European Journal of Operational Research 171 (2006) 390–411

www.elsevier.com/locate/ejor

O.R. Applications

A fuzzy optimization model for QFD planning process

using analytic network approach
Cengiz Kahraman a, Tijen Ertay b,*
, Gülçin Büyüközkan c

a
Department of Industrial Engineering, Istanbul Technical University, Macka 34367, Istanbul, Turkey
b
Department of Management Engineering, Istanbul Technical University, Macka 34367, Istanbul, Turkey
c
Department of Industrial Engineering, Galatasaray University, Ortaköy 34357 Istanbul, Turkey

Received 16 September 2003; accepted 2 September 2004

Available online 30 October 2004

Abstract

In both the quality improvement and the design of a product, the engineering characteristics affecting product per-
formance are primarily identified and improved to optimize customer needs (CNs). Especially, the limited resources and
increased market competition and product complexity require a customer-driven quality management and product
development system achieving higher customer satisfaction. Quality function deployment (QFD) is used as a powerful
tool for improving product design and quality, and procuring a customer-driven quality system. In this paper, an inte-
grated framework based on fuzzy-QFD and a fuzzy optimization model is proposed to determine the product technical
requirements (PTRs) to be considered in designing a product. The coefficients of the objective function are obtained
from a fuzzy analytic network process (ANP) approach. Fuzzy analytic hierarchy process (AHP) is also used in the
proposed framework. An application in a Turkish Company producing PVC window and door systems is presented
to illustrate the proposed framework.

2004 Elsevier B.V. All rights reserved.

Keywords: Fuzzy sets; QFD; AHP; ANP; Optimization

1. Introduction

Today, business decisions in many companies involve selecting the products providing a high degree of
customer satisfaction to meet multiple objectives. Especially, global competitiveness has recently become
the largest concern of many companies, which consider ‘‘continuous improvement’’ to catch up with the

*
Corresponding author. Tel.: +90 212 293 1300x2059; fax: +90 212 240 7260.
E-mail address: ertay@itu.edu.tr (T. Ertay).

0377-2217/$ - see front matter
2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejor.2004.09.016
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 391

rapid changing development throughout the world stimulated by technological innovations and changing
customersÕ needs (CNs). In general, a poor product definition commonly leads to either failure of that prod-
uct in market place or extended product development time. Understanding of CNs leads to successful prod-
ucts and shorter development time. For this reason, product technical requirements (PTRs) or design
requirements are generally established by the product design team at an early design stage according to
the companyÕs strategic goals. Quality function deployment (QFD) is also a cross-functional planning tool
used to help the product development team. The purpose of the technique is to reduce two types of conflict:
First, the conflict that the product specification does not comply with the needs of the predetermined target
group of customers; second, the conflict that the final product does not comply with the PTRs. To reduce
the first conflict, the product specifications, or PTRs must take CNs that are the voice of the customers into
account. The second conflict is reduced by additional transformations of PTRs into product parts, ingre-
dients, processes, and production specifications.
There have been some studies on quantifying the planning subjects in HoQ within past decade. Some of
these studies focused on being employed fuzzy set theory for prioritizing PTRs or CNs (Chan et al., 1999;
Khoo and Ho, 1996; Zhou, 1998; Kim et al., 2000; Shen et al., 2001). Other studies are related to AHP
(Saaty, 1980) applied to the HoQ to identify CNs and to generate the relative importances (Armacost
et al., 1994; Park and Kim, 1998; Doukas et al., 1995; Fukuda and Matsuura, 1993).
In this study, we propose the use of the analytic network process (ANP) (Saaty, 1996) to incorporate the
innerdependence issues into CNs and PTRs in HoQ. ANP enables us to take into consideration the degree
of interdependences between CNs and PTRs by means of AHP. Using crisp ANP in QFD has been con-
sidered by Partovi (2001), Partovi and Corredoira (2002), and Karsak et al. (2002). We extend these studies
by employing linguistic parameters to emphasize impreciseness and vagueness in ANP because of human
judgmentsÕ subjectivity on the importance of PTRs related to CNs. Besides, in order to determine the set
of PTRs, which will be considered in product design, we construct a mixed integer linear programming
model to optimize target improvements. The proposed integrated framework can take the difficulties into
consideration due to the uncertainty of data and lack of quantitative tools. It prioritizes engineering char-
acteristics through a fuzzy ranking procedure and optimizes the improvements using a mixed integer
program.
The paper is organized as the follows. In Section 2, we describe QFD briefly and its literature review.
Section 3 presents the ANP and its usage in QFD. In Section 4, we indicate the representation of ANP
structure in QFD model. Section 5 summarizes fuzzy QFD and proposed optimization framework. In Sec-
tion 6, we illustrate an application of the proposed fuzzy framework.

2. QFD and literature review

Quality function deployment (QFD) is namely a key tool for application of concurrent engineering and
implementing total quality management (TQM) (Guinta and Praizler, 1993). QFD emphasizes multifunc-
tional teams required for integrating all corporate functions to be responsive to the customerÕs requirements
so that product planning, product design, process planning, and production planning provide a coherent
response to CNs. In other words, QFD can be seen as a set of planning tools, which help introducing
new or improved products faster to market by focusing on customer satisfaction.
QFD was originally developed and implemented in Japan at the Kobe Shipyards of Mitsubishi Heavy
Industries in 1972. It was observed that Toyota was able to reduce start up pre-production costs by 60%
from 1977 to 1984 and to decrease the time required for its development by one-third through the use
of QFD (Hauser and Clausing, 1988; Ertay, 1998; Hsiao, 2002). Early users of QFD include Toyota, Ford
Motor Company, Procter, 3M Corporation, Gamble, AT&T, Hewlett Packard, Digital Equipment Corpo-
ration, etc. (Cohen, 1995). Besides, the American Supplier Institute (ASI) in Dearborn, Michigan and
392 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

GOAL/QPC (Growth Opportunity Alliance of Lawrence/Quality Productivity Center) in Methuen, Mas-

sachusetts have been the primary organizations offering an overview and workshop type training since
QFD was introduced to the United States in the early 1980s (Prasad, 1998).
The basic concept of QFD is to translate the needs of customers (CNs), in other words voice of customer
(VoC), into product technical requirements (PTRs) or engineering characteristics, and subsequently into
parts characteristics, process plans, and production requirements related to its manufacture. Each transla-
tion uses a chart, called ‘‘house of quality’’ (HoQ). This cart contains information on ‘‘what to do related to
CNs’’, ‘‘how CNs to do related to PTRs’’, relationships between CNs and PTRs and among the PTRs, and
benchmarking data. The components of HoQ are displayed in Fig. 1 (Hauser and Clausing, 1988).
A more comprehensive way of implementing QFD is to construct many different houses. After establish-
ing the first HoQ, the design specifications are translated to the second HoQ as inputs for the development
of part characteristics. Thus, in the second HoQ matrix, design specifications are linked to product compo-
nent characteristics deployment. In the third HoQ, the product component characteristics are similarly
linked to manufacturing processes. In the final phase, the fourth HoQ is established for setting process
parameters and control limits that part standards and customer needs are met appropriately. The four
HoQs are shown in Fig. 2.
In general, there are two unique aspects for first HoQ of the QFD methodology. First of them is related
to determine relationship between customer needs and technical specifications. The other is related to deter-
mine the correlation between CNs and the correlation between PTRs. For this reason, it can be said that the
HoQ is a kind of conceptual maps which are the means for interfunctional planning and communication
(Shen et al., 2001).
QFD as a complete planning tool in translating the VoC to product and process specifications as well as
to operating decisions has been utilized in a few studies (Ansari and Modarress, 1994; Crowe and Cheng,
1996; Partovi and Corredoira, 2002; Partovi and Epperly, 1999). But, HoQ as a conceptual map of QFD

Technical
Correlation
Roof

The Voice of The

Company

The Voice of Relationship Room Strategic

The Customer Planning Room

Technical Priorities
Room

Fig. 1. The components of HoQ.

 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 393

Engineering Parts Key Process Production

Characteristics Characteristics Operations Requirements

Engineering Characteristics

Key Process Operations

Parts Characteristics
Customer Attributes

I. II. III. IV.

House of Parts Deployment Process Planning Production

Quality Planning

Fig. 2. Deploying the voice of customers from design through production.

has been used extensively in manufacturing areas in recent years (Pulat, 1994; Khoo and Ho, 1996; Kar-
markar and Pitbladdo, 1997) and applied to diversified manufacturing areas, such as the design of a secure
music-toy for children (Hsiao, 2002), the ketchup quality improvement in the food industry (Costa et al.,
2001), a hypothetical application in the lodging industry considering both service management issues and
service innovations based on the framework of QFD (Jeong and Oh, 1998), a new product development
process structured according to a QFD technique in a Danish butter cookie (Holmen and Kristensen,
1998) an illustrative QFD application for evaluation of engine filters for heavy duty trucks (Zhang
et al., 1999), QFD approach based on mathematical programming model for determining the optimal set-
tings for engineering characteristics related to value functions at pencil design (Askın and Dawson, 2000), a
QFD application to cope with the obligation for delivering a quick release top nozzle to several customers
(Crow, 1999), the use of QFD in the area of ergonomics in the case study to translate the needs of safety
shoe users in the cold climate into product characteristics (Bergquist and Abeysekara, 1996).

3. ANP and its usage in QFD

The ANP procedure generalizes the AHP as a widely used multi criteria decision-making tool by replacing
hierarchies with network. More recently, a more general form of AHP approach, which incorporates feed-
back and interdependent relationships among decision attributes and alternatives, has been proposed as a
more accurate approach for modeling complex decision environments. While AHP is a well-known tech-
nique that decomposes a problem into several levels in such a way that they form a hierarchy, ANP enables
interrelationships among the decision levels and attributes to be taken into consideration in a more general
form. Thus, the ANP can be used as an effective tool in those cases where the interactions among the elements
of a system form a network structure (Saaty, 1996). ANP uses ratio scale measurements based on pairwise
comparisons. However, it does not impose a strict hierarchical structure as in AHP, and models a decision
problem using a system with feedback approach. The systems-with-feedback are related to imply how to
include both inner and outer dependencies with feedback. While outer dependence implies the dependence
among components in a way to allow for the feedback circuits, inner dependence is related to the dependence
within a component combined with feedback among components. The systems-with-feedback can be illus-
trated graphically in a way represented by the hierarchical structure. In the graphically representation, in gen-
eral, there are two-way arcs among levels. First of them, a looped arc, is related to show the innerdependency
394 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

Goal

……………
Criteria

…………….
Alternatives

Fig. 3. Linear hierarchy with no feedback and innerdependence (Saaty and Takizawa, 1986).

relationships that occur within the same level of analysis. Second of them, a hierarchical arc is also related to
show a dominance or control of one level of attributes over another set of sub-components or attributes. A
linear hierarchy with no feedback and innerdependency representation is given in Fig. 3.
This representation is also suitable for a general network of QFD model. After constituting the construc-
tion of the network, the calculation of the priorities of elements is required. In order to construct the struc-
ture of the problem, both the inneractions and the interactions among the elements should be considered.
All of these relations evaluated by pairwise comparisons and a supermatrix, which is a matrix of influence
among the elements, are obtained by these priority vectors. The supermatrix is raised to limiting powers to
calculate the overall priorities, and thus the cumulative influence of each element on every other element
with which it interacts is obtained (Saaty and Vargas, 1998). When a network consists of only two clusters
apart from the goal, namely criteria and alternatives, the matrix manipulation approach proposed by Saaty
and Takizawa (1986) can be employed to deal with the dependencies of the elements of a system. This
approach will be considered to incorporate the dependencies inherent in QFD network process. The super-
matrix representation of QFD network process constitutes from three levels as follows:

ð1Þ

where w21 is a vector that represents the impact of the goal on the criteria, W32 is a matrix that represents
the impact of the criteria on each of the alternatives, and I is the identity matrix. There are also some re-
searches using AHP to determine the degrees of importance of the customer needs (Armacost et al., 1994;
Lu et al., 1994; Park and Kim, 1998). More recently some studies are related to the application of ANP:
Hämäläinen and Seppäläinen (1986) describe the application of the so-called supermatrix technique of
the ANP to a complex energy decision problem. Their model describes a nuclear power plant-licensing
problem in Finland. Meade and Sarkis (1998) use the ANP to evaluate logistics strategies for an organiza-
tion that seeks to be adaptive to dynamic competitive environments. Lee and Kim (2000, 2001) suggest an
improved IS project selection methodology which reflects interdependencies among evaluation criteria and
candidate projects using ANP within a zero-one goal programming model. Partovi and Corredoira (2002)
present a QFD model based on ANP for prioritizing and designing rule changes for the game of soccer in
order to make it more attractive to soccer enthusiasts. Partovi (2001) presents an analytical method for
quantifying HeskettÕs ‘‘Strategic service vision’’. In the model, AHP is used to determine the intensity of
the relationship between the row and column variables of each matrix, while ANP is used to determine
the intensity of synergy effects among column variables. More recently, Karsak et al. (2002) realize the
product planning in QFD using a combined ANP and goal programming approach.
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 395

4. Representation of ANP structure in QFD model

The network representation in QFD model is based on the structure of a hierarchy with inner depend-
encies within components and no feedback. In this situation, the CNs correspond to the alternatives, which
have inner dependencies within themselves. The first step of the network representation in QFD model is
the identification of the CNs and PTRs. Then, the determination of the importance of the CNs, which cor-
responds to the first step of the matrix manipulation concept of the ANP, follows (Lee and Kim, 2000;
Saaty and Takizawa, 1986). Next, the body of the house by the weights obtained will be filled through com-
paring the PTRs with respect to each CN. Finally, the interdependent priorities of the PTRs by analyzing
dependencies among the CNs and PTRs will be obtained. The general network representation of QFD
model based on Fig. 3 is given in Fig. 4. W3 and W4 are the inner dependency matrices of the CNs and
PTRs respectively. W2 is a outer dependency matrix including the column eigenvectors with respect to each
CN.
Based on the representation of Fig. 4, the supermatrix representation of the QFD model used in this
study is as follows:

ð2Þ

where w1 is a vector on the CNs that represents the impact of the goal, namely manufacturing a product
that satisfies the customer. W2 is a matrix that denotes the impact of the CNs on each of the PTRs. W3
and W4 are the matrices that represent the inner dependence of the CNs and the inner dependence of
the PTRs, respectively.

5. Fuzzy QFD

Zadeh (1965) introduced the fuzzy set theory to deal with the uncertainty due to imprecision and vague-
ness. A major contribution of fuzzy set theory is its capability of representing vague data. The theory also
allows mathematical operators and programming to apply to the fuzzy domain. A fuzzy set is a class of
objects with a continuum of grades of membership. Such a set is characterized by a membership function,
which assigns to each object a grade of membership ranging between zero and one. Fig. 5 demonstrates a

Goal

(w1)

Customer
(W3) (Inner dependence)
Criteria needs

(W2) (Outer dependence)

Design
Alternatives requirements (W4) (Inner dependence)

Fig. 4. The network representation of the HoQ model.

396 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

0
Fig. 5. Left and right representation of a TFN, Pe .

triangular fuzzy number (TFN) Pe . Fuzzy set theory is a perfect means for modeling uncertainty (or impre-
cision) arising from mental phenomena, which are neither random nor stochastic. Human beings are heav-
ily involved in the process of decision analysis. A rational approach toward decision-making should take
into account human subjectivity, rather than employing only objective probability measures. This attitude
towards the uncertainty of human behavior led to the study of a relatively new decision analysis field: Fuzzy
decision-making.
Numerical data obtained across a human subjectivity are called fuzzy data. The motivation for the use of
words or sentences rather than numbers is that linguistic characterizations are, in general, less specific than
numerical ones (Zadeh, 1973). Linguistic variables differ from numerical variables in that their values are
not numbers but are words or phrases. The use of linguistic variables allows precise modeling of imprecise
statements such as ‘‘very important’’ or ‘‘some important’’. The successful use of linguistic variables is
highly dependent on the determination of a valid membership function. When there is both a normal
and a convex fuzzy set with membership functions that satisfy both normality and convexity, arithmetic
operations can be performed to obtain fuzzy numbers. A fuzzy number is a fuzzy set of the real line with
a normal, (fuzzy) convex and continuous membership function of bounded support.
There are two main characteristics of fuzzy systems that give them better performance for specific appli-
cations. Fuzzy systems are suitable for uncertain or approximate reasoning, especially for the system with a
mathematical model that is difficult to derive. Fuzzy logic allows decision-making with estimated values
under incomplete or uncertain information.
In traditional QFD, most of the input variables are assumed to be precise and are treated as numerical
data. However, QFD as a concept and mechanism for translating the voice of the customer into product
attributes through various stages of product planning, engineering, processing, and production is required
linguistic data to be inherently vague and ambiguity. Linguistic data can be treated to approximate exact-
ness with the help of fuzzy set theory. When implementing QFD using linguistic data, some factors may
affect the final results such a ranking of technical characteristics. The factors include the type of fuzzy num-
bers, defuzzification strategies, and the degree of fuzziness of fuzzy numbers. Besides, capturing the elastic-
ity of imprecise requirements is an important issue. CustomersÕ preferences are often fuzzy and imprecise.
In addition, relationships between CNs and PTRs are identified qualitatively. This qualitative identification
requires to be translated into numerical scales. For example, the human is asked to denote if a relationship
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 397

is weak, moderate, or strong and their answer is translated to a scale like 1–3–5; 1–3–9; or 1–5–9. The
important problem in traditional QFD is that the choice of scales can dramatically influence the outcome.
For this reason, it can be more appropriate to treat linguistic data as fuzzy rather than precise.
As looking out for the related literature, Masud and Dean (1993) investigate how QFD analysis can be
performed when input variables are treated as linguistic variables with values expressed as fuzzy numbers.
Bahrami (1994) introduces a method for performing routine design by using information content and fuzzy
QFD based on the concept of linguistic variable. Kim et al. (1994, 2000) present an integrated approach
that allows a design team to mathematically consider tradeoffs among various customer attributes as well
as the inherent fuzziness in the system by combining multi-attribute value theory with fuzzy linear regres-
sion and fuzzy optimization theory. Khoo and Ho (1996) develop an approach centered on applying pos-
sibility theory and fuzzy arithmetic to address the ambiguity involved in various relationships. Fung et al.
(1998) propose a hybrid system that incorporates the principles of QFD, analytic hierarchy process, and
fuzzy set theory to tackle the complex and often imprecise problem domain encountered in customer
requirement management. Zhou (1998) suggests an approach to address the difficulties due to the uncer-
tainty of data and lack of quantitative tools. This approach prioritizes engineering characteristics through
a fuzzy ranking procedure and optimizes the improvements using a mixed integer program. Wang (1999)
considers the QFD planning as a multi-criteria decision problem and proposes a new fuzzy outranking ap-
proach to prioritize design requirements recognized in QFD. The inputs required for QFD are represented
with linguistic terms that are characterized by fuzzy sets. The fuzzy outranking relation is used to model the
imprecise preference relations between design requirements. Shen et al. (2001) examine the sensitivity of the
ranking of technical characteristics to the defuzzification strategy and the degree of fuzziness of fuzzy num-
bers. This proposed fuzzy approach allows QFD users to avoid subjective and arbitrary quantification of
linguistic data. Sohn and Choi (2001) develop a fuzzy quality function deployment (QFD) model in order
to convey fuzzy relationship between customers needs and design specification for reliability in the context
of supply chain management. More recently, Yang et al. (2003) presented the findings of a research effort to
adapt HoQ to meet the needs of buildable designs in the construction industry and to develop a fuzzy QFD
system for buildability evaluation. In this system, the fuzzy set theory is integrated into HoQ to capture the
inherent impreciseness and vagueness of design-relevant inputs and facilitate the analysis of design-relevant
QFD information. Erol and Ferrell (2003) present a methodology to assist decision-makers in selecting
from a finite number of alternatives when there are more than one objective and both qualitative and quan-
titative factors must be considered. The methodology uses fuzzy QFD to convert qualitative information
into quantitative parameters and then combines this data with other quantitative data to parameterize a
multi-objective mathematical programming model.
In this study, first of all, fuzzy ANP model is used for the prioritizing of PTRs in QFD. The steps of this
model are given in Table 1. In Section 5.2, the results of the model are used for the estimation of objective
functionÕs coefficients in an optimization model based on mixed-integer programming.

5.1. ChangÕs extent analysis method

As we mentioned in the previous section, to calculate w1, W2, W3, W4, and wAG, it is necessary to use a
fuzzy AHP methodology. There are various types of fuzzy AHP methods in the literature (such as Laarh-
oven and Pedrycz, 1983; Buckley, 1985; Chang, 1996; Leung and Cao, 2000). In this study, we prefer
ChangÕs (1992, 1996) extent analysis method since the steps of this approach are relatively easier than
the other fuzzy AHP approaches and similar to the crisp AHP. Recently, Bozdag et al. (2003) have used
this approach in the evaluation of computer integrated manufacturing alternatives. Using the same ap-
proach, Kahraman et al. (2003, 2004) also used this approach in the evaluation of the catering firms in Tur-
key and in the selection of the best location for a facility respectively. Büyüközkan et al. (2004) applied the
same approach to select the best software development strategy. Kwong and Bai (2003) used it to prioritize
398 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

Table 1
The evaluation algorithm steps for determining the overall priorities of the PTRs
Step 1. Identifying CNs and determining the PTRs matching the CNs
Step 2. Determining the importance degrees of CNs with linguistic data by assuming that there is no dependence among the CNs:
Calculation of w1
Step 3. Determining the importance degrees of PTRs with respect to each CN with linguistic data by assuming that there is no
dependence among the PTRs: Calculation of W2
Step 4. Determining the inner dependency matrix of the CNs with respect to each CN with linguistic data by utilizing the schematic
representation of inner dependence among CNs: Calculation of W3
Step 5. Determining the inner dependency matrix of the PTRs with respect to each PTR with linguistic data by utilizing the schematic
representation of inner dependence among PTRs: Calculation of W4
Step 6. Determining the interdependent priorities of the CNs: Calculation of wC = W3 · w1
Step 7. Determining the interdependent priorities of the PTRs: Calculation of WA = W4 · W2
Step 8. Determining the overall priorities of the PTRs: Calculation of wANP = WA · wC

customer requirements in QFD. The fuzzy scale regarding relative importance to measure the relative
weights is given in Fig. 6 and Table 2. This scale will be used in ChangÕs fuzzy AHP model.
The steps of ChangÕs (1992, 1996) extent analysis approach are as follows: Let X = {x1, x2, . . ., xn} be an
object set, and G = {g1, g2, . . ., gm} be a goal set. According to the method of ChangÕs (1992) extent analysis,
each object is taken and extent analysis for each goal, gi, is performed respectively. Therefore, m extent
analysis values for each object can be obtained, with the following signs:
M 1gi ; M 2gi ; . . . ; M mgi ; i ¼ 1; 2; . . . ; n; ð3Þ
where all the M jgi (j = 1, 2, . . ., m) are triangular fuzzy numbers (TFNs).

µ RI

EI WMI SMI VSMI AMI

1.0

1/2 1 3/2 2 5/2 3 7/2 RI

Fig. 6. Linguistic scale for relative importance.

Table 2
Linguistic scales for difficulty and importance
Linguistic scale for difficulty Linguistic scale for importance Triangular fuzzy scale Triangular fuzzy
reciprocal scale
Just equal Just equal (1, 1, 1) (1, 1, 1)
Equally difficult (ED) Equally important (EI) (1/2, 1, 3/2) (2/3, 1, 2)
Weakly more difficult (WMD) Weakly more important (WMI) (1, 3/2, 2) (1/2, 2/3, 1)
Strongly more difficult (SMD) Strongly more important (SMI) (3/2, 2, 5/2) (2/5, 1/2, 2/3)
Very strongly more difficult (VSMD) Very strongly more important (VSMI) (2, 5/2, 3) (1/3, 2/5, 1/2)
Absolutely more difficult (AMD) Absolutely more important (AMI) (5/2, 3, 7/2) (2/7, 1/3, 2/5)
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 399

The steps of ChangÕs extent analysis can be given as follows:

Step 1. The value of fuzzy synthetic extent with respect to the ith object is defined as
" #1
Xm n X
X m
j j
Si ¼ M gi
M gi : ð4Þ
j¼1 j¼1 j¼1
Pm j
To obtain j¼i M gi , perform the fuzzy addition operation of m extent analysis values for a particular matrix
such that
!
X
m X
m X
m X
m
M jgi ¼ lj ; mj ; uj ð5Þ
j¼i j¼1 j¼1 j¼1
hP P i1
n m j
and to obtain i¼1 j¼1 M gi , perform the fuzzy addition operation of M jgi (j = 1, 2, . . ., m) values such
that
!
X
n X
m X
n X
n X
n
M jgi ¼ li ; mi ; ui ð6Þ
i¼1 j¼1 i¼1 i¼1 i¼1

and then compute the inverse of the vector in Eq. (6) such that
" #1

Xn X m
1 1 1
j
M gi ¼ Pn ; Pn ; Pn : ð7Þ
i¼1 j¼1 i¼1 ui i¼1 mi i¼1 li

Step 2. The degree of possibility of M2 = (l2, m2, u2) P M1 = (l1, m1, u1) is defined as
 
V ðM 2 P M 1 Þ ¼ sup minðlM 1 ðxÞ; lM 2 ðyÞÞ ð8Þ
yPx

and can be equivalently expressed as follows:

8
< 1;
> if m2 P m1 ;
V ðM 2 P M 1 Þ ¼ hgtðM 1 \ M 2 Þ¼lM 2 ðdÞ ¼ 0; if l1 P u2 ; ð9Þ
>
: l1 u2
ðm2 u2 Þðm1 l1 Þ
; otherwise;

where d is the ordinate of the highest intersection point D between lM1 and lM2 (see Fig. 7).
To compare M1 and M2, we need both the values of V(M1 P M2) and V(M2 P M1).
Step 3. The degree possibility for a convex fuzzy number to be greater than k convex fuzzy numbers Mi
(i = 1, 2, . . ., k) can be defined by
V ðM P M 1 ; M 2 ; . . . ; M k Þ ¼ V ½ðM P M 1 Þ and ðM P M 2 Þ and    and ðM P M k Þ
¼ min V ðM P M i Þ; i ¼ 1; 2; 3; . . . ; k: ð10Þ
Assume that
d 0 ðAi Þ ¼ min V ðS i P S k Þ ð11Þ
for k = 1, 2, . . ., n; k 5 i. Then the weight vector is given by
T
W 0 ¼ ðd 0 ðA1 Þ; d 0 ðA02 Þ; . . . ; d 0 ðAn ÞÞ ð12Þ
where Ai (i = 1, 2, . . ., n) are n elements.
400 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

µ M~ ~ ~
M1 M2

V (M2 ≥ M 1 )

M
0
l1 m1 l2 d u1 m2 u2

e 1 and M
Fig. 7. The intersection between M e 2.

Step 4. Via normalization, the normalized weight vectors are

T
W ¼ ðdðA1 Þ; dðA2 Þ; . . . ; dðAn ÞÞ ð13Þ
where W is a nonfuzzy number.

5.2. Fuzzy optimization framework

A mixed integer linear programming model is proposed to optimize target improvements. In this model,
ai is defined as the wANP
i value belonging to each PTR:
ai ¼ wANP
i : ð14Þ
Let xi be the improvement of PTRi then ai * xi represents the utility value contributed from the improve-
ment of PTRi. The total utility value is simply the sum of utility contributions from each member of PTRs.
A clear objective is to identify the best improvements such that the total utility value is maximized and all
constraints satisfied. The considered optimization model is followed in Eq. (15).
X
Maximize ai d 1
i xi ð15Þ
subject to xi 6 zi P i 8i;
X
ðDi zi þ ci xi Þ 6 B;
i
li 6 xi 6 ui 8i;
zi 2 f0; 1g 8i.
In this model, the first constraint represents that no improvement is made to a PTR that is not selected.
Pi is any number that can make ziPi P ui when zi = 1. The second constraint ensures that total cost of
improvements does not exceed the given budget limit. Due to limited resource, a budget constraint is often
necessary to ensure that total cost of improvements must not exceed the given budget. The third constraint
enforces the competition requirement and technological feasibility. In other words, to be competitive, a
company usually sets its targets better or at least not worse than the competitorsÕ. Any improvement must
be subjected to technological feasibility enforced by available resource. In this situation, li indicates mini-
mum improvement to match competitorsÕ performance, in addition, ui indicates maximum obtainable
improvement. To reduce the effects of different magnitudes, the variables in the objective function are scaled
by dividing xi by di where di is the range that restricts the value of PTR. The rest of the notation is intro-
duced as follows: ci = unit cost of improving PTRs; B = Budget limit, li = the lover bound of xi, ui = the
upper bound of xi.
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 401

Fuzzy mathematical programming is defined as the term that is usually used in operation research, i.e. as
an algorithmic approach to solve models of the type

Maximize f ðxÞ ð16Þ

such that gi ðxÞ 6 0:

Here, a special model of the problem ‘‘maximize an objective function subject to constraints’’, namely the
‘‘linear programming model’’ will be considered:

Maximize Z ¼ cT x ð17Þ
such that Ax 6 b;
x 6 0;

where c and x are n-vectors, b is an m-vector, and A is an m · n matrix.

It is assumed that the decision-maker has upper and lower bounds
ci and ^ci for the attainment of the
objectives. The decision-maker can establish these aspiration levels for himself, or they can be computed
as a function of the solution space. The constraints can be hard or soft. If soft, it is assumed that the deci-
sion-maker has upper and lower bounds
ci and ^ci for the, the right hand side b can be exceeded by the
amount p, which is also under the discretion of the decision-maker (Zimmermann, 1994).
The membership function of the fuzzy objective function i, lGi(x), should be 0 for aspiration levels equal
to or less than ^ci , 1 for aspiration levels equal to or greater than
ci , and monotonically increasing from 0 to
1, i.e.,
8
> 0 if cTi x 6 ^ci ;
>
< T
lGi ðxÞ ¼ c
ci x^ ci
if ^ci < cTi x 6
ci ; ð18Þ
>
> i ^
ci
:
1 if cTi x P
ci :

The membership function of the fuzzy set representing constraint j, lCj(x), should be 0 if the constraint is
strongly violated (i.e., if it exceeds bj + pj), 1 if it is satisfied in the crisp sense (i.e., if equal to or less than bj),
and should decrease monotonically from 1 to 0 over the tolerance interval (bj, bj + pj),
8
>
> 1 if ðAxÞj 6 bj ;
>
<
ðAxÞ bj
lCj ðxÞ ¼ 1  pjj if bj < ðAxÞj 6 bj þ pj ; ð19Þ
>
>
>
:0 if ðAxÞj > bj þ pj :

The membership function of the decision set, lD(X), is given by

n o
lD ðX Þ ¼ min lGi ðxÞ; lCj ðxÞ i ¼ 1; . . . ; k; j ¼ 1; . . . ; m; for all x 2 X : ð20Þ
ij

The min-operator is used to model the intersection of the fuzzy sets of objectives and constraints. Since the
decision maker wants to have a crisp decision proposal, the maximizing decision will correspond to the
value of x, xmax, that has the highest degree of membership in the decision set:
n o
lD ðxmax Þ ¼ max min lGi ðxÞ; lCj ðxÞ i ¼ 1; . . . ; k; j ¼ 1; . . . ; m: ð21Þ
xP0 ij
402 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

This problem is equivalent to solving the following crisp LP problem:

Maximize k ð22Þ

T 
c x  ^c
subject to P k; i ¼ 1; 2; . . . ; k;

c  ^c i


Ax  b
1 P k; j ¼ 1; 2; . . . ; m;
p j

x P 0;

which can be rewritten as

Maximize k ð23Þ
subject to kð
c  ^cÞi  cTi x
6 ^ci ; i ¼ 1; 2; . . . ; k;
kpj þ ðAxÞj 6 bj þ pj ; j ¼ 1; 2; . . . ; m;
x P 0:
Now, we can convert the crisp LP problem in Eq. (15) to the fuzzy case:

Maximize k ð24Þ
X
subject to kð
c  ^cÞi  ai d 1
i xi 6 ^ci ; i ¼ 1; 2; . . . ; k;
i

kp1 þ xi  zi P i 6 p1 ; 8i;
X
kp2 þ ðDi zi þ ci xi Þ 6 B þ p2 ; 8i;
i

kp3 þ li  xi 6 p3 ; 8i;
kp4 þ xi  ui 6 p4 ; 8i;
zi 2 f0; 1g; 8i:

6. Application of the proposed fuzzy framework

In this study, the considered decision methodology is applied to Kompen Co., which is a producer of
door and window systems with PVC material. As known, these systems being produced with plastic mate-
rial are more economical than the ones being produced with wooden. Moreover, in present day, sale prices
and preventive maintenance costs against the corruption of wooden-made products are gradually increas-
ing because of high labor costs. Kompen profiles are produced with respect to Turkish Standards (TS)
Norm and the company has possessed TS-ISO 9002 Quality System Certificate since 1998. Since the begin-
ning of the year 1996, Kompen PVC factory uses German and Austrian technologies. Kompen with 250
sellers in Turkey takes 10–12% share in the market, aiming at achieving the percentage of 20%. Kompen
has a distributor network in Romanya, Russia, Belarus, Holland, Ukrayna, Uzbekistan, Mauritius, France,
Kazakhstan, and Turkmenistan.
In this case study, the PVC window system in Kompen Co. is considered for application in QFD.
Longitudinal section of the considered PVC window system is illustrated in Fig. 8.
The data were obtained by interviewing KompenÕs experts. The pairwise comparison matrices among
CNs to obtain w1 and W3 considering the inner dependencies and among PTRs to obtain W4 considering
the inner dependencies and among CNs and PTRs to obtain W2 were used.
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 403

Fig. 8. Example of a PVC window system.

The attributes to improve the PVC window system have been determined as follows: Heat insulating
(HI), sound insulating (SI), waterproofing (WP), Robustness (R), bending (B), and stability of color
(SC). The determined PTRs are as follows: PVC conductivity coefficient (PVCCC), PVC cell thickness
(PVCCT), Internal drainage system and gaskets (IDS-G), Anticorrosion and CaCO3 (A-CACO3), Support
sheets and nuts (SSN-N), and TiO2 pigment (TiO2-P).
Two sample pairwise comparison matrices for sound insulating (SI) in Table 5 and for heat insulating
(HI) in Table 7 are given in the following (Tables 3 and 4). The reciprocal values are not given since they
are the inverses of their reciprocals.
The steps of the proposed fuzzy ANP methodology for HoQ according to Table 1 are described as
follows:

Table 3
Relative importance of the PTRs for sound insulating (SI)
SI PVCCC PVCCT IDS-G SS-N Relative
importance
weights
PVCCC Weakly more important Very strongly more important Very strongly more important 0.58
PVCCT Strongly more important Strongly more important 0.39
IDS-G Equally important 0.00
SS-N 0.03

Table 4
The inner dependence matrix of the CNs with respect to heat insulating (HI)
HI HI SI B Relative importance weights
HI Weakly more important Strongly more important 0.56
SI Weakly more important 0.34
B 0.10
404 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

Steps 1 and 2. Identifying CNs and PTRs and importance degrees of CNs: Calculation of w1.
Assuming that there is no dependence among the CNs, the following eigenvector for the CNs is obtained
by performing the extent analysis of fuzzy AHP methodology with respect to the goal of achieving the best
PVC design.
0 1 0 1
Heat insulating ðHIÞ 0:27
B Sound insulating ðSIÞ C B 0:21 C
B C B C
B C B C
B Water proofing ðWPÞ C B 0:13 C
w1 ¼ BB C B C
Robustness ðRÞ C ¼ B 0:13 C
B C B C
B C B C
@ Bending ðBÞ A @ 0:13 A
Stability of color ðSCÞ 0:13
Step 3. Determining the importance degrees of PTRs with respect to each CN: Calculation of W2.
The relative importance weights with respect to each CN are given in Table 5.
Step 4. Determining the inner dependency matrix of the CNs with respect to each CN: Calculation of W3.
The inner dependencies among the CNs are determined by analyzing the impact of each CN on the other
CNs and using pairwise comparisons. The schematic representation of the relationships among the CNs is
shown in Fig. 9. The inner dependencies among the CNs are depicted in Table 6. An arrow indicates that
the CN in the row affects the CN in the column. A dash indicates no effect between two elements. The Kom-
penÕs experts confirmed that the directions of the relationships among CNs were as in Table 6. The resulting
eigenvectors obtained from pairwise comparisons are presented in Table 7.
Step 5. Determining the inner dependency matrix of the PTRs with respect to each PTR: Calculation of W4.
In this step, we deal with the dependence among the PTRs. As previously accomplished for CNs,
the inner dependencies are determined and required pairwise linguistic comparisons are performed. The

Table 5
The column eigenvectors with respect to each CN
W2 HI SI WP R B SC
PVCCC 0.68 0.58 0 0 0 0
PVCCT 0.32 0.39 0 0.35 0 0
IDS-G 0 0 1 0 0 0
A-CaCO3 0 0 0 0.23 0.32 0
SS-N 0 0.03 0 0.41 0.68 0
TiO2-P 0 0 0 0 0 1

HI SI WP R B SC

Fig. 9. The inner dependence among the customer needs.

 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 405

Table 6
The inner dependency matrix among the customer needs
CNs HI SI WP R B SC
HI – – – –
SI – – – –
WP – – –
R – – – – –
B – – – –
SC – – – – –

Table 7
The inner dependency matrix among CNs
W3 HI SI WP R B SC
HI 0.56 0.43 0 0 0 0
SI 0.34 0.57 0.09 0 0 0
WP 0 0 0.46 0 0 0.07
R 0 0 0 0.93 0.59 0
B 0.1 0 0.46 0 0.41 0
SC 0 0 0 0.07 0 0.93

TiO 2-P A-CACO 3 PVCCC PVCCT SS-N IDS-G

Fig. 10. The inner dependence among the PTRs.

schematic representation of the relationships among the PTRs is shown in Fig. 10. The inner dependen-
cies among the PTRs which the KompenÕs experts confirm are depicted in Table 8. An arrow indicates
that the PTR in the row affects the PTR in the column. A dash indicates no effect between two elements.
The relative importance weights obtained from the pairwise comparisons are presented in Table 9.
Step 6. Determining the interdependent priorities of the CNs: Calculation of wC = W3 · w1.
In this step, we obtain the interdependent priorities of the CNs: wC = W3 · w1
0 1
0:2410
B C
B 0:2229 C
B C
B 0:0689 C
B C
wC ¼ B C
B 0:1971 C
B C
B C
@ 0:1401 A
0:1300

Step 7. Determining the interdependent priorities of the PTRs: Calculation of WA = W4 · W2.

406 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

Table 8
The inner dependency matrix among PTRs
PTRs TiO2-P A-CACO3 PVCCC PVCCT SS-N IDS-G
TiO2-P – – – – – –
A-CACO3 – – – – – –
PVCCC – –
PVCCT – – – –
SS-N – – –
IDS-G – – – –

Table 9
The inner dependency matrix of the PTRs
W4 PVCCC PVCCT IDS A-CaCO3 SS-N TiO2-P
PVCCC 0.48 0.68 0 1 0.28 0
PVCCT 0.29 0.32 0 0 0.4 0
IDS 0.13 0 1 0 0 0
A-CaCO3 0.1 0 0 0 0 0
SS-N 0 0 0 0 0.32 0
TiO2-P 0 0 0 0 0 1

In this step, the interdependent priorities of the PTRs, WA, are calculated as follows: WA = W4 · W2,
0 1
0:5440 0:5520 0:0000 0:5828 0:5104 0:0000
B 0:2996 0:3050 0:0000 0:2810 0:2720 0:0000 C
B C
B C
B 0:0884 0:0754 1:0000 0:0000 0:0000 0:0000 C
WA ¼ B B 0:0680 0:0580 0:0000 0:0000 0:0000
C:
B 0:0000 C
C
B C
@ 0:0000 0:0096 0:0000 0:1362 0:2176 0:0000 A
0:0000 0:0000 0:0000 0:0000 0:0000 1:0000
Step 8. Determining the overall priorities of the PTRs: Calculation of wANP = WA · wC.
The overall priorities of the PTRs, wANP, reflecting the interrelationships within the HoQ, are calculated
by as follows:
0 1 0 1
PVC conductivity coefficient 0:4413
B PVC cell thickness C B 0:2331 C
B C B C
B C B C
B Inner drainage system and gaskets C B 0:1072 C
w ANP
¼ W A  wC ¼ B B C ¼BB C:
Anti-corrosion and CaCO3 C C
B C B 0:0296 C
B C B C
@ Support sheets and nuts A @ 0:0586 A
TiO2 pigment 0:1302
With respect to the results of fuzzy ANP, the most important PTR is PVC conductivity coefficient. And
then PVC cell thickness, TiO2 pigment, inner drainage system and gaskets, support sheets and nut, and
anti-corrosion and CaCO3.
To compare the results of the used methods for the cases of crisp and fuzzy data, we obtained Table 10.
The crisp data for crisp AHP and crisp ANP methods were obtained using the largest possible values of
fuzzy numbers.
As it is clearly seen from Table 10, in all the methods PVC conductivity coefficient has the largest weight
of all. PVC cell thickness has the second importance level in all the methods except crisp AHP. While inner
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 407

Table 10
Comparison of the used methods
PTRs Crisp AHP Crisp ANP Fuzzy AHP Fuzzy ANP
PVCCC 0.3277 0.2204 0.3054 0.4413
PVCCT 0.1266 0.2176 0.2138 0.2331
IDS-G 0.1363 0.1435 0.1310 0.1072
A-CaCO3 0.0572 0.2078 0.0715 0.0296
SS-N 0.2178 0.1840 0.1480 0.0586
TiO2-P 0.1344 0.1612 0.1303 0.1302

drainage system has the fourth importance levels in fuzzy methods, it has the least importance level in crisp
ANP. Anti-corrosion and CaCO3 has the least importance level except crisp ANP method. The other PTRs
have different rankings with respect to the used methods.
In Table 11, the data necessary for the optimization model defined in Section 5.2 are given. The budget
limit is $94,000 but may be expanded to $110,000. Thus, the right hand side B can be exceeded by the
amount p = $16,000.
Using the above data and the optimization model in Section 5.2, the following formulation is obtained:
Maximize 0:40118x1 þ 0:0777x2 þ 0:0002144x3 þ 0:000074x4 þ 0:0293x5 þ 0:0651x6
subject to x1  1:7z1 6 0;
x2  6z2 6 0;
x3  1200z3 6 0;
x4  800z4 6 0;
x5  4z5 6 0;
x6  6z6 6 0;
40;000z1 þ 10; 000z2 þ 8000z3 þ 4000z4 þ 1000z5 þ 3000z6 þ 150x1 þ 50x2
þ 30x3 þ 10x4 þ 5x5 þ 6x6 6 94;000;
0:6 6 x1 6 1:7;
3 6 x2 6 6;
700 6 x3 6 1200;
400 6 x4 6 800;
2 6 x5 6 4;
4 6 x6 6 6;
zi 2 f0; 1g and xi P 0:

Table 11
Data for the case study
PTRs Data
li ui di Di ($) ci ($)
2 2 2
PVCCC 0.6 W/m K 1.7 W/m K 1.1 W/m K 40,000 150
PVCCT 3 mm 6 mm 3 mm 10,000 50
IDS 700 Pa 1200 Pa 500 Pa 8000 30
A-CaCO3 400 kg 800 kg 400 kg 4000 10
SSN-N 2 mm 4 mm 2 mm 1000 5
TiO2-P 4 units 6 units 2 units 3000 6
408 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

Using LINDO, the solution to the above problem is found as z1 = 1, z2 = 1, z3 = 1, z4 = 1, z5 = 1, z6 = 1,

x1 = 1.7, x2 = 6, x3 = 700, x4 = 638.9, x5 = 4, x6 = 6 and the maximum value of the objective function is
1.8534. When we expand the budget limit to $110,000, the budget constraint becomes
40;000z1 þ 10;000z2 þ 8000z3 þ 4000z4 þ 1000z5 þ 3000z6 þ 150x1 þ 50x2 þ 30x3
þ 10x4 þ 5x5 þ 6x6 6 110;000:
The LINDO solution to the expanded problem is z1 = 1, z2 = 1, z3 = 1, z4 = 1, z5 = 1, z6 = 1, x1 = 1.7,
x2 = 6, x3 = 1179.63, x4 = 800, x5 = 4, x6 = 6 and the maximum value of the objective function is 1.9681.
Using the fuzzy optimization model in Eq. (24), we obtain the following formulation:
Maximize k
subject to 0:1147k  ð0:40118x1 þ 0:0777x2 þ 0:0002144x3 þ 0:000074x4
þ 0:0293x5 þ 0:0651x6 Þ 6 1:8534;
x1  1:7z1 6 0;
x2  6z2 6 0;
x3  1200z3 6 0;
x4  800z4 6 0;
x5  4z5 6 0;
x6  6z6 6 0;
16;000k þ 40;000z1 þ 10;000z2 þ 8000z3 þ 4000z4 þ 1000z5
þ 3000z6 þ 150x1 þ 50x2 þ 30x3 þ 10x4 þ 5x5 þ 6x6 6 110;000;
0:6 6 x1 6 1:7;
3 6 x2 6 6;
700 6 x3 6 1200;
400 6 x4 6 800;
2 6 x5 6 4;
4 6 x6 6 6;
zi 2 f0; 1g and xi P 0:
When the above fuzzy optimization problem is solved using LINDO, z1 = 1, z2 = 1, z3 = 1, z4 = 1, z5 = 1,
z6 = 1, x1 = 1.7, x2 = 6, x3 = 912.51, x4 = 800, x5 = 4, x6 = 6 and k = 0.501 are obtained. This is a crisp
solution generated from the fuzzy decision set by selecting the solution that has the highest degree of mem-
bership in the decision set. When compared with the crisp results, the variable values for the PTRs IDS and
A-CaCO3 changed. While in the crisp case IDS is at its lower value, x3 = 700, in the fuzzy case it has a value
of 912.51. This means we can make an improvement for this PTR. In a similar way, while in the crisp case
A-CaCO3 has a value of 638.9, in the fuzzy case it is at its upper value, x4 = 800. This means we can make
the largest possible improvement for this PTR.

7. Conclusions

In this study, an integrated fuzzy ANP approach to formulate and solve a QFD problem has been pre-
sented. The proposed methodology (framework) could be expanded to allow extended linkages among the
houses of quality across the design/production chain. A systematic decision procedure in this study aims at
considering the interdependencies between CNs and PTRs, and the inner dependencies within themselves.
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 409

This paper includes a combined ANP and a fuzzy logic approach to incorporate the CNs and the PTRs
systematically into the product design phase in QFD. The application of the decision procedure is demon-
strated via a PVC window system producer. Since in a product design linguistic expressions like a strongly
more important CN or PTR than another CN or PTR may be used in case of incomplete information, the
fuzzy model in this paper can capture this vagueness. The ranking of the PTRs is almost the same for these
two proposed methods. When considering the interdependencies among PTRs, especially, the ranks for sup-
port sheets and nut and TiO2 pigment change. The rank of support sheets and nut changed from 3 in fuzzy
AHP to 5 in fuzzy ANP while the rank of TiO2 pigment changed from 5 in fuzzy AHP to 3 in fuzzy ANP.
According to the obtained results, the company should improve the PVC conductivity coefficient. Taking
into account the proposed fuzzy optimization framework, a budget planning which used the weights of fuz-
zy ANP given in Table 10 was made. Since the budget limit was uncertain, it was handled as a fuzzy con-
straint and the solution with the highest membership degree was obtained. This solution does not force the
decision-maker into a precise formulation because of mathematical reasons. For further research, different
fuzzy AHP approaches and fuzzy optimization models may be used to compare with the results obtained in
this paper.

References

Ansari, A., Modarress, B., 1994. Quality function deployment: The role of suppliers. International Journal of Purchasing and
Materials Management 30 (4), 47–62.
Armacost, R.L., Componation, P.J., Mullens, M.A., Swart, W.W., 1994. An AHP framework for prioritizing customer requirements
in QFD: An industrialized housing application. IIE Transactions 26 (4), 72–79.
Askın, R.G., Dawson, D.W., 2000. Maximizing customer satisfaction by optimal specification of engineering characteristics. IIE
Transactions 32, 9–20.
Bahrami, A., 1994. Routine design with information content and fuzzy quality function deployment. Journal of Intelligent
Manufacturing 5, 203–210.
Bergquist, K., Abeysekara, J., 1996. Quality function deployment (QFD)—A means for developing usable products. International
Journal of Industrial Ergonomics 18 (4), 269–275.
Bozdag, C.E., Kahraman, C., Ruan, D., 2003. Fuzzy group decision making for selection among computer integrated manufacturing
systems. Computers in Industry 51 (1), 13–29.
Buckley, J.J., 1985. Fuzzy hierarchical analysis. Fuzzy Sets and Systems 17 (3), 233–247.
Büyüközkan, G., Kahraman, C., Ruan, D., 2004. A fuzzy multi-criteria decision approach for software development strategy selection.
International Journal of General Systems, Taylor & Francis 33 (2–3), 259–280.
Chang, D.Y., 1992. Extent Analysis and Synthetic Decision, Optimization Techniques and Applications, vol. 1. World Scientific,
Singapore, p. 352.
Chang, D.Y., 1996. Applications of the extent analysis method on fuzzy AHP. European Journal of Operational Research 95, 649–
655.
Chan, L.K., Kao, H.P., Ng, A., Wu, M.L., 1999. Rating the importance of customer needs in quality function deployment by fuzzy
and entropy methods. International Journal of Production Research 37 (11), 2499–2518.
Cohen, L., 1995. Quality Function Deployment How to Make QFD Work for You. Addition-Wesley, Reading, MA.
Costa, A.I.A., Dekker, M., Jongen, W.M.F., 2001. Quality function deployment in the food industry: A review. Trends in Food
Science & Technology 11, 306–314.
Crow, K., 1999. QFD and Target Costing Case Study. DRM Associates, Palos Verdes, CA, Available from: <http://www.
npd-solutions.com/qrtncasestudy.htm>.
Crowe, T.J., Cheng, C.C., 1996. Using quality function deployment in manufacturing strategic planning. International Journal of
Operations and Production Management 16 (4), 35–48.
Doukas, L., William, P.W., Jeyaratnam, C., (1995). Integrating quality factors into system design. In: Proceedings of the 1995 IEEE
International Engineering Management Conference, pp. 235–240.
Erol, I., Ferrell, W.G., 2003. A methodology for selection problems with multiple, conflicting objectives and both qualitative and
quantitative criteria. International Journal of Production Economics 86 (3), 187–199.
Ertay, T., 1998. Simulation approach in comparison of a pull system in a cell production system with a push system in a conventional
production system according to flexible cost: A case study. International Journal of Production Economics 56–57, 145–155.
410 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411

Fukuda, S., Matsuura, Y., (1993). Prioritizing the customerÕs requirements by AHP for concurrent design. In: Design for
Manufacturability—1993 American Society of Mechanical Engineering, Design Engineering Division, vol. 52, pp. 13–19.
Fung, R.Y.K., Popplewell, K., Xie, J., 1998. An intelligent hybrid system for customer requirements analysis and product attribute
targets determination. International Journal of Production Research 36, 13–34.
Guinta, L.R., Praizler, N.C., 1993. The QFD Book: The Team Approach to Solving Problems and Satisfying Customers through
Quality Function Deployment. Amacom, New York.
Hämäläinen, R.P., Seppäläinen, T.O., 1986. The analytic network process in energy policy planning. Socio-Economic Planning
Sciences 20 (6), 399–405.
Hauser, J.R., Clausing, D., 1988. The house of quality. Harvard Business Review 66, 63–73.
Holmen, E., Kristensen, P.S., 1998. Supplier Roles in Product development: Interaction versus task partitioning. European Journal of
Supply Management 4, 185–193.
Hsiao, S.W., 2002. Concurrent design method for developing a new product. International Journal of Industrial Ergonomics 29,
41–55.
Jeong, M., Oh, H., 1998. Quality function deployment: An extended framework for service quality and customer satisfaction in the
hospitality industry. International Journal of Hospitality Management 17 (4), 375–390.
Kahraman, C., Ruan, D., Dogan, I., _ 2003. Fuzzy group decision making for facility location selection. Information Sciences 157,
135–153.
Kahraman, C., Cebeci, U., Ruan, D., 2004. Multi-attribute comparison of catering service companies using fuzzy AHP: The case of
Turkey. International Journal of Production Economics 87 (2), 171–184.
Karmarkar, U.S., Pitbladdo, R.C., 1997. Quality, class and competition. Management Science 43 (1), 27–39.
Karsak, E.E., Sozer, S., Alpteki, S.E., 2002. Product planning in quality function deployment using a combined analytic network
process and goal programming approach. Computers & Industrial Engineering 44, 171–190.
Khoo, L.P., Ho, N.C., 1996. Framework of a fuzzy quality function deployment system. International Journal of Production Research
34, 299–311.
Kim, K.J., Moskowitz, H., Dhingra, A., Evans, G., 1994. Fuzzy multicriteria models and decision support system for quality function
deployment. CMME Working Paper, Krannert Graduate School of Management, Purdue University, West Lafayette, IN.
Kim, K.J., Moskowitz, H., Dhingra, A., Evans, G., 2000. Fuzzy multicriteria models for quality function deployment. European
Journal of Operational Research 121 (3), 504–518.
Kwong, C.K., Bai, H., 2003. Determining the importance weights for the customer requirements in QFD using a fuzzy AHP with an
extent analysis approach. IIE Transactions 35, 619–626.
Laarhoven, P.J.M., Pedrycz, W., 1983. A fuzzy extension of SaatyÕs priority theory. Fuzzy Sets and Systems 11 (3), 229–241.
Lee, J.W., Kim, S.H., 2000. Using analytic network process and goal programming for interdependent information system project
selection. Computers & Operations Research 27, 367–382.
Lee, J.W., Kim, S.H., 2001. An integrated approach for interdependent information system project selection. International Journal of
Project Management 19, 111–118.
Leung, L.C., Cao, D., 2000. On consistency and ranking of alternatives in fuzzy AHP. European Journal of Operational Research 124,
102–113.
Lu, M., Madu, C.N., Kuei, C., Winokur, D., 1994. Integrating QFD, AHP, and benchmarking in strategic marketing. Journal of
Business and Industrial Marketing 9 (1), 41–50.
Masud, A.S.M., Dean, E.B., (1993). Using fuzzy sets in quality function deployment. In: Proceedings of the 2nd Industrial Engineering
Research Conference, pp. 270–274.
Meade, L., Sarkis, J., 1998. Strategic analysis of logistics and supply chain management systems using the analytical network process.
Transportation Research Part E: Logistics and Transportation Review 34 (3), 201–215.
Park, T., Kim, K., 1998. Determination of an optimal set of design requirements using house of quality. Journal of Operations
Management 16, 569–581.
Partovi, F.Y., 2001. An analytic model to quantify strategic service vision. International Journal of Service Industry Management 12
(5), 476–499.
Partovi, F.Y., Corredoira, R.A., 2002. Quality function deployment for the good of soccer. European Journal of Operational Research
137 (3), 642–656.
Partovi, F.Y., Epperly, J.M., 1999. A quality function deployment approach to task organization in peacekeeping force design. Socio-
Economic Planning Sciences 33 (2), 131–149.
Prasad, B., 1998. Review of QFD and related deployment techniques. Journal of Manufacturing Systems 17 (3), 221–234.
Pulat, B.M., 1994. Total quality management: A framework for application in manufacturing. The TQM Magazine 6 (1), 44–49.
Saaty, T.L., 1980. The Analytic Hierarchy Process. McGraw-Hill, New York.
Saaty, T.L., 1996. Decision Making with Dependence and Feedback: The Analytic Network Process. RWS Publications, Pittsburgh.
Saaty, T.L., Takizawa, M., 1986. Dependence and independence: From linear hierarchies to nonlinear networks. European Journal of
Operational Research 26, 229–237.
 C. Kahraman et al. / European Journal of Operational Research 171 (2006) 390–411 411

Saaty, T.L., Vargas, L.G., 1998. Diagnosis with dependent symptoms: Bayes theorem and the analytic hierarchy process. Operations
Research 46 (4), 491–502.
Shen, X.X., Tan, K.C., Xie, M., 2001. The implementation of quality function deployment based on linguistic data. Journal of
Intelligent Manufacturing 12 (1), 65–75.
Sohn, S.Y., Choi, I.S., 2001. Fuzzy QFD for supply chain management with reliability consideration. Reliability Engineering & System
Safety 72 (3), 327–334.
Wang, J., 1999. Fuzzy outranking approach to prioritize design requirements in quality function deployment. International Journal of
Production Research 37 (4), 899–916.
Yang, Y.Q., Wang, S.Q., Dulaimi, M., Low, S.P., 2003. A fuzzy quality function deployment system for buildable design decision-
makings. Automation in Construction 12, 381–393.
Zadeh, L.A., 1965. Fuzzy sets. Information and Control 8, 338–353.
Zadeh, L.A., 1973. The concept of a linguistic variable and its application to approximate reasoning. Memorandum ERL-M 411
Berkeley.
Zhang, Y., Wang, H.P., Zhang, C., 1999. Green QFD-II: A life cycle approach for environmentally conscious manufacturing by
integrating LCA and LCC into QFD matrices. International Journal of Production Research 37 (5), 1075–1091.
Zhou, M., 1998. Fuzzy logic and optimization models for implementing QFD. Computers and Industrial Engineering 35 (1–2),
237–240.
Zimmermann, H.-J., 1994. Fuzzy Set Theory and its Applications. Kluwer Academic Publishers, London.