Sei sulla pagina 1di 11

J Intell Manuf (2012) 23:14331443

DOI 10.1007/s10845-010-0473-5

An optimum design selection approach for product


customization development
Chen Liu Alejandro Ramirez-Serrano Guofu Yin

Received: 25 March 2009 / Accepted: 18 October 2010 / Published online: 6 November 2010
Springer Science+Business Media, LLC 2010

Abstract Consumer preferences and information on Introduction


product choice behavior can be of significant value in the
development processes of innovative products. In this paper, To attain competitive advantage for the product organiza-
product customization evaluation and selection model is tion, product design confronts unprecedented pressure for
introduced to support imprecision inherent of qualitative satisfying different market demands from todays global and
inputs from customers and designers in the decision mak- competitive business environment. The design decision pro-
ing process. Focusing on customer utility generation, an cedure of selecting an appropriate design result for a given
optimum design selection approach based on fuzzy set deci- product is perhaps the most critical stage in the product design
sion-making is proposed, where design attributes priority is when further product development is required.
identified from customer preferences using an analytical hier- Product design selection is influenced by many factors
archy process. A multi-attribute analysis diagram is devel- such as market demand based on customers preference, and
oped to visualize the preference of each attribute from the designers knowledge and experience in handling design and
experts group decision. Conjoint analysis is used in the prod- market issues (Besharati et al. 2006). Generally, custom-
uct customization to focus on customer utility generation in ers preference reflects the demand of market side, while
terms of multiple criteria. The use of the decision-making designers technical consideration is obviously critical in pre-
method is illustrated with a case example that highlights the dicting product performance with respect to its attributes.
utility of the proposed method. In fact, there are numerous difficulties associated with the
product evaluation and selection process which lie in the
Keywords Product customization selection Fuzzy sets complexity of solving design problems, adopting decision
Conjoint analysis Analytical hierarchy process criteria, and assessing product performance (Jiao and Tseng
1998; Zha et al. 2005), etc, about which customers may not
know or consider less at the point of purchase (Besharati et al.
2006), since consumers are mainly aware of their own needs
C. Liu
China Institute of Atomic Energy, without thinking about technical attributes that can be easily
Beijing, Peoples Republic of China used for evaluating design alternatives. Therefore, design-
e-mail: liuchenyy@gmail.com ers will inevitably confront the problem of comprehensively
regarding the customers preference and their own technical
C. Liu A. Ramirez-Serrano
Department of Mechanical and Manufacturing Engineering, consideration in product design.
University of Calgary, Calgary, AB, Canada There are a number of solutions reported in the literature
A. Ramirez-Serrano that have been commonly used for product design selection
e-mail: aramirez@ucalgary.ca and evaluation, such as Multi-Attribute Utility (MAU) the-
ory (Farquhar et al. 1977). MAU theory is a structured meth-
G. Yin (B)
odology designed in handling the tradeoffs among multiple
School of Manufacturing Science and Engineering,
Sichuan University, Chengdu, China objectives. It is applied to model a decision makers prefer-
e-mail: yingf.jim@gmail.com ences based on utility functions developed on a set of multiple

123
1434 J Intell Manuf (2012) 23:14331443

attributes. Such utility functions are used to quantify the or imprecise inputs from customers and designers. Numer-
desirability of certain alternatives in the areas of engineering ous publications and research works indicate that fuzzy set
and marketing for decision making. Besharati et al. (2006) analysis is a useful tool in decision-making problems with
proposed a generalized modeling approach for supporting the multiple goals or criteria (Shao et al. 2006; Zha et al. 2005).
selection in product design, where MAU approach is applied Moreover, fuzzy set analysis is appropriate in making the
to capture the designers considerations. best selection among several alternatives under defined cri-
In order to regard customer preferences in the multi- teria which are represented in fuzzy terms.
attribute design evaluation process, some existed evaluation In this paper, we develop an approach to comprise the cus-
procedures take into account the relative importance of (or tomers preferences and designers technical considerations
weighting factor for) each criterion being considered (Jiao in the process of selection and evaluation of product design,
and Tseng 1998). Further, the Analytical Hierarchy Process and thus can help consumers select the right product to meet
(AHP) (Saaty 1980) has been proven to be a more rigorous their needs. For this aim, a product evaluation and selection
procedure for determining customer preferences because it is model is defined to formulate the technical criteria and con-
relatively easy to use and can effectively handle both qualita- sider customers requirements. Then, a multi-attribute anal-
tive and quantitative data. This data is compared in a pairwise ysis diagram is proposed to visualize the preference of each
fashion with respect to the relative importance of (or prefer- attribute from experts viewpoints. Finally, a case study on
ence for) each criterion between hierarchical decision levels mobile robot selection and evaluation is exemplified to vali-
(Chiang et al. 2008; Kwong and Bai 2002). We note that Ayag date the modeling approach and the decision-making method.
(2005) also proposed to use the AHP technique to evaluate Technically, we use fuzzy set technique to analyze the uncer-
conceptual design alternatives in new product development. tainties of customers and designers, and fuzzy analysis tech-
Conjoint analysis (Green and Srinivasan 1978) is another nique to investigate customers responses to design options,
approach of solving the multi-attribute decision problems the proposed approach thus can make product design more
where the focus is preference elicitation at the individual effective and accurate, and could be adapted to different qual-
customer level (Besharati et al. 2006). In addition, in order itative design and customers criteria.
to overcome preference aggregation problems and account We mention that the method of making design decisions
for choice uncertainty, the conjoint model can be based on on the basis of customers requirements is an extension of
a discrete choice model that utilizes an individuals selec- the previous work (Liu et al. 2009), where the goal was to
tion behavior (Louviere et al. 2000). Chen et al. (2005) pro- provide an effective strategy for associating customer infor-
posed a product definition and customization system using mation with a given product concept.
an integrated methodology of conjoint analysis and Kohonen This paper is organized as follows. In Sect. Fuzzy deci-
association techniques. sion-making method, a description of the fuzzy decision-
Although there are numerous models that have been used, making method is introduced. Section An optimum selection
the deterministic models (such as the first choice model based approach for product customization development describes
on the customers utilities) have been applied in practice to an optimum selection approach for product customization
select the product with the highest customer utility. However, development based on the fuzzy decision-making method.
such techniques cannot deal with intangible criteria, since the Section A case study exemplifies an the proposed approach.
utility analysis approach requires the expected performance Finally, the concluding remarks of the paper are provided in
with respect to each criterion to be represented in a quantita- Sect. Conclusions.
tive form, for this reason it is difficult to quantify intangible
criteria such as some of the qualitative design criteria in early Fuzzy decision-making method
design stages (Thurston and Crawford 1994).
Indeed, to search for the optimal design alternatives with A good design decision-making model must be able to
the development of trade-off techniques for making deci- accommodate vagueness because fuzziness is a common
sions, effective product evaluation has to consider both tan- characteristic encountered in any design decision making
gible (e.g., physical dimensions and production cost) and problems. Under these conditions, the transformation of
intangible (e.g., reliability and compatibility) criteria, along qualitative preferences to precise numerical values is usually
with quantitative and qualitative performance measures. inappropriate, and decision-makers often therefore provide
From a practical viewpoint, consumers preference reflects uncertain estimates rather than precise answers.
the psychological requirements which are conceptually
vague with uncertainty and frequently presented in linguis- Representation of triangular fuzzy numbers
tic forms (Tsai and Hsiao 2004). Fuzzy analysis, based on
fuzzy set theory, excels in capturing semantic uncertainty A fuzzy set is defined as a pair (A, A ), where A R is a
with linguistic terms. It is capable of dealing with qualitative subset of real number set, and A : A [0, 1] is called

123
J Intell Manuf (2012) 23:14331443 1435

Fig. 1 Membership function of a triangular fuzzy number Fig. 2 Maximizing/minimizing triangular fuzzy numbers

the member function on A. A fuzzy number is a convex,


f i be the corresponding membership function. Let xmin =
normalized fuzzy set whose membership function is at
mini { pi1 } and xmax = maxi { pi3 }. Define the maximizing/min-
least segmentally continuous and has the functional value
imizing triangular fuzzy numbers as M = (xmin , xmax , xmax )
A (x) = 1 at precisely one element. A triangular fuzzy num-
and G = (xmin , xmin , xmax ), with their membership func-
ber, denoted by p = ( p 1 , p 2 , p 3 ), is a special case of fuzzy
tions as f M and f G respectively. Figure 2 gives an example
number, where p 1 , p 2 and p 3 R satisfy: p 1 p 2 p 3 ,
for n = 2. By the property of triangular fuzzy numbers, f G
and p ( p 1 ) = p ( p 3 ) = 0, p ( p 2 ) = 1. Figure 1 shows
and f M in a given design problem will inevitably intersect
the graph of function p . For simplicity, we will sometimes
the boundaries of the n membership functions f i . Let U M (i)
call the increasing boundary of the shape left-edge, while the
be the cross point of left-edge of f M and the right-edge of
decreasing boundary right-edge of p. See Zadeh (1965) for
f i , UG (i) be the cross point of right-edge of f G and the left-
details.
edge of f i . Formally, they are defined as
Triangular fuzzy numbers have a number of advantages
such as simple representation, speedy calculation and suffi- U M (i) = sup ( f M (x) f i (x)), (1)
x
cient precision that are relatively more effective than other
methods to represent customers expectations and experts UG (i) = sup ( f G (x) f i (x)), (2)
x
satisfactions (Xu et al. 2007). There are many applications
where the notion means the intersection operation of
that use triangular fuzzy numbers in engineering design.
fuzzy numbers, and sup means the supremum function.
Bykzkan et al. (2007) proposed a fuzzy group deci-
The total utility value of each fuzzy number is then defined
sion making approach to fuse multiple preference styles to
by
respond to customer needs with a set of triangular fuzzy num-
bers in the product development. Vanegas and Labib (2005) U M (i) + 1 UG (i)
UT (i) = . (3)
presented a number of fuzzy set approaches for design eval- 2
uation where weights of criteria and performance levels are By calculating the total utility value of each fuzzy number,
captured by triangular fuzzy numbers. the fuzzy numbers can be ranked in consequence.
Moreover, the fuzzy approach for ranking a set of alterna-
tives against a set of criteria is also needed to cope with uncer-
tain and incomplete information. In what follows, a fuzzy An optimum selection approach for product
ranking approach is described based on fuzzy set theory. customization development

Ranking fuzzy numbers In this paper, an optimum design selection approach for
product customization development based on fuzzy set deci-
In practice, ranking fuzzy numbers is a very important deci- sion-making is proposed. The approach is used to identify the
sion making procedure. For example, the design concept of optimal product design from a set of product design alterna-
optimum or best choice to come true are completely based on tives. The details of the approach are presented in this section.
ranking or comparison. As in Chen (1985), we will employ
the concept of maximizing set and minimizing set to decide Identification of design attributes priority
the order of the fuzzy numbers.
Assume there are n N triangular fuzzy numbers, The attributes priority is identified by an Analytical Hierar-
denoted by ( pi1 , pi2 , pi3 ) for i {1, . . . , n} respectively, and chy Process (AHP) from customer preference. The linguistic

123
1436 J Intell Manuf (2012) 23:14331443

variables on customer expectation preferences are considered Fuzzy evaluation of relative importance of individual
as the input for the product customization selection. Here, the customer expectation
customer expectation preferences are modeled with linguis-
tic sets. Fuzzy evaluation of customers expectations is analyzed by
their relative importance in a pairwise comparison way. Let
pql = ( pql 1 , p 2 , p 3 ) denote the measure value of expec-
ql ql
tation of the qth customer for the lth need, where q
Customer expectations modeling
{1, . . . , s}, l {1, . . . , k}, and s, k denote the number of
customers and their needs respectively.
The preference reflection of customers is frequently pre-
To obtain pairwise comparison ratios of the fuzzy num-
sented in linguistic forms in statements such as the cost 2 of fuzzy number p is taken to
bers, the peak value pql ql
should be low, the quality of the product should be good,
represent the importance intensity for the lth need. The rel-
etc. Customer expectations are captured difficultly by means
ative importance alv is considered as the important ratio of
of a definite structure. Fuzzy set theory is a rational approach
the lth need compared with the vth need. Formally, alv = 1
that uses linguistic characterizations toward decision-mak-
for the case of l = v; for other cases it is defined as
ing by taking into account human subjective judgments. The
linguistic variables are used to allow precise modeling of alv = pql
2
pqv
2
. (4)
imprecise statements such as very important or less satis-
2 p 2 , and
if pql
faction to linguistically evaluate the importance of the cus- qv
tomer needs. In the proposed approach any number of fuzzy
alv = 1/avl = 1/( pqv
2
pql
2
) (5)
sets can be used to model the customers expectation prefer-
ences. The number of fuzzy sets will depend on the problem otherwise, where l, v {1, . . . , k} and q {1, . . . , s}.
at hand. For example, in order to explain the level of expec- Thus, the relative importance of individual customer
tation of customer needs easily, six linguistic sets have been expectation is obtained.
developed: (1) Very Low, (2) Low, (3) Medium Low, (4)
Medium High, (5) High, (6) Very High. Moveover, these lin- Ranking the importance of customer needs
guistic sets can be quantified with corresponding triangular
fuzzy numbers as shown in Table 1, Fig. 3. To obtain the importance of all customer needs, the pairwise
comparison method, developed by Saaty (1980), is utilized
for ranking customer needs by their importance. A k k
matrix A is obtained from the pairwise comparison with rel-
Table 1 Linguistic variables for the importance and the ratings
ative importance of fuzzy customer expectations. The matrix
Linguistic variables Triangular A satisfies
fuzzy
numbers a11 a12 . . . a1k w1 w1
a21 a22 . . . a2k w2 w2

Very Low (VL) (0,0,1) .. .. .. .. .. = k .. (6)
Low (L) (0,1,3) . . . . . .
Medium Low (ML) (1,3,5) ak1 ak2 . . . akk wk wk
Medium High (MH) (3,5,7) We note that the values in the upper triangle of the matrix
High (H) (5,7,8) are the reciprocal values of the elements in the lower triangle
Very High (VH) (7,8,8) of this matrix, so only k(k1)
2 comparisons are needed for
ranking. Equation (6) can be denoted as
AW = kW, (7)
or
(A k I )W = 0, (8)
where I is an k k identity matrix. From this equation, it
is apparent that k is an eigenvalue of A , and W is an corre-
sponding eigenvector for k. Since the judgment matrix A is
created based on the decision makers subjective judgments,
the estimation errors will more or less exist in this pairwise
Fig. 3 Fuzzy membership function for expectation of customer needs comparison matrix. In such a case, the vector W satisfies

123
J Intell Manuf (2012) 23:14331443 1437

Table 2 Consistency indices for random reciprocal matrices with different orders
Size of matrix 1 2 3 4 5 6 7 8 9

RI 0 0 0.58 0.90 1.12 1.24 1.32 1.42 1.45

AW = max W, (9)

where max is the maximum eigenvalue of A considering


estimation errors, and it has

max k. (10)

Saaty (1980) introduced the consistency index CI to judge


the consistency of the judgments between max and k. CI is
the measure to evaluate the deviation from consistency of the
pairwise ratios and is calculated as
max k
CI = . (11)
k1
Fig. 4 The analytic hierarchy process structure for the selection of
When the elements of a comparison matrix are generated product alternatives
randomly, the CI for such matrix is represented as a random
consistency index RI. The values of RI for different orders
of matrices are summarized in Table 2. of product customization is structured into different hierar-
The consistency ratio CR is calculated to monitor the con- chical levels to provide representing and quantifying relat-
sistency of the comparison matrix as ing criteria. The overall customer goals are obtained from
multi-level structure for evaluating alternative solutions. The
CR = CI/RI. (12)
used analytic hierarchy process structure is shown in Fig. 4
If CR 0.1, the level of consistency is acceptable. The which reveals the relation between various criteria used for
pairwise comparison matrix is considered to be adequately the selection of product customization development. With the
consistent. If the CR exceeds this, the decision-maker is AHP, the objectives, criteria and alternatives are arranged in
advised to check the elements of matrix A to produce a more a hierarchical structure similar to a family tree (Albayrak and
consistent matrix. Erensal 2004).
In the process the importance of customers needs can be As shown in Fig. 4, the overall goal is expressed as a root
obtained by identifying the maximum eigenvalue max and node. Each criterion is categorized from the goal level. All
the corresponding eigenvector, W = (w1 , w2 , . . . , wk ), of attribute nodes, as design subcriterion, are the sub-nodes of
matrix A. corresponding criterion nodes and are listed in the subcriteri-
on (attribute) level. In this multilevel taxonomy, each node is
Identification of the attributes priority from customer stemmed from the high-level design goal, and each criterion
preferences influences on design alternatives in the lowest level of the
hierarchy. In practice, the number and content of nodes may
Since each individual design alternative is described by sev- be changed depending on the different goals that can be set
eral attributes based on the design specifications, evaluation in the problem.
of design candidates are easily accomplished based on design Customers needs are derived from each node of the cri-
attributes. The design attributes, i.e., technical terms used by terion level in the hierarchy. To quantify relative priority of
design engineers, are seen as techniques criteria. However, a given set of customers needs, the importance measures of
the designer deals with technical attributes that are not of k customer needs in the criterion level are identified as:
any interest to a customer (or are beyond customers knowl- 
edge) and customer needs are usually described by qualitative W X Y = w y 1 , w y 1 , . . . , w y 1 , . . . , w y 1 (13)
1 2 l k
expressions, evaluation of design candidates in terms of cus-
tomer satisfaction is difficult to be conducted directly. For this The nodes relative meaning and importance are judged
reason, the relationship between customer needs and product using pairwise compare method in attribute level. WY Z
attributes has to be developed when the design candidates are denotes the priority of attributes in the subcriterion level
evaluated based on customer needs. Hence, the development is identified from corresponding each criterion from the

123
1438 J Intell Manuf (2012) 23:14331443

criterion level. Thus, the overall priority of attributes W X Z


is achieved as:

W X Z = W X Y WY Z

= wy1 wy2 , . . . , wy1 wy2 ,
1 Y Z l Y Z

. . . , wy1 wy2 (14)
k Y Z

Accordingly, the consistency of the judgment matrix is


satisfactory when every CR is less than 0.1. It is said that Fig. 5 Fuzzy membership function for satisfaction of experts
the judgment matrix is logical. The overall consistency index
CI X Z and random consistency index RI X Z are computed
using the following functions: usually causes imprecise evaluation and serious difficulties
during the decision-making process. For this reason, the cur-

k
rent study proposes to use a group decision that improves the
CI X Z = CIY Z W X Y = CI y 1 w y 1 (15)
l=1
l l original evaluation based on various experts independent
evaluation.

k
RI X Z = RIY Z W X Y = RI y 1 w y 1 (16) Suppose there are Q experts {e1 , e2 , . . . , e Q }. Each expert
l l eq evaluates ri alternatives with respect to design attributes
l=1
CI X Z metrics t j for i {1, . . . , m} and j {1, . . . , n}. Denote by
CR X Z = < 0.1 (17) pqi j = ( pqi
1 , p 2 , p 3 ) the evaluation value of alternative i
RI X Z j qi j qi j
by expert q with respect to criterion j. pi j denotes the overall
Fuzzy evaluation of product alternatives from the experts evaluation for alternative i with criterion j and is calculated
point of view as follows:
T
pi j = p1i j , p2i j , . . . , pqi j , . . . , p Qi j
In this section, the overall fuzzy measure matrix of design
alternatives with respect to design attributes is achieved from = ( pi1j , pi2j , pi3j ), (18)
satisfaction of different experts. The level of satisfaction of
where
experts with respect to the given attributes of alternatives
are modeled in linguistic sets, and a multi-attribute analy- pi1j = min pqi j,
1
sis diagram is used to visualize the comparison between the q

different alternatives.
Q
pi2j = pqi2
j /q,
Expert satisfactions using linguistic sets q=1

pi3j = max pqi


3
j, (19)
q
The design attributes are technical terms used by design engi-
neers. It is thus necessary to identify the designers prefer- where i {1, . . . , m}, j {1, . . . , n}.
ences from the experts point of view. Due to the lack of suf- The overall measure matrix Pmn is calculated as
ficient information, the design attributes of alternatives are
usually evaluated by the experts in ambiguous terms such as p11 . . . p1 j . . . p1n

Reliability is not good. Thus, the satisfaction of experts Pmn = ... ... ... ... ... , (20)
with respect to design attributes of product in the subcriteri- pm1 . . . pm j . . . pmn
on (attribute) level is described by a measure corresponding
to six linguistic sets, which are used to model this measure: where m denotes the number of design alternatives and n
(1) Very Poor (VP), (2) Poor (P), (3) Medium Poor (MP), (4) denotes the number of design criteria.
Medium Good (MG), (5) Good (G), (6) Very Good (VG), as
shown in Fig. 5. Visual comparison on the alternatives

Fuzzy measure matrix of product alternatives with design Each individual design alternative may contain several attri-
attributes butes of the design specifications. The metrics for this level
of evaluation include not only the product functionality, but
Generally, different experts have different specific skill or also other criteria including some life-cycle issues, such as
knowledge about product design. If this is the case, it assembability, manufacturability, maintainability, reliability,

123
J Intell Manuf (2012) 23:14331443 1439

Conjoint analysis for customer utility generation in terms of


multiple criteria

The conjoint analysis, proposed by Green and Srinivasan


(1978), is used to evaluate multi-attribute alternatives on
the assumption of individual that customers responses are
approximately intervals in a measurement level. It is an
appropriate method that makes decision for the select of prod-
uct customization.
In this method, conjoint analysis strategy is used to obtain
the best satisfied product from various alternatives with
respect to the customer preferences based on the customers
utility generation. By integrating design attributes priority
from customer preferences and product alternatives from the
experts point of view, a multi-attribute preference ranking
Fig. 6 Multi-attribute analysis diagram approach for product customization using customers utility
metric is proposed. Herein, a customer utility metric forms
the basis for a decision modeling that supports the selection
and efficiency. In order to obtain the most desirable alterna- in product design. The utility function representing the cus-
tive, the comparison between attributes is necessary. From tomer preference on design alternatives is defined as
the experts point of view, the importance of each attribute
must be measured quantitatively. The multi-attribute priority U = Osn Pmn
T
, (21)
measurement of design alternatives need to be interpreted to where U is the customer utility for design alternatives, Osn
analyze preference judgments for ranking the relative pref- denotes the preference matrix from customer with respect to
erence levels of the alternatives. Therefore, it is necessary to the design criteria, and Pmn denotes the measure matrix of
represent the design alternatives in a visual way. Herein, a product alternatives from various experts.
visualization aid is presented to quantify the relative prefer- This conjoint analysis procedure has to be performed for
ence levels of attributes of design alternatives. the overall design alternatives under consideration, and the
As the illustrative multi-attribute analysis diagram shown output of the decision-making is the alternative with the high-
in Fig. 6, each attribute of alternatives indicates the quan- est customers utility value that meets both designers prefer-
titative measure of experts satisfaction level. The measure ences and market demand the best. For instance, the design
value is the peak value of the corresponding fuzzy number pi j alternative ri , which is ranked first as the customer preferred
which is the satisfaction of experts with respect to alternatives choice, is related to the maximum utility value Mi as
i. Some attributes of one design alternative under consider-
ation (e.g., y 2j ) have a superior value when compared to others Mi = max u i j , (22)
1 jm
alternatives, but other attributes of the same alternative are
inferior to others. Therefore, the multiple attributes analysis for i {1, . . . , s}. Therefore, the customer information is
diagram attempts to identify strong and weak points of each input into conjoint analysis procedure in terms of customer
attribute of alternatives. The diagram shows a gap between and designer preferences ratings, which affect the final results
the alternatives in each attribute, and then along with cus- of the design decision-making in terms of the customer
tomer preferences provides more information in the whole utility.
decision-making process.
However, since the selection of product customization
has different level of priority to the customer when obtain- A case study
ing the desirable design alternative, the final selection of
the optimum design scheme is obtained through a series A case study is applied to demonstrate the effect of the intro-
of evaluation, weighting, transformation, and computation duced optimum selection approach for product customization
processes. It is apparently that the multi-attribute decision- development. In this example, product design alternatives of
making method under customer expectations is essential for mobile robots with tracked locomotion are analyzed using
ranking the candidates. With these results, we will intro- the fuzzy decision support method. As a demonstration of
duces the conjoint analysis method in the following to obtain our approach, we only consider three robot design alterna-
an optimum selection based on customers and designers tives as shown in Table 3. The proposed approach meets indi-
preferences. vidual customer requirements via product customization for

123
1440 J Intell Manuf (2012) 23:14331443

Table 3 The basic information of three robot alternatives


Alternative Controller Speed Vision Wireless Rain Hill Price
sensor communication sensor capability ($)

r1 Manual and 50 km/h Omnidirectional camera Yes Yes 30 degree 5,200


Autonomous control
r2 Manual and 20 km/h Radar Optional No 15 degree 3,800
Autonomous control
r3 Manual and 15 km/h Radar and camera Optional No 18 degree 3,000
Autonomous control

different customers having different requirements. Specifi- Firstly, customers are chosen from different groups to con-
cally, the identification of the best design candidate is carried tribute their preferences of design needs. This is a key step
out for different customers in the following three steps. of the product development of robot design. For simplicity,
Step 1: Identifying the attributes priority from different two extreme types of customers are considered in our case.
customer preferences. In detail, one type of customers expect robots with high per-
The selection of alternatives is a process of integrating dif- formance and suitability and hope the selected robots can be
ferent views from the angles of engineering, economy and really used in industrial applications, while another type of
environment. For clarity, it is depicted as a hierarchy structure customers show more interesting to the cheap price and ele-
as shown in Fig. 7. Specifically, engineering performance mentary functions of robots as they will be mainly used in
explains the meaning of technical parameters from the view laboratories for teaching and experimental purposes.
of product engineering, and embodies as the attributes of Then, customers are asked to select the corresponding
kinematic accuracy, motion efficiency, and reliability. Econ- linguistic sets based on personal requirement (See Table 1;
omy performance involves two attributesproduction cost Fig. 3). In our case, the customer C1 expects three perfor-
and work life. Further, environment performance reflects the mance needs, i.e., VH, L and H; while they are ML,
ability of robots adapt for different environment and can be VH and L for customer C2 correspondingly.
detailed as three attributes: the structural complexity attri- The importance of customer expectations is determined
bute reflects the state or quality of robot; the noise attribute by the pairwise comparison method shown in Table 4 (See
denotes the interference level between robots and the sur- Sect. Fuzzy evaluation of relative importance of individual
rounding; while environment suitability attribute denotes the customer expectation and Ranking the importance of cus-
ability that robots adapt to different conditions. Therefore, tomer needs). Accordingly, the comparison matrixes of layer
the technical criteria for evaluating the robot can be naturally Y-Z are shown in Table 5. The pairwise comparison method
described as the above eight design attributes as a whole. can be performed to evaluate attributes priority from differ-
ent customer preferences. The priority of the eight design
attributes can be obtained from the two specific customers
using Eq. (14) as shown in Table 6 (See Sect. 3.1.4).
Step 2: Fuzzy evaluating the product alternatives from the
experts point of view.
In this example, four experts from design, manufacture,
usage and maintenance fields respectively are asked to con-
tribute their satisfaction on each design alternative with
respects to design criteria. These experts pick out suitable lin-
guistic sets according to the method shown in Fig. 5. to eval-
uate each alternative. As a result, Table 7. shows the measure
value of four experts and the overall evaluation result with
respects to design alternative r1 calculated using Eqs. (18)
and (19). Moreover, Table 8. lists the fuzzy measure matrix
of three robot alternatives with different design criteria from
four experts using Eq. (20).
Step 3: Conjointly analyzing customer utility generation
Fig. 7 The analytic hierarchy process structure for the selection of in terms of multiple criteria.
alternatives

123
J Intell Manuf (2012) 23:14331443 1441

Table 4 The result of pairwise


Layer W max CI CR
comparison matrix
X1 Y (0.9232, 0.1102, 0.3683) 3.0324 0.0162 0.0279
X2 Y (0.2181, 0.9682, 0.1228) 3.0142 0.0071 0.0122
y11 Z (0.1862, 0.8527, 0.4881) 3.0183 0.0091 0.0158
y21 Z (0.9487, 0.3162) 2.0 0 0
y31 Z (0.1102, 0.3683, 0.9232) 3.0324 0.0162 0.0279

Table 5 Paired comparison results for the Engineering (y11 ), Econ- The alternatives fuzzy evaluation membership functions
omy (y21 ) and Environment (y31 ) in the subcriterion level (Y) are ranked as shown in Figs. 8 and 9 in a descending order of
y11 y12 y22 y32 W y 1 Z U using Eqs. (1), (2) and (3). We note that the product design
1
alternative with the highest customer utility value will be the
y12 1 1/4 3 0.1862 best choice for product customization.
y22 4 1 2 0.8527 A simple analysis of the design evaluation results shows
y32 3 1/2 1 0.4881 the following:

y31 y62 y72 y82 W y 1 Z 1) The customer utility with respect to design alternatives is
3
treated as the customer preference level to those selected
y62 1 1/4 1/7 0.1102 alternatives.
y72 4 1 1/3 0.3683 2) The customer utility matrix is obtained with qualitative
y82 7 3 1 0.9232 inputs from customers and designers. The expectation of
customer needs and satisfaction of experts are modeled
y21 y42 y52 W y 1 Z
2 in the triangular fuzzy numbers.
y42 1 3 0.9487 3) Two customers show different preferences in selecting
y52 1/3 1 0.3162 design alternatives, i.e. customer C1 selects robot alter-
native r1 which has high performance, while customer
C2 prefers robot alternative r3 which satisfies the need
The customer utility matrix for different alternatives is of low price.
obtained from Eq. (21). For two specific customers, Table 9
lists their corresponding customer utility values with respect As demonstrated in the above example, the approach
to each alternative. The results of the fuzzy decision-making guides the designer to determine which of the alternatives
method for each design alternative are used to help the final could possibly satisfy corresponding customers and thus gain
decision making process of the product manager or customer. a higher potential demand for further product development.

Table 6 Attributes priority of two specific customers (Osn )


Customer Design attributes
ranking
t1 (y12 ) t2 (y22 ) t3 (y32 ) t4 (y42 ) t5 (y52 ) t6 (y62 ) t7 (y72 ) t8 (y82 )

C1 0.1719 0.7872 0.4506 0.1045 0.0348 0.0406 0.1356 0.3400


C2 0.0406 0.1860 0.1065 0.9185 0.3061 0.0135 0.0452 0.1134

Table 7 Measurement of four experts and the overall evaluation result with respects to design alternative r1
Design Design attributes

Alternative (r1 ) t1 t2 t3 t4 t5 t6 t7 t8

E1 VG G VG P MG MP VG MG
E2 VG VG G VP MP MG G MP
E3 G G MG MP MG MG VG MG
E4 MG G G P P MP G G
Overall(E) (3,7,8) (5,7.25,8) (3,6.75,8) (0,1.25,5) (0,3.5,7) (1,4,7) (5,7.5,8) (1,5,8)

123
1442 J Intell Manuf (2012) 23:14331443

Table 8 Fuzzy measure matrix of product alternatives (Pmn )


Design Design attributes
Alternative t1 t2 t3 t4 t5 t6 t7 t8

r1 (3,7,8) (5,7.25,8) (3,6.75,8) (0,1.25,5) (0,3.5,7) (1,4,7) (5,7.5,8) (1,5,8)


r2 (3,6,8) (3,5.5,8) (1,4.5,8) (1,3.5,7) (1,3.5,7) (1,3.5,7) (1,4,7) (0,3.5,7)
r3 (0,3.5,8) (0,1.75,7) (3,6,8) (5,7.8,8) (5,7.8,8) (1,4.5,7) (0,1.75,5) (0,3,8)

Table 9 Customer utility U


from two customers Customer Design alternatives
r1 r2 r3

C1 (6.8621,12.0619,14.4975) (3.6434,9.7508,15.8661) (2.0889,7.2094,15.2870)


C2 (1.7242,5,5218,10.7633) (2.0696,6.1976,12.4417) (6.4560,11.1318,13.5033)

enables engineers to evaluate design alternatives with related


criteria such as the customers preferences and engineerings
view points.

Acknowledgments The research reported in this paper is supported


by the National Natural Science Foundation of China through Grant
No. 50575153 and Technology Research & Development Foundation
of Sichuan Province of China through Grant No. 2008GZ0063. We
acknowledge the support from the China Scholarship Council.
Fig. 8 Membership function for the utility ranking of customer C1

References

Albayrak, E., & Erensal, Y. C. (2004). Using analytic hierarchy pro-


cess (AHP) to improve human performance: An application of
multiple criteria decision making problem. Journal of Intelligent
Manufacturing, 15, 491503.
Ayag, Z. (2005). A fuzzy AHP-based simulation approach to concept
evaluation in a NPD environment. IIE Transactions, 37(9), 827
Fig. 9 Membership function for the utility ranking of customer C2 842.
Besharati, B., Azarm, S., & Kannan, P. (2006). A decision support
system for product design selection: A generalized purchase
modeling approach. Decision Support Systems, 42(1), 333350.
Conclusions
Bykzkan, G., Feyzioglu, O., & Ruan, D. (2007). Fuzzy group deci-
sion-making to multiple preference formats in quality function
We introduce an optimum selection method for product cus- deployment. Computers in Industry, 58(5), 392402.
tomization development using fuzzy decision-making tech- Chen, C., Khoo, L., & Yan, W. (2005). PDCS-a product definition and
customisation system for product concept development. Expert
nique. The proposed approach trades off customers and Systems with Applications, 28(3), 591602.
designers preferences by depicting them as fuzzy sets to Chen, S. (1985). Ranking fuzzy numbers with maximizing set and
make it possible to better reflect the designers estimate of minimizing set. Fuzzy Sets and Systems, 17(2), 113129.
the performance of design alternatives as well as the cus- Chiang, D., Guo, R., & Pai, F. (2008). Improved customer satisfaction
with a hybrid dispatching rule in semiconductor back-end facto-
tomers expectations. It facilitates more accurate and precise ries. International Journal of Production Research, 46(17), 4903
linguistic inputs, and provides a ways to fuzzify numeric 4923.
inputs. By adopting the conception of fuzzy triangular num- Farquhar, P. (1977). A survey of multiattribute utility theory and
bers, estimate values with linguistic sets are discussed in this applications. TIMS Studies in the Management Sciences, 6, 59
89.
paper. This paves the way for the analytical hierarchy pro- Green, P., & Srinivasan, V. (1978). Conjoint analysis in con-
cess to assist designers to determinate the attributes weights. sumer research: issues and outlook. Journal of Consumer
This method is beneficial in improving design capability as it Research, 5(2), 103123.

123
J Intell Manuf (2012) 23:14331443 1443

Jiao, J., & Tseng, M. (1998). Fuzzy ranking for concept evalua- Thurston, D., & Crawford, C. (1994). A method for integrating end-user
tion in configuration design for mass customization. Concurrent preferences for design evaluation in rule-based systems. Journal
Engineering, 6(3), 189206. of Mechanical Design, 116(2), 522530.
Kwong, C. K., & Bai, H. (2002). A fuzzy AHP approach to the Tsai, H., & Hsiao, S. (2004). Evaluation of alternatives for product
determination of importance weights of customer requirements customization using fuzzy logic. Information Sciences, 158, 233
in quality function deployment. Journal of Intelligent Manufac- 262.
turing, 13, 367377. Vanegas, L., & Labib, A. (2005). Fuzzy approaches to evaluation in
Louviere, J., Hensher, D., & Swait, J. (2000). Stated choice methods: engineering design. Journal of Mechanical Design, 127(1), 2433.
analysis and applications. : Cambridge University Press. Xu, L., Li, Z., Li, S., & Tang, F. (2007). A decision support system
Liu, C., Ramirez-Serrano, A., & Yin, G. F. (2009). Customer-driven for product design in concurrent engineering. Decision Support
product design and evaluation method for collaborative design Systems, 42(4), 20292042.
environments. Journal of Intelligent Manufacturing. doi:10.1007/ Zadeh, L. (1965). Fuzzy sets. Information and Control, 8, 338353.
s10845-009-0334-2. Zha, X., Sriram, R., & Lu, W. (2005). Evaluation and selection in
Saaty, T. L. (1980). The analytic hierarchy process. New York: product design for mass customization: A knowledge decision
McGraw-Hill. support approach. AI EDAM, 18(01), 87109.
Shao, X. Y., Wang, Z. H., Li, P. G., & Feng, X. J. (2006). Integrating
data mining and rough set for customer group-based discovery of
product configuration rules. International Journal of Production
Research, 44(14), 27892811.

123

Potrebbero piacerti anche