Int J Adv Manuf Technol DOI 10.1007/s00170-014-6233-5
ORIGINAL ARTICLE
Fuzzy PROMETHEE GDSS for technical requirements ranking in HOQ Seyyed Mahdi Hosseini Motlagh & Majid Behzadian & Joshua Ignatius & Mark Goh & Mohammad Mehdi Sepehri & Tan Kim Hua
Received: 5 January 2012 / Accepted: 4 August 2014 # Springer-Verlag London 2014
Abstract This paper provides a fuzzy Preference Ranking Organization Method for Enrichment Evaluation (PROMETHEE) in a Group Decision Support System (GDSS) approach to ranking the technical requirements for the house of quality (HOQ) process in multi-criteria product design. The problem under study involves incorporating the design alternatives of a group of designers located in different geographies who often provide vague and imprecise linguistic design information to the HOQ process. As such, the proposed fuzzy PROMETHEE GDSS allows the quality function deployment (QFD) team of designers to minimize any deviation arising from the individual designer preferences and to capture the ambiguity of the imprecise design information when S. M. Hosseini Motlagh School of Industrial Engineering, Iran University of Science and Technology, Tehran, Iran M. Behzadian School of Industrial Engineering, Shomal University, Tehran, Iran J. Ignatius (*) School of Mathematical Sciences, Universiti Sains Malaysia, Penang, Malaysia e-mail:
[email protected] M. Goh Decision Sciences Department, School of Business, National University of Singapore, Singapore, Singapore M. Goh School of Business IT and Logistics, RMIT University, Melbourne, Victoria 3000, Australia M. M. Sepehri Department of Industrial Engineering, Tarbiat Modares University, Tehran, Iran T. K. Hua Nottingham Business School,, University of Nottingham, Nottingham, UK
expressing the importance of customer needs and to delineate the linkage between customer needs and the technical requirements. The approach advances the HOQ group decisionmaking context in two important aspects. First, it treats each criterion and decision maker (DM) as unique in terms of the preference function and threshold levels. Second, it facilitates a rapid communication among DMs for the HOQ process. A case of a design team for an ergonomic chair manufacturer serves to validate this approach. Keywords Fuzzy . GDSS . PROMETHEE . Quality function deployment . House of quality
1 Introduction Today, firms need to develop products and services that meet customer needs (CNs) and hence customer satisfaction. Often, quality function deployment (QFD), developed in the 1960s [1], is used as a strategic tool to ensure that the voice of the customer is heard throughout the phases of product design and manufacturing, in translating the CNs into the product-based technical requirements (TRs) by integrating the diverse ideas across the marketing, design, engineering, manufacturing, and other key functions of a firm [16]. QFD uses four sets of matrices: (i) product planning (also known as the house of quality (HOQ)), (ii) part development, (iii) process planning, and iv) production planning [15]. The HOQ, as shown in Fig. 1, contains valuable information about the CNs and their relative importance, the relationships between the CNs and TRs, the correlation among the TRs, and the relative importance of the TRs. This paper focuses on the HOQ matrix, which describes the translation of the CNs into a list of TRs, which can sometimes entail complicated decision making.
Int J Adv Manuf Technol Fig. 1 Representation of HOQ chart and its steps Step 6
Step 4 Step 1: Identify CNs Step 2: Find relative importance of CNs.
Step 5
Step 6: Prepare correlation matrix.
Strong
●
9
Medium
◘
3
Weak
○
1
Step 3
Step 5: Prepare relationship matrix.
Step 2
Step 4: Determine TRs
Step 1
Step 3: Customer competitive assessment.
Step 7: Rank TRs and define target Step 7
Although QFD has been successfully applied across many industries, several drawbacks have been reported in its implementation process. According to an in-depth review of the QFD publications, the implementation problems can be categorized as follows: (1) the lack of time to go through all the processes [22, 11, 10]; (2) the large size of the matrices [42, 13]; (3) the inability to discriminate among diverse and conflicting CNs (Benner et al. [5]; [43]); (4) the difficulty in reaching an agreement arising from the conflicting TRs [21, 9]; (5) the inaccuracy in ranking CNs and TRs with the conventional scaling [26, 27]; (6) the inability to respond to the dynamic changes in CNs [45, 2, 38]; (7) poor understanding of CNs across various segments [46, 23, 29]; (8) the focus on the relationship between TRs and cost intead of extending it to other factors affecting the supply chain [18, 40, 28, 32]; (9) the vagueness and ambiguity in the ranks, relationship matrix, and correlation matrix are ignored [30, 31, 12]; and (10) the ineffective knowledge management during the product development process [47]. Hence, this research attempts to improve and illustrate the applicability of the QFD technique. This paper focuses on the HOQ matrix, which describes the transition of CNs into a list of TRs. It aims to highlight the problem of ranking TRs among a group of experts in the HOQ process that are virtually located. In such environment, rapid communication and resolving trade-offs among stakeholders are essential in achieving product development excellence. Hence, successful QFD implementation requires an effective decision-making method to guide a common consensus among the stakeholders. We note the absence of the Preference Ranking Organization Method for Enrichment Evaluation (PROMETHEE) Group Decision Support System (GDSS) method [7] in the literature coverage and seek to introduce its usefulness in ranking a set of TRs for product planning. Incorporating the fuzzy set theory approach into PROMETHEE GDSS enables the group to properly account for the ambiguity and uncertainty in group decision making.
The rest of the paper is organized as follows. Sect. 2 presents the literature review on QFD and PROMETHEE. Sect. 3 details the fuzzy PROMETHEE GDSS approach. Sect. 4 validates the methodology through a case of an ergonomic chair design. Sect. 5 concludes the paper with suggestions for future research.
2 Literature review In the literature, only a few applications of PROMETHEE were applied to group decision making [3] let alone QFD. This is pale in comparison to the more popular analytic hierarchy process approach in QFD [37] although PROMETHEE has the advantage of allowing the decision makers to specify their preference functions. Raju et al. [39] apply the PROMETHEE GDSS method to rank the alternative strategies for the planning scenario of an irrigation system. Haralambopoulos and Polatidis [20] employ the PROMETHEE method to evaluate and rank renewable energy projects according to a multi-criteria group decision-making action plan. Likewise, to select the best location for an electricity power plant in the EU, Leyya-Lopez and Fernandez-Gonzalez [34] conduct a comparative study of PROMETHEE for group decision making with an extension of the ELECTRE III multi-criteria outranking methodology. Morais and de Almeida [36] later propose a procedure with four stakeholders to develop a leakage management strategy. Although the PROMETHEE GDSS method has been designed for use in a colocated group environment, it is equally applicable to a virtual setting using tele or video conferencing systems [7]. The literature contains even fewer studies on group decision-making techniques for ranking the TRs that incorporate the subjectivity of the decision makers in the modeling framework. Since decision maker subjectivity often affects the ranking of the TRs, allowing the decision makers to express values in linguistic terms and accounting for the fuzziness of their value judgment are imperative to group decision making.
Int J Adv Manuf Technol
Hence, fuzzy set theory is often relied upon to deal with the vague and imprecise linguistic terms in the HOQ process. In terms of the TR analysis, Kim et al. [30] combine fuzzy regression and fuzzy optimization theory to deal with the QFD team’s performance trade-offs. Buyukozkan and Feyzioglu [9] introduce a new group decision-making approach that accounts for multiple preference formats, which combines different expressions into a uniform group decision using fuzzy set theory. Subsequently, Liu and Wu [35] present a hybrid version of the group decision-making problem for the customer requirement management of the QFD process. Recently, Buyukozkan et al. [10] propose a fuzzy preference relation technique for aggregating opinions across a crossfunctional team. Nonetheless, current studies do not clarify the challenges arising from evaluating TRs in QFD group decision making, especially the means to resolve poor communication among group members. In many studies, the decision-making process is assumed to be on an ad hoc once-off “face-to-face” basis. However, getting expert opinions resolved in a colocated setting is extremely difficult, and this often would delay the QFD process. Therefore, the main contribution of this paper is to resolve the problems associated with poor communication among team members, who are virtually colocated. These team members may exhibit different weights and tolerances in evaluating the TRs of a QFD project. For this purpose, this study proposes a PROMETHEE GDSS method for rapidly aggregating individual preferences expressed by the team members. Incorporating the fuzzy set theory into PROMETHEE GDSS enables to better capture the ambiguity and uncertainty in group decision making. The PROMETHEE method combined with fuzzy set theory has been applied to environment management, hydrology and water management, energy management, and logistics and transportation. For instance, Le Teno and Mareschal [33] develop a version of PROMETHEE with interval criteria and fuzzy set theory to evaluate the environmental quality of building products through life cycle assessment (LCA). Geldermann et al. [19] also combine PROMETHEE with fuzzy set theory to rank sinter plants through LCA based on 12 impact factors. In logistics and transportation, Radojevic and Petrovic (1997), recognizing that PROMETHEE handles human judgment in a simplistic manner, introduce fuzzy if-then rules that linked the criterion value difference and its preference function. Likewise, Fernandez-Castro and Jimenez [17] apply fuzzy PROMETHEE to rank and select distribution centers in Belgium. Other applications of fuzzy PROMETHEE are hospital Web site evaluation [6], eco-technology model evaluation [14], material handling equipment selection [43], training provider evaluation [25], and optimal choice of investments [44].
3 Fuzzy PROMETHEE GDSS To rank the alternatives from a group of DMs, the fuzzy PROMETHEE GDSS goes through the following procedure. Steps 0.1 and 0.2 are the data preprocessing needed for the fuzzy numbers prior to conducting the PROMETHEE GDSS computation. Step 0.1: Generate alternatives and criteria The process begins with generating alternatives and criteria by the facilitator who poses decision problems to the DMs. For this purpose, the facilitator meets the DMs together or individually so that each DM proposes possible alternatives and criteria including their extended description. All the proposed alternatives are collected and displayed by the facilitator, who then provides a frame of the evaluation table. Step 0.2: Fuzzy data elicitation and preparation The DMs express their ratings linguistically, i.e., “very strong” and “weak.” The linguistic variables are modeled as a series of triangular fuzzy numbers instead of crisp numbers. A triangular e takes the form (L, M, U) with fuzzy number A eðxÞ (Fig. 2). membership function μA 8 0; x < L; > > > x−L > < ; L ≤ x ≤ M; eðxÞ ¼ M−L ð1Þ μA x−H > > ; M ≤ x≤ H; > > : H−M 0; x>H The triangular fuzzy number is presented in the LR form of Geldermann et al. [31]: ðL; M ; U Þ ¼ ðM ; M −L; U −M ÞLR
ð2Þ
The term LR is used to denote the left and right spreads of the triangle. Table 1 contains the basic fuzzy operations in the LR mode. The evaluation of alternatives and criteria is expressed using this fuzzy form. For instance, very strong is expressed as (8,9,9)=(9,1,0)LR
Fig. 2 Graphical representation of a triangular fuzzy number
Int J Adv Manuf Technol Table 1 Basic fuzzy LR operations Addition
(m,α,β)LR ⊕(n,γ,δ)LR =(m+n,α+γ,β+δ)LR
Opposite −(m,α,β)LR =(−m,β,α)LR Subtraction (m,α,β)LR −(n,γ,δ)LR =(m−n,α+δ,β+γ)LR Multiplication by (m,α,β)LR ⊗(n,0,0)LR =(mn,αn,βn)LR scalar Multiplication by fuzzy For m>0,n>0 (m,α,β)LR ⊗(n,γ,δ)LR =(mn,mγ+nα,mδ+nβ)LR For m<0,n>0 (m,α,β)LR ⊗(n,γ,δ)LR =(mn,nα−mδ,nβ−mγ)LR For m<0,n<0 (m,α,β)LR ⊗(n,γ,δ)LR =(mn,−nβ−mδ,nα−mγ)LR
Step 1: Evaluate alternatives pairwise for each criterion and DM The difference score between two alternatives for criterion j is derived as follows: e dj ¼ Cj e a −C j e b
ð3Þ
where Cj ðe aÞ represents the membership function for criterion j on triangular fuzzy number e a in the LR mode. For instance, if C j ðe aÞ ¼ very strongð9; 1; 0Þ and Cj e b ¼ very weak ð1; 0; 1Þ , t h e n e dj ¼ ð9−1; 1 þ 1; 0 þ 0Þ ¼ ð8; 2; 0Þ . Step 2: Set the preference function and threshold selection for each criterion and DM Brans et al. [8] present the shape of the six possible choices of the preference functions to assist the DMs with this selection. For example, a DM who selects a type 3 function (Fig. 3) is required to specify a threshold value of p (strict preference). This preference threshold can be interpreted as the smallest tolerance that the DM can bear to reflect its sufficiency condition to be considered as maximizing full preference. The type 3 function can be expressed as follows: 8 0 < e > d P e dj ¼ > p : 1
where e d ¼ ðM ; M −L; U −M ÞLR . Assume that e d j ¼ ð6; 3; 2Þ and the DMs agree on a threshold value of p = 8. Then, P e dj ¼ 6 3 2 p ; p ; p ¼ ð0:75; 0:375; 0:25Þ . Step 3: Weigh the preferences for each DM The weights of the criteria are incorporated into the rating of alternatives with ∑wj =1. The weighted preference indicator for fuzzy PROMETHEE can be expressed as follows: X π e dj ¼ dj w jP e
ð5Þ
Consider the simple case of a pair of alternatives (1, 2) of two criteria. Hence, π e d j ¼ πð1; 2Þ and suppose w1 =0.3, w2 =0.7 and P e d 1 ¼ ð1; 0; 2Þ and P e d 2 ¼ ð3; 2; 2Þ , then π(1,2) = 0.3(1,0,2) ⊕ 0.7(3,2,2) = (0.51,1.4,2) Step 4: Calculate the outranking flows for each DM There are two types of flows, positive (ϕ+) and negative (ϕ−). For an alternative to be preferred over another, φ+(a) > φ+(b) and φ−(a) <φ−(b) where ϕ+ sums all outgoing paths from node a in the form π(a,x), with a as the origin node in an outranking graph while x represents the rest of the destination nodes. For instance, π(a,x) for x=b,c where π(a,b)=(0.7,0.4,0.2) and π(a,c)=(0.6,0.3,0.2). Then, ϕ+(a)=1/2[(0.7,0.4,0.2)⊕(0.6,0.3,0.2)]= 1/2(1.3,0.7,0.4)=(0.65,0.35,0.2). The computation for ϕ− is similar to ϕ+, but the origin node is now the destination node. For instance, let π(x,a) for x=b,c, where π(b,a)=(0.55,0.4,0.25) and π(c,a)=(0.35,0.25, 0.2).
if L ≤ 0 if 0 ≤ L and U ≤ p if U ≤ p
ð4Þ
ϕ− ðaÞ ¼ 1=2½ð0:55; 0:4; 0:25Þ⊕ð0:35; 0:25; 0:2Þ ¼ 1=2ð0:9; 0:65; 0:45Þ ¼ ð0:45; 0:325; 0:225Þ: Step 5: Calculate the net flow for each DM For a net comparison, the positive flow for alternative a is as follows:
Fig. 3 Type 3 preference function for PROMETHEE
ϕnet ðaÞ ¼ ϕþ ðaÞ−ϕ− ðaÞ
ð6Þ
Int J Adv Manuf Technol
In our two-criteria example, the net flow is calculated as ϕnet ðaÞ ¼ ð0:65; 0:35; 0:2Þ−ð0:45; 0:325; 0:225Þ ¼ ð0:2; 0:575; 0:525Þ:
ðM ; M −L; U −M ÞLR ¼ ½3M −ðM −LÞ þ ðU −M Þ=3
Step 6: Calculate the net flows for all DMs The net flow vectors of all the DMs are then put into a global decision matrix. Steps 1–5 are repeated for the group net flow. However, instead of using e d j to represent the fuzzy difference between the two alternatives for criterion j, e d k is used to represent the fuzzy difference between two alternatives for each of the k DMs. The equations below are a generalization of steps 1–5 in a group decision-making context. Thus, the fuzzy PROMETHEE GDSS can be summarized as follows: h i Pi e a; e b ¼ F i ϕi e a −ϕi e b i ¼ 1tok
ð7Þ
where Pi(a,b) denotes the preference of alternative a with respect to alternative b for k DMs. m X πgdss e Pi e a; e b ¼ a; e b wi
ð8Þ
i¼1
where πgdss e a; e b is defined as the fuzzy weighted sum Pi e a; e b for all DMs. The PROMETHEE GDSS partial and complete rankings are obtained as shown. 1 X πgdss e a ¼ a; ex and ϕþ gdss e m−1 x∈A 1 X ϕ− gdss e πgdss ex; e a ¼ a m−1 x∈A
ϕgdss e a ¼ ϕgdss þ e a −ϕgdss − e a
Step7: Defuzzification In order to provide a ranking of alternatives, the LR triangular fuzzy number is defuzzified as follows [41]:
ð9Þ
ð11Þ F o r ex a m p l e , th e d e f u z z i f i e d va l ue o f (0.5, 0.6, 0.9) = 3(0.5) −(0.6) + (0.9)/ 3 is 0.6. The defuzzified value is then used to rank the alternatives, where a higher alternative is given greater preference. The algorithm for the fuzzy PROMETHEE GDSS is programmed via MATLAB 2008.
4 Case example We consider the design of high-end ergonomic office chairs, by a manufacturing company identified for its well-known brand of ergonomic office chairs, to illustrate the performance of the proposed methodology. The company usually offers a range of models that are designed based on CNs and ergonomic standards. Recently, the after sales department has registered a considerable number of complaints arising from the innovation in engineering and quality of the products, which saw customer satisfaction rate reduced by 20 %. Hence, the company embarked on a QFD project to improve the development process of an existing product line based on CNs. In the first attempt, a joint task force was formed, which comprises three product designers working in different locations and a facilitator from the marketing department. Since all designers (who are also the DMs in the QFD team) were geographically scattered, it was decided that a virtual team with online video communication be established. The facilitator played a crucial role in ensuring smooth coordination throughout the video conference sessions (i.e., 3 h per session). The stages involved in the ranking of the TRs in a group decision-making context are provided below. 4.1 Stage I: Identify CNs and their relative importance
ð10Þ
where ϕgdss − ðe aÞ and ϕgdss ðe aÞ denote the positive, negative, and the net outranking flow for each alternative, respectively, in fuzzy terms.
The facilitator initiated the study with a thorough evaluation of the needs of the customers across a set of criteria such as sturdy arm rest (CN1), flexibility of swivelling (CN2), smooth and easy seat position adjustment (CN3), height suitability (CN4), lumbar support (CN5), durability and life expectancy (CN6), and aesthetics (CN7). This includes an evaluation of the previous customer complaints related to these criteria. The
Int J Adv Manuf Technol
matrix. To measure the relationships between the CNs and TRs, five linguistic terms in the following scale (VW to VS) were introduced.
facilitator also initiated a questionnaire survey and numerous face-to-face interviews with prospective customers. Prospective customers were asked to rate the relative importance of the seven CNs using the following five linguistic terms (VU to VI). Very unimportant Unimportant Medium Important Very important (VU) (U) (M) (I) (VI) [1, 0, 1]
[3, 1, 1]
[5, 2, 2]
[7, 1, 1]
[9, 1, 0]
Very weak (VW)
Weak (U)
Moderate (M)
Strong (S)
Very strong (VS)
[1, 0, 1]
[3, 1, 1]
[5, 1, 1]
[7, 1, 1]
[9, 1, 0]
For example, DM1 perceives that CN1 is very strongly associated with TR1 but very weakly associated with TR8 (see Table 3). Table 3 further provides an overall HOQ matrix, i.e., the relationships between the CNs and TRs for all DMs. The final weights of each DM are obtained as follows:
The linguistic assessments of the CNs and transformed results are shown in Table 2. For example, customer 1 considers CN1 as “important,” which was represented by a triangular fuzzy number [7, 1, 1]LR. The relative importance ratings of each CN is obtained by averaging the fuzzy linguistic terms of 50 customers (see final ratings in Table 2). The fuzzy arithmetic operations can be obtained from Table 1.
e dk ¼
7 X
eclet m ∀k∈k;
ð12Þ
l¼1
where e d k refers to the final fuzzy rating of each DM k based on the relationships between the lCNs and m TRs. For instance, the final fuzzy rating for DM1 for TR1 can be computed as follows:
4.2 Stage II: Generate TRs and the criteria for decision making The QFD team in the first round of online communications identified eight TRs: height and width adjustable arms (TR1), degree of base movement (TR2), pneumatic height adjustment (TR3), the range of seat pan height (TR4), back height adjustment (TR5), back angle adjustment (TR6), the quality of materials (fabric and textile) used (TR7), and the shape of the components (TR8).
ð½4:6; 0:9; 0:8⊗ ½9; 1; 0Þ⊕ð½7:2; 1; 0:66⊗½5; 1; 1Þ⊕ ð½4:4; 0:92; 0:9⊗½1; 0; 1Þ ¼ ½110:6; 29:82; 20:06
4.4 Stage IV: Rate the TRs for each DM 4.3 Stage III: Build relationship matrix of CNs and TRs of each DM
Aside from establishing the strength of the association between the CNs and TRs or the relative importance from the customers’ perspective, another two criteria are used for assessing the TRs, namely technical difficulty and estimated expenditure. Technical difficulty describes the level of
The QFD team is required to determine the strength of the association between the CNs and the TRs. To begin the process, each member filled in his own HOQ
Table 2 Relative importance rating of CNs Customer 1
CN1 CN2 CN3 CN4 CN5 CN6 CN7 CN8
Customer 2
Customer i
Customer n
Final ratings
Perception
Fuzzy
Perception
Fuzzy
Perception
Fuzzy
Perception
Fuzzy
I VU U U VI I M VI
[7, 1, 1] [1,0,1] [3, 1, 1] [3, 1, 1] [9,1,0] [7, 1, 1] [5, 1, 1] [9,1,0]
I U M M VI I I U
[7, 1, 1] [3, 1, 1] [5, 1, 1] [5, 1, 1] [9,1,0] [7, 1, 1] [7, 1, 1] [3, 1, 1]
… … … … … … … …
… … … … … … … …
VU U U I U VI M VU
[1, 0, 1] [3, 1, 1] [3, 1, 1] [7, 1, 1] [3, 1, 1] [9,1,0] [5, 1, 1] [1,0,1]
VU very unimportant, U unimportant, M medium, I important, VI very important
[4.6, 0.9, 0.8] [1.6, 0.42, 1] [2.8, 0.54, 0.96] [4, 0.88, 0.9] [5.8, 0.94, 0.82] [7.2, 1, 0.66] [4.2, 0.9, 0.88] [4.4, 0.92, 0.9]
[95.5, 32.24, 27.46] [14.4, 5.38, 9] [31, 8.6, 15.26] [81.8, 25.82, 24,62] [14.4, 5.38, 9] [19.6, 6.58, 9.52]
Final rating for DM1
Final rating for DM2
Final rating for DM3
Relative Importance
L
VH
H
VH
D
VD
D
L
L
L
E
VE
E
[87.8, 28.96, 24.16]
VL very low [2, 1.5, 1.5], L low [5, 2, 2], M modest [9, 3, 3], H high[14, 3, 3], VH very high [20, 5, 5]
VE very easy [1, 0, 2], E easy[3, 2, 2], MD moderately difficult [5, 2, 2], D difficult[7, 2, 2], VD very difficult [9, 2, 0]
VS [35.6, 12.68, 8.1]
M
L
M
MD
E
VE
M
M
L
D
D
MD
VH
VH
VH
E
E
MD
M
M
H
VD
D
VD
[64.8, 16.2, 5.94] [64.8, 16.2, 5.94] [72.2, 24.18, 26.14] [49.4, 19.98, 15.64]
[102.2, 31.96, 25.48] [50.4, 14.2. 11.82] [64.8, 16.2, 5.94] [95, 34.18, 25.2]
VW very weak [1, 0,1], W weak [3, 1, 1], M moderate[5, 1, 1], S strong[7, 1, 1], VS very strong[9, 1, 0]
L
VL
M
MD
Final rating for DM3
MD
Final rating for DM3
MD
MD
VL
E
Final rating for DM2 L
D
Final rating for DM1
VS
DM3 VW
VS
VS
W
VW
TR8
[64.8, 16.2, 5.94] [64.8, 16.2, 5.94] [75.8, 24.14, 26.3] [40.2, 13.58, 13.68]
S VS
DM2 W
S
DM1 VW
DM3
VS
VW
W
VW
TR7
VS
Estimated Final rating for DM1 Expenditure Final rating for DM2
Technical Difficulty
[110.6, 29.82, 20.06] [14.4, 5.38, 9] [46.2, 16.36, 17.24] [61, 18.68, 13.06]
[4.4, 0.92, 0.9]
CN7
M
VS
VS
VS
TR6
DM2
DM1
[4.2, 0.9, 0.88]
W
DM3 M
CN6
S VS
VS M
DM2 S
VS
TR5
DM1 M
[7.2, 1, 0.66]
DM3
VS VS
VW
M
DM2
S
[5.8, 0.94, 0.82] DM1
DM3
M
W
TR4
CN5
CN4
VS
DM3 VS
VS
DM2
TR3
VS
CN3
VS
DM1
DM3 VS
TR2
DM2
[1.6, 0.42, 1]
CN2
DM2 VS
DM1 VS
TR1
[2.8, 0.54, 0.96] DM1
[4.6, 0.9, 0.8]
CN1
Table 3 HOQ rating
Int J Adv Manuf Technol
Int J Adv Manuf Technol Table 4 Multi-criteria information Criteria
Preference functions
Min/Max p
Relative importance V-shape criterion (type3) Max Technical difficulty V-shape criterion (type3) Min Estimated expenditure V-shape criterion (type3) Min
wj
0.05 0.4 3 0.3 3 0.3
difficulty in the following linguistic terms (VE to VD) and measures how much complexity is associated with carrying out each TR. Very easy (VE)
Easy (E)
Moderately difficult (MD)
Difficult (D)
Very difficult (VD)
[1, 0, 2]
[3, 2, 2]
[5, 2, 2]
[7, 2, 2]
[9, 2, 0]
Finally, estimated expenditure, in the following linguistic terms, indicates the degree of the necessary expenses on labor and material (in $‘000), which were required to meet each TR. Very low (VL)
Low (L)
Modest (M)
High (H)
Very high (VH)
[2, 1.5, 1.5]
[5, 2, 2]
[9, 3, 3]
[14, 3, 3]
[20, 5, 5]
Table 3 shows the HOQ for all the DMs and the criteria. 4.5 Stage V: Set the parameter p Aside from the relationship between the CNs and TRs (i.e., relative Importance), the DMs decided that it would be more useful to consider a multi-criteria approach. Therefore, the second round of virtual communication was organized to decide on the parameter setting for the weights and type of preference function for the additional criteria of technical difficulty and estimated expenditure. A Delphi method was
used to generate consensus on the type of preference function and the weights of criteria among the DMs. All criteria except relative importance are to be minimized. Table 4 contains the parameter setting p for the type 3 preference function and the weights which were set by the DMs through a Delphi approach. 4.6 Stage VI: Conduct the fuzzy PROMETHEE GDSS algorithm With the HOQ matrix and parameter setting finalized, the fuzzy PROMETHEE GDSS algorithm was programmed and computed via MATLAB 2008 in two phases. The first phase considers up to the first five steps of the fuzzy PROMETHEE GDSS. The results of the three DMs are given in Table 5. While the DMs presented slightly different concerns and opinions for assessing and ranking the proposed alternatives, they provided similar rankings for the best alternative (TR7) and the worst alternative (TR2). In the second phase, another round of the Delphi process was initiated for the parameter p, and each DM finally agreed to a value of 0.15 or a 15 % tolerance. The DMs also agreed that all of their preferences were of equal importance in this decision-making process, which is reflected by the weights of 0.333 in Table 5. However, our approach can cater to varying importance across the DMs and threshold values. Subsequently, we compute steps 5 and 6 of the fuzzy PROMETHEE GDSS to arrive at the fuzzy global net flows for the TRs (see Table 6). To provide a ranking of the alternatives, the LR triangular fuzzy numbers were defuzzified according to step 7 of the fuzzy PROMETHEE GDSS. The defuzzified value was then used to rank the alternatives, where a higher alternative was given greater preference. The final results show that alternatives TR7, TR1, and TR3 were preferred, whereas TR2 and TR5 were considered as the
Table 5 Data matrix for group decision making DM1
DM2
DM3
Weight Max/Min Type p
0.333 Max 3 0.15
Rank
0.333 Max 3 0.15
Rank
0.333 Max 3 0.15
Rank
TR1 TR2 TR3 TR4 TR5 TR6 TR7 TR8
[0.267, 0.514, 0.503] [−0.432, 0.376, 0.407] [0.233, 0.332, 0.319] [−0.227, 0.259, 0.263] [−0.235, 0.257, 0.322] [−0.149, 0.235, 0.225] [0.271, 0.303, 0.303] [0.272, 0.351, 0.285]
2 8 4 7 6 5 1 3
[0.153, 0.367, 0.378] [−0.447, 0.438, 0.452] [0.121, 0.385, 0.336] [0.014, 0.346, 0.4] [−0.249, 0.270, 0.276] [0.177, 0.229, 0.192] [0.294, 0.419, 0.431] [−0.064, 0.326, 0.315]
3 8 4 5 7 2 1 6
[0.007, 0.210, 0.221] [−0.375, 0.299, 0.335] [0.042, 0.505, 0.529] [−0.046, 0.253, 0.264] [−0.038, 0.132, 0.124] [0.025, 0.164, 0.145] [0.296, 0.425, 0.435] [0.089, 0.234, 0.170]
5 8 3 7 6 4 1 2
Int J Adv Manuf Technol Table 6 Global ranking of alternatives Alternative
Fuzzy global net flows in LR mode
Defuzzification
Ranking
TR1 TR2 TR3 TR4
[0.147, 0.133, 0.169] [−0.396, 0.201, 0.270] [0.142, 0.205, 0.189] [−0.103, 0.050, 0.045]
0.159 −0.373 0.137 −0.105
2 8 3 6
TR5 TR6 TR7 TR8
[−0.142, 0.177, 0.210] [−0.065, 0.038, 0.015] [0.291, 0.218, 0.185] [0.084, 0.163, 0.225]
−0.131 −0.073 0.280 0.105
7 5 1 4
worst alternatives as a group (Table 6). The results also show the inconsistencies and similarities among the individual and global rankings. While individual and global evaluations agree on the best (TR7) and worst (TR2) alternatives, the results for the six other alternatives were quite different. Also, the result for the global evaluation was similar to DM1, whereas DM2 and DM3 expressed rather different concerns and preferences in the six alternative TRs to global ranking. 4.7 Comparison with fuzzy TOPSIS Table 7 presents ranking results obtained from fuzzy TOPSIS, which can be compared with the results derived earlier with the Fuzzy PROMETHEE GDSS method. The TOPSIS approach considers a preferred alternative to be one that has the closest distance to the positive ideal solution and the farthest distance from the negative ideal solution [24]. A positive ideal solution is composed of the best performance values of each criterion whereas the negative ideal solution consists of the worst performance values. The closeness coefficient determines the ranking order of alternatives by calculating the distances of alternatives to both the positive ideal and negative ideal solutions. The various applications of fuzzy TOPSIS for group decision making are discussed in Behzadian et al. [4]. This paper uses an extended fuzzy TOPSIS method that was originally
Table 7 Ranking of alternatives by fuzzy TOPSIS
Alternative
CCi
Ranking
TR1 TR2 TR3 TR4 TR5 TR6 TR7 TR8
0.685 0.323 0.654 0.419 0.431 0.517 0.761 0.588
2 8 3 7 6 5 1 4
proposed by Chen [13] for comparison purposes. In order to pool the decision makers’ opinions, the aggregated data matrix and weights for criteria are obtained from Tables 3 and 4, respectively. According to the closeness coefficient (CC) of alternatives, the ranking order of eight alternatives is determined as TR7>TR1>TR3>TR8>TR6>TR5>TR4>TR2. In this problem, there is no difference between the final solution produced by the two MCDA methods. It indicates that the ranks obtained from PROMETHEE GDSS are highly reliable for the purposes of ranking TRs. The main advantage of PROMETHEE GDSS was the rapid communication among team members. It allows each member to express their own preferences rapidly, while comparing their individual rankings with the global ranking.
5 Conclusion This article presents a fuzzy PROMETHEE GDSS method to handle the ranking and selection of the TRs in the final stage of the HOQ process for a high-end ergonomic chair manufacturer. The approach allows a facilitator to collect the relevant data from each DM located in different geographical areas and aggregate into a group decision matrix for use by the QFD team of designers. The approach also allows each criterion to have different tolerance levels contributing to the TR, as well as each DM to possess a different threshold level toward the TR. We show that applying the fuzzy PROMETHEE GDSS can prevent the loss of valuable evaluation data and overcome the difficulty of integrating subjective linguistic assessments into the group decision-making process. Future research can explore the assignment of weights for the alternatives and DMs. During the decision-making process, serious conflicts may arise from different concerns and preferences between DMTs. A further geometrical analysis for interactive assistance (GAIA) allowed the DMTs to check the consistency, independency, and similarity of preferences among the DMTs and to ensure the quality of the group decision-making process, based on the quality of information preserved. Acknowledgments Dr. Joshua Ignatius would like to express his gratitude to Universiti Sains Malaysia for partially supporting this research under the research grant no. 1001/pmaths/817060.
References 1. Akao Y (1990) Quality function deployment. Integrating customer requirements into product design. Productivity Press, Cambridge 2. Bae SM, Ha SH, Park SC (2005) A web-based system for analyzing the voices of call center customers in the service industry. Expert Syst Appl 28:29–41
Int J Adv Manuf Technol 3. Behzadian M, Kazemzadeh R, Albadvi A, Aghdasi M (2010) PROMETHEE: a comprehensive literature review on applications and methodologies. Eur J Oper Res 200(1):198–215 4. Behzadian M, Khanmohammadi Otaghsara S, Yazdani M, Ignatius J (2012) A state-of the-art survey of TOPSIS applications. Expert Syst Appl 39:13051–13069 5. Benner M,Linnemann AR, Jongen WMF, Folstar P (2003) Quality Function Deployment (QFD)—can it be used to develop food products? Food Qual Prefer 14(4):327–339 6. Bisel RU, Buyukozkan G, Ruan D (2006) A fuzzy preferenceranking model for a quality evaluation of hospital websites. Int J Intell Syst 21(11):1181–1197 7. Brans JP, Mareschal B (2005) PROMETHEE methods. In: Figueira J, Greco S, Ehrgott M, Multiple criteria decision analysis: state of the art surveys. Springer Science + Business Media, 163–196 8. Brans JP, Mareschal B, Vincke P (1984) PROMETHEE: a new family of outranking methods in MCDM, IFORS 84, North Holland. Amsterdam, 477–490 9. Buyukozkan G, Feyzioglu O (2005) Group decision making to better respond customer needs in software development. Comput Ind Eng 48(2):427–441 10. Buyukozkan G, Feyzioglu O, Rual D (2007) Fuzzy group decisionmaking to multiple preference formats in quality function deployment. Comput Ind 58(5):392–402 11. Chan LK, Wu ML (2005) A systematic approach to quality function deployment with a full illustrative example. Omega 33:119–139 12. Chen Y, Chen L (2006) A non-linear possibilistic regression approach to model functional relationships in product planning. Int J Adv Manuf Technol 28:1175–1181 13. Chen CT (2000) Extensions of the TOPSIS for group decisionmaking under fuzzy environment. Fuzzy Sets Syst 114:1–9 14. Chou WC, Lin WT, Lin CY (2007) Application of fuzzy theory and PROMETHEE technique to evaluate suitable ecotechnology method: a case study in Shihmen Reservoir Watershed, Taiwan. Ecol Eng 31(4):269–280 15. Christiano JJ, Liker JK, White CC III (2000) Customer-driven product development through quality function deployment in the US and Japan. J Prod Innov Manag 17(4):286–308 16. Cohen L (1995) Quality function deployment: inhow to make QFD work for you. Addison-Wesley, Reading 17. Fernandez-Castro AS, Jimenez M (2005) PROMETHEE: an extension through fuzzy mathematical programming. J Oper Res Soc 56(1):119–122 18. Fung RYK, Tang J, Tu PY, Chen Y (2003) Modelling of quality function deployment planning with resource allocation. Res Eng Des 14(4):247–255 19. Geldermann J, Spengler T, Rentz O (2000) Fuzzy outranking for environmental assessment. Case study: iron and steel making industry. Fuzzy Sets Syst 115(1):45–65 20. Haralambopoulos DA, Polatidis H (2003) Renewable energy projects: structuring a multicriteria group decision-making framework. Renew Energy 28(6):961–973 21. Ho ESSA, Lai YJ, Chang SI (1999) An integrated group decisionmaking approach to quality function deployment. IIE Trans 31(6): 553–567 22. Hsiao SW (2002) Concurrent design method for developing a new product. Int J Ind Ergon 29:41–55 23. Hsiao SW, Liu E (2005) A structural component-based approach for designing product family. Comput Ind 56:13–28 24. Hwang C, Yoon K (1981) Multiple attribute decision-makings: methods and application. Springer, New York 25. Ignatius J, Motlagh SMH, Sepehri MM, Behzadian M, Mustafa A (2010) Hybrid models in decision making under uncertainty:
the case of training provider evaluation. J Intell Fuzzy Syst 21(1): 147–162 26. Iranmanesh SE, Salimi MH (2003) An investigation of rank reversal when using fuzzy importance lrvrls in QFD analysis. Int J Reliab Qual Saf Eng 10(2):185–203 27. Kahraman C, Ertay T, Büyüközkan G (2006) A fuzzy optimization model for QFD planning process using analytic network approach. Eur J Oper Res 171(2):390–411 28. Karsak EE (2004) Fuzzy multiple objective decision making approach to prioritize design requirements in quality function deployment. Int J Prod Res 42(18):3957–3574 29. Kazemzadeh RB, Behzadian M, Aghdasi M, Albadvi A (2009) Integration of marketing research techniques into house of quality and product family design. Int J Adv Manuf Technol 41:1019–1033 30. Kim KJ, Moskowitz H, Dhingra A, Evans G (2000) Fuzzy multicriteria models for quality function deployment. Eur J Oper Res 121(3):504–518 31. Kwong CK, Bai H (2002) A fuzzy AHP approach to the determination of importance weighted of customer requirement in quality function deployment. J Intell Manuf 13(5):367–377 32. Lai X, Xie M, Tan KC (2005) Dynamic programming for QFD optimization. Qual Reliab Eng Int 21(8):769–780 33. Le Teno JF, Mareschal B (1998) An interval version of PROMETHEE for the comparison of building products’design with ill-defined data on environmental quality. Eur J Oper Res 109(2):522–529 34. Leyya-Lopez JC, Fernandez-Gonzalez E (2003) A new method for group decision support based on ELECTRE III methodology. Eur J Oper Res 148(1):14–27 35. Liu CH, Wu HH (2008) A fuzzy group decision-making approach in quality function deployment. Qual Quant 42(4):527–540 36. Morais DC, De Almeida AT (2007) Group decision-making for leakage management strategy of water network. Resour Conserv Recycl 52(2):441–459 37. Mayyas A, Shen Q, Mayyas A, Abdelhamid M, Shan D, Qattawi A, Omar M (2011) Using quality function deployment and analytical hierarchy process for material selection of body-in-white. Mater Des, 2771–2782 38. Raharjo H, Xie M, Brombacher AC (2006) Prioritizing quality characteristics in dynamic quality function deployment. Int J Prod Res 44(23):5005–5018 39. Raju KS, Duckstein L, Arondel C (2000) Multicriterion analysis for sustainable water resources planning: a case study in Spain. Water Resour Manag 14(6):435–456 40. Reich Y, Levy E (2004) Managing product design quality under resource constraints. Int J Prod Res 42(13):2555–2572 41. Roychowdhury S, Pedrycz W (2001) A survey of defuzzification strategies. Int J Intell Syst 16(6):679–695 42. Shen XX, Tan KC, Xie M (2000) An integrated approach to innovative product development using Kano_s model and QFD. Eur J Innov Manag 3(2):91–99 43. Tuzkaya G, Gulsun B, Kahraman C, Ozgen D (2010) An integrated fuzzy multi-criteria decision making methodology for material equipment selection problem and an application. Expert Syst Appl 37(4): 2853–2863 44. Vranegl S, Stanojevi M, Stevanovi V, Lui M (1996) INVEX: investment advisory expert system. Expert Syst 13(2):105–119 45. Wu HH, Liao AYH, Wang PC (2005) Using grey theory in quality function deployment to analyse dynamic customer requirements. Int J Adv Manuf Technol 25(11):1241–1247 46. Xie M, Tan KC, Goh TN (2003) Advanced QFD application. ASQ Quality Press, Milwaulee 47. Yan W, Khoo LP, Chen CH (2005) A QFD-enabled product conceptualisation approach via design knowledge hierarchy and RCE neural network. Knowl-Based Syst 18(6):279–293