ratios can be calculated as shown in Table 4.
Based on the compromised settings, the customer satisfaction values for ‘comfortable to hold’,‘easy to use’,H–B, and
S–C are calculated again using (6), (7), (8) and (9), respectively. The customer satisfaction values obtained based on the
combined standalone QFD and standalone affective design with compromised settings, and the proposed methodology
are compared in Table 4.
From the table, it can be observed that the customer satisfaction values of ‘comfortable to hold’,‘easy to use’,H–B,
and S–C based on the proposed methodology are higher than those based on the combined standalone QFD and stand-
alone affective design method. The total customer satisfaction value based on the proposed methodology is 35.6%,
11.92%, 30.29%, 34.23% and 24.4%, respectively higher than that based on the combined method with the ratios of
0.3:0.7, 0.7:0.3, 0.5:0.5, 0.4:0.6, and 0.6:0.4.
6. Discussion
The proposed methodology provides a scientific and systematic way to determine the optimal settings of the design
attributes and engineering requirements with simultaneous consideration of affective design and the determination of
engineering specifications. The Pareto optimal solutions can be generated based on the proposed methodology. Thus,
product development teams can select solutions in view of different scenarios, such as high simplicity and high quality.
The methodology can help reduce product development time, improve the attractiveness of new products and reduce
possible conflicts between engineers and designers in determining the settings of particular design variables. Although,
in this paper, a case study of mobile phone design was used to demonstrate the proposed methodology, the methodology
is generic and can be applied to the design of different kinds of products, such as household appliance and furniture,
where affective design of products plays an important role in influencing decisions on choice by consumers. However,
the proposed methodology has some limitations. The predefined number of generations was adopted as the stopping cri-
terion for the chaos-based NSGA-II approach. Single objective optimisation can be terminated when its convergence
remains stable over several generations and a satisfactory solution is obtained. However, the determination of proper
trade-off solutions among multi-objective functions is difficult. A higher maximum number of generations can be
adopted to ensure the convergence, but it requires a longer computational time. On the other hand, the proposed
Table 4. Comparison of results.
The combined standalone QFD and
standalone affective design The proposed methodology
of simultaneous
consideration of affective
design and engineering
specifications
Ratios for
compromised
settings 0.3:0.7 0.7:0.3 0.5:0.5 0.4:0.6 0.6:0.4
Design
variables (units)
Weight x
1
(g) 78.94 78.59 78.77 78.85 78.68 78
Height x
2
(mm) 119.64 119.64 119.64 119.64 119.64 119.99
Width x
3
(mm) 60.48 60.48 60.48 60.48 60.48 57.62
Thicknessx
4
(mm) 13.78 11.57 12.68 13.23 12.12 9.9
Menu layers x
5
22222 2
Screen sizex
6
(in) 2.37 2.68 2.53 2.45 2.6 2.73
Side shape x
7
5th
alternative
5th
alternative
5th
alternative
5th
alternative
5th
alternative
3rd
alternative
Customer
satisfaction
Comfortable to
hold
3.06 3.14 3.1 3.08 3.12 3.64
Easy to use 3.31 3.99 3.69 3.51 3.83 4.08
Handy–bulky 2.97 2.31 2.38 2.61 2.28 3.25
Simple–complex 2.43 4.82 3.08 2.69 3.6 4.99
Total customer
satisfaction value
11.77 14.26 12.25 11.89 12.83 15.96
International Journal of Production Research 2483
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

methodology may not be suitable for the design of breakthrough new products as very few or even no competitive
products can be identified.
7. Conclusion
This paper proposes and describes a methodology of simultaneous consideration of affective design and the determina-
tion of engineering specifications to determine design attribute settings and engineering requirement settings for a new
product. The proposed methodology mainly involves the generation of two types of customer satisfaction models,
respectively, based on QFD and affective design, as well as formulation of a multi-objective optimisation model. A
chaos-based NSGA-II approach is proposed to solve the optimisation problem by which optimal settings of engineering
requirements and design attributes for a new product can be determined.
A case study of mobile phone design was conducted to evaluate the proposed methodology. A multi-objective
optimisation model was formulated based on the four customer satisfaction models, which are for ‘comfortable to hold’,
‘easy to use’,H–B and S–C, respectively. Optimal settings of the engineering requirements and design attributes were
determined based a chaos-based NSGA-II approach. Validation test was conducted to evaluate the effectiveness of the
proposed methodology. Determination of optimal settings of the engineering requirements based on the standalone QFD
and determination of optimal settings of the design attributes based on the standalone affective design are described.
The customer satisfaction values obtained from the proposed methodology are compared with those based on the
combined standalone QFD and standalone affective design method. The validation results indicate that higher customer
satisfaction values can be obtained based on the proposed methodology.
In the development of the proposed methodology, it is assumed that the perceptions, needs and desires of customers
regarding particular products are static. In reality, they could be quite dynamic. Future work could consider dynamic
effects in the simultaneous consideration of affective design and the determination of engineering specifications. Further-
more, the methodology can be further extended to perform product line design, which involves the design of various
product variants in a product line to satisfy the needs of various market segments.
Funding
The work described in this paper was substantially supported by a grant from the Research Grants Council of the Hong Kong Special
Administrative Region, China (Project No. PolyU 517113). The work described in this paper was also partially supported by a grant
from the Hong Kong Polytechnic University.
References
Abu-Assab, S., and D. Baier. 2010. “Designing Products Using Quality Function Deployment and Conjoint Analysis: A Comparison
in a Market for Elderly People.”In Advances in Data Analysis, Data Handling and Business Intelligence Studies in Classifica-
tion, Data Analysis, and Knowledge Organization, edited by A. Fink, B. Lausen, W. Seidel, and A. Ultsch, 515–526. Berlin:
Springer.
Aktar Demirtas, E., A. S. Anagun, and G. Koksal. 2009. “Determination of Optimal Product Styles by Ordinal Logistic Regression
Versus Conjoint Analysis for Kitchen Faucets.”International Journal of Industrial Ergonomics 39 (5): 866–875.
Bai, H., and C. K. Kwong. 2003. “Inexact Genetic Algorithm Approach to Target Values Setting of Engineering Requirements in
QFD.”International Journal of Production Research 41 (16): 3861–3881.
Barnes, C., and S. P. Lillford. 2007. “Affective Design Decision-Making –Issues and Opportunities.”CoDesign 3 (sup1): 135–146.
Barnes, C., and S. P. Lillford. 2009. “Decision Support for the Design of Affective Products.”Journal of Engineering Design 20 (5):
477–492.
Barone, S., A. Lombardo, and P. Tarantino. 2007. “A Weighted Logistic Regression for Conjoint Analysis and Kansei Engineering.”
Quality and Reliability Engineering International 23 (6): 689–706.
Chan, K. Y., C. K. Kwong, T. S. Dillon, and K. Y. Fung. 2011. “An Intelligent Fuzzy Regression Approach for Affective Product
Design That Captures Nonlinearity and Fuzziness.”Journal of Engineering Design 22 (8): 523–542.
Chan, K. Y., C. K. Kwong, and T. C. Wong. 2011. “Modelling Customer Satisfaction for Product Development Using Genetic
Programming.”Journal of Engineering Design 22 (1): 55–68.
Chang, C. C. 2008. “Factors Influencing Visual Comfort Appreciation of the Product Form of Digital Cameras.”International Journal
of Industrial Ergonomics 38 (11–12): 1007–1016.
Chen, L. H., and W. C. Ko. 2009. “Fuzzy Approaches to Quality Function Deployment for New Product Design.”Fuzzy Sets and
Systems 160 (18): 2620–2639.
2484 H. Jiang et al.
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

Chen, C. H., L. P. Khoo, and W. Yan. 2006. “An Investigation into Affective Design Using Sorting Technique and Kohonen
Self-Organising Map.”Advances in Engineering Software 37 (5): 334–349.
Chen, Y., J. Tang, R. Y. K. Fung, and Z. Ren. 2004. “Fuzzy Regression-Based Mathematical Programming Model for Quality
Function Deployment.”International Journal of Production Research 42 (5): 1009–1027.
Chuang, M. C., and Y. C. Ma. 2001. “Expressing the Expected Product Images in Product Design of Micro-Electronic Products.”
International Journal of Industrial Ergonomics 27 (4): 233–245.
Cong, S., G. Li, and X. Feng. 2010. “An Improved Algorithm of Chaos Optimization.”Proceedings of the 8th IEEE International
Conference on Control and Automation, Xiamen, June 9–11.
Dawson, D., and R. G. Askin. 1999. “Optimal New Product Design Using Quality Function Deployment with Empirical Value
Functions.”Quality and Reliability Engineering International 15 (1): 17–32.
Deb, K., A. Pratap, S. Agarwal, and T. Meyarivan. 2002. “A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II.”IEEE
Transactions on Evolutionary Computation 6 (2): 182–197.
Deng, W. J., and W. Pei. 2009. “Fuzzy Neural Based Importance-Performance Analysis for Determining Critical Service Attributes.”
Expert Systems with Applications 36 (2): 3774–3784.
Fung, C. K. Y., C. K. Kwong, K. Y. Chan, and H. Jiang. 2014. “A Guided Search Genetic Algorithm Using Mined Rules for Optimal
Affective Product Design.”Engineering Optimization 46 (8): 1094–1108.
Fung, K. Y., C. K. Kwong, K. W. M. Siu, and K. M. Yu. 2012. “A Multi-Objective Genetic Algorithm Approach to Rule Mining for
Affective Product Design.”Expert Systems with Applications 39 (8): 7411–7419.
Fung, R., D. Law, and W. Ip. 1999. “Design Targets Determination for Inter-Dependent Product Attributes in QFD Using Fuzzy
Inference.”Integrated Manufacturing System 10 (6): 376–383.
Fung, R. Y. K., K. Popplewell, and J. Xie. 1998. “An Intelligent Hybrid System for Customer Requirements Analysis and Product
Attribute Targets Determination.”International Journal of Production Research 36 (1): 13–34.
Fung, R. Y. K., J. Tang, Y. Tu, and D. Wang. 2002. “Product Design Resources Optimization Using a Non-Linear Fuzzy Quality
Function Deployment Model.”International Journal of Production Research 40 (3): 585–599.
Hartono, M., T. K. Chuan, S. Ishihara, and J. B. Peacock. 2012. “Incorporating Markov Chain Modelling and QFD into Kansei
Engineering Applied to Services.”International Journal of Human Factors and Ergonomics 1 (1): 74–97.
Hong, S. W., S. H. Han, and K. J. Kim. 2008. “Optimal Balancing of Multiple Affective Satisfaction Dimensions: A Case Study on
Mobile Phones.”International Journal of Industrial Ergonomics 38 (3–4): 272–279.
Hsiao, S. W., and H. C. Tsai. 2005. “Applying a Hybrid Approach Based on Fuzzy Neural Network and Genetic Algorithm to
Product from Design.”International Journal of Industrial Ergonomics 35 (5): 411–428.
Jiang, H. M., C. K. Kwong, W. H. Ip, and T. C. Wong. 2012. “Modeling Customer Satisfaction for New Product Development Using
a PSO-Based ANFIS Approach.”Applied Soft Computing 12 (2): 726–734.
Jiang, Huimin, C. K. Kwong, W. H. Ip, and Zengqiang Chen. 2013. “Chaos-Based Fuzzy Regression Approach to Modeling
Customer Satisfaction for Product Design.”IEEE Transactions on Fuzzy Systems 21 (5): 926–936.
Jiao, J. R., Y. Zhang, and M. Helander. 2006. “A Kansei Mining System for Affective Design.”Expert Systems with Applications 30
(4): 658–673.
Jindo, T., and K. Hirasago. 1997. “Application Studies to Car Interior of Kansei Engineering.”International Journal of Industrial
Ergonomics 19 (2): 105–114.
Kahraman, C., T. Ertay, and G. Buyukozkan. 2006. “A Fuzzy Optimization Model for QFD Planning Process Using Analytic
Network Approach.”European Journal of Operational Research 171 (2): 390–411.
Kim, K. J., H. Moskowitz, A. Dhingra, and G. Evans. 2000. “Fuzzy Multicriteria Models for Quality Function Deployment.”
European Journal of Operational Research 121 (3): 504–518.
Kim, K., and T. Park. 1998. “Determination of an Optimal Set of Design Requirements Using House of Quality.”Journal of
Operations Management 16 (5): 569–581.
Kusiak, A. 1999. Engineering Design: Products, Processes and Systems. Orlando, FL: Academic Press.
Kwong, C. K., Y. Chen, K. Y. Chan, and X. Luo. 2010. “A Generalized Fuzzy Least- Squares Regression Approach to Modeling
Functional Relationships in QFD.”Journal of Engineering Design 21 (5): 601–613.
Kwong, C. K., T. C. Wong, and K. Y. Chan. 2009. “A Methodology of Generating Customer Satisfaction Models for New Product
Development Using a Neuro-Fuzzy Approach.”Expert Systems with Applications 36 (8): 11262–11270.
Lai, H. H., Y. M. Chang, and H. C. Chang. 2005. “A Robust Design Approach for Enhancing the Feeling Quality of a Product: A
Car Profile Case Study.”International Journal of Industrial Ergonomics 35 (5): 445–460.
Lai, H. H., Y. C. Lin, and C. H. Yeh. 2005. “Form Design of Product Image Using Grey Relational Analysis and Neural Network
Models.”Computers & Operations Research 32 (10): 2689–2711.
Lau, T. W., P. C. L. Hui, F. S. F. Ng, and K. C. C. Chan. 2006. “A New Fuzzy Approach to Improve Fashion Product Development.”
Computers in Industry 57 (1): 82–92.
Lin, Y. C., H. H. Lai, and C. H. Yeh. 2007. “Consumer-Oriented Product Form Design Based on Fuzzy Logic: A Case Study of
Mobile Phones.”International Journal of Industrial Ergonomics 37 (6): 531–543.
Nagamachi, M. 1995. “Kansei Engineering: A New Ergonomic Consumer-Oriented Technology for Product Development.”Interna-
tional Journal of Industrial Ergonomics 15 (1): 3–11.
International Journal of Production Research 2485
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

Nagamachi, M. 2008. “Perspectives and the New Trend of Kansei/Affective Engineering.”The TQM Journal 20 (4): 290–298.
Noble, C. H., and M. Kumar. 2008. “Using Product Design Strategically to Create Deeper Consumer Connections.”Business
Horizons 51 (5): 441–450.
Orsborn, S., J. Cagan, and P. Boatwright. 2009. “Quantifying Aesthetic Form Preference in a Utility Function.”Journal of Mechani-
cal Design 131 (6): 061001-1–061001-10.
Park, J., and S. H. Han. 2004. “A Fuzzy Rule-Based Approach to Modeling Affective User Satisfaction towards Office Chair
Design.”International Journal of Industrial Ergonomics 34 (1): 31–47.
Schütte, S. 2002. “Design Feelings into Products: Integrating Kansei Engineering Methodology in Product Development.”PhD diss.,
Linkopings universitet.
Sekkeli, G., G. Koksal, I. Batmaz, and O. T. Bayrak. 2010. “Classification Models Based on Tanaka’s Fuzzy Linear Regression
Approach: The Case of Customer Satisfaction Modeling.”Journal of Intelligent & Fuzzy Systems 21 (5): 341–351.
Sener, Z., and E. E. Karsak. 2010. “A Fuzzy Regression and Optimization Approach for Setting Target Levels in Software Quality
Function Deployment.”Software Quality Journal 18 (3): 323–339.
Sener, Z., and E. E. Karsak. 2011. “A Combined Fuzzy Linear Regression and Fuzzy Multiple Objective Programming Approach for
Setting Target Levels in Quality Function Deployment.”Expert Systems with Applications 38 (4): 3015–3022.
Shi, L., S. Olafsson, and Q. Chen. 2001. “An Optimization Framework for Product Design.”Management Science 47 (12):
1681–1692.
Sun, N., X. Mei, and Y. Zhang. 2009. “A Simplified Systematic Method of Acquiring Design Specifications from Customer Require-
ments.”Journal of Computing and Information Science in Engineering 9 (3): 031004-1–031004-10.
Vonderembse, M. A., and T. S. Raghunathan. 1997. “Quality Function Deployment’s Impact on Product Development.”International
Journal of Quality Science 2 (4): 253–271.
Wasserman, G. S. 1993. “On How to Prioritize Design Requirements During the QFD Planning Process.”IIE Transactions 25 (3):
59–65.
Withanage, C., T. Park, and H. J. Choi. 2010. “A Concept Evaluation Method for Strategic Product Design with Concurrent Consider-
ation of Future Customer Requirements.”Concurrent Engineering 18 (4): 275–289.
Withanage, C., T. Park, T. T. H. Duc, and H. J. Choi. 2012. “Dynamic Partial Least Square Path Modeling for the Front-End Product
Design and Development.”Journal of Mechanical Design 134 (10): 100907-1–100907-14.
Yang, C. C., and M. D. Shieh. 2010. “A Support Vector Regression Based Prediction Model of Affective Responses for Product Form
Design.”Computers & Industrial Engineering 59 (4): 682–689.
Zhai, L. Y., L. P. Khoo, and Z. W. Zhong. 2009. “A Rough Set Based Decision Support Approach to Improving Consumer Affective
Satisfaction in Product Design.”International Journal of Industrial Ergonomics 39 (2): 295–302.
Zhang, X. P., J. Bode, and S. J. Ren. 1996. “Neural Networks in Quality Function Deployment.”Computer & Industrial Engineering
31 (3–4): 669–673.
Zhou, M. 1998. “Fuzzy Logic and Optimization Models for Implementing QFD.”Computers & Industrial Engineering 35 (1–2):
237–240.
2486 H. Jiang et al.
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

Appendix 1
Calculate the objective
functions of the new population
Initialization
Calculate the objective functions of
each chromosome
Selection
Rank the population according to
non-dominating criteria
Combine parent population and
offspring population
Non-dominating ranking on the
combined population
Calculate crowding distance of
all the solutions
Get the N member from the
combined population on the
basis of rank and crowding
distance
Replace parent population by
the better members of the
combined population
Final generation?
Pareto optimal
solutions
No
Yes
Obtain the optimal
settings of
engineering
requirements and
design attributes of
a new product
Generation =1
Crossover
Mutation
Appendix 2. Adaptive chaos search
COA is a stochastic search algorithm in which chaos is introduced into the optimisation strategy to accelerate the optimum seeking
operation and find the global optimal solution (Cong, Li, and Feng 2010). COA employs chaotic dynamics to solve optimisation
problems. In the optimisation process, the carrier wave method is used in linearly mapping the chosen chaos variables to the space of
optimisation variables, and then the optimal solutions are searched based on the ergodicity of the chaos variables. The logistic model
used in chaos optimisation is shown in (A1), and the logistic mapping can generate chaos variables by iteration.
cnþ1¼fðcnÞ¼lcnð1cnÞ(A1)
where μ= 4 and cn20;1½is the nth iteration value of the chaos variable c.
COA is incorporated into the process of NSGA-II for two purposes: initialization and refining the search range to conduct narrow
searching. The initialization of population P
1
is conducted by COA. Chaos variables c
ij
,1iN;jV, are generated using (A1),
where Nis the population size and Vis the number of design variables. The individuals x
ij
in P
1
are initialized by mapping the gener-
ated chaos variables into the ranges of the design variables using (A2).
International Journal of Production Research 2487
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015

xij ¼xjmin þðxjmax xjminÞcij (A2)
where xjmin and xjmax:are the minimum and maximum values of the jth design variable, respectively.
To avoid converging prematurely in the process of NSGA-II, COA is introduced to facilitate the algorithm to further search in
the new ranges which are adjusted adaptively based on the current population. The range of the design variables jis adjusted to be
½x0
jmin;x0
jmaxas follows:
x0
jmin ¼xij
^cðxjmax xjminÞ
x0
jmax ¼xij
^þcðxjmax xjminÞ
((A3)
where xij
^is the best individual which has the largest crowding distance in the first front F
1
of the current population, and γis the con-
striction factor that is usually set as 0.5.
In order to avoid values of x0
jmin and x0
jmax exceeding the range ½xjmin;xjmax , the following examination is performed.
If x0
jmin\xjmin , then x0
jmin ¼xjmin;ifx0
jmax [xjmax, then x0
jmax ¼xjmax
The chaos variables c
ij
generated based on (A1) are mapped into the new ranges as follows:
x0
ij ¼x0
jmin þðx0
jmax x0
jminÞcij (A4)
The new individuals are obtained by linearly combining the generated x0
ij and the individuals x
ij
in the current population as follows:
x00
ij ¼ð1lÞx0
ij þlxij (A5)
where μis the adaptive adjustment parameter and can be calculated by Equation (A6).
l¼1k1
k

M
(A6)
where kis the iteration number of COA and Mis the number of objective functions.
Algorithms of chaos-based NSGA-II
The algorithms of the proposed chaos-based NSGA-II are as follows:
Step 1: The initialization of the parameters is first conducted, including the population size N, the number of generations, the dis-
tribution index for crossover η
c
, the crossover probability, the distribution index for mutation η
m
, the mutation probability,
the size of the mating pool, the tournament size and the iteration number of COA. The description of objective functions
is given, such as the number of objective functions M, the number of design variables V, and the range of the design vari-
ables j,½xjmin;xjmax ,1≤j≤V.
Step 2: A parent population P
1
is initialized based on COA using (A1) and (A2). The values of the objective functions of each
individual are calculated. The parent population P
1
is then sorted based on non-domination, and the crowding distances
are calculated for each individual. The generation is set as t=1.
Step 3: The crowded tournament selection is applied to create a mating population. In the selection process, two individuals from
the parent population P
t
are selected at random for a tournament. The winners chosen are inserted in the mating pool for
reproduction and the selection is repeated until the size of the mating pool reaches the predefined value.
Step 4: The SBX operator and the polynomial mutation are conducted in the mating pool to create the offspring population Q
t
.
Step 5: A combined population R
t
is generated by combining the parent population P
t
and the offspring population Q
t
,
R
t
=P
t
∪Q
t
.
Step 6: The combined population R
t
is sorted based on non-domination, and different fronts F
i
,i¼1;, are identified. The new
population P
t+1
with size Nis obtained based on the process of combination and selection.
Step 7: In the new population P
t+1
, the number of individuals that belong to the first front F
1
is calculated. If the number is less
than the population size N, it indicates the current population is not the optimal population, and the COA is then exe-
cuted. The population in COA is initialized by P1
C¼Ptþ1, and the iteration starts from k= 1. The first 10% of individu-
als in Pk
Care updated by the new individuals. The new ranges are defined using (A3) and the new individuals are
generated using (A4), (A5) and (A6). The updated Pk
Cis sorted based on non-domination, then Steps 3–6 are performed
and the new population Pkþ1
Cis obtained. The iteration continues by k+1→ktill the predefined iteration number of
COA is reached. The population P
t+1
is updated by the final population generated from COA.
Step 8: The generation counter is increased by t+1→t. The algorithm is again executed from Step 3 and it stops after the num-
ber of generations reaches the predefined value. Finally, the Pareto front is obtained and the individuals belonging to the
Pareto front are the optimal settings of the design attributes and engineering requirements.
2488 H. Jiang et al.
Downloaded by [Hong Kong Polytechnic University] at 22:57 03 March 2015