RECYCL-2546; No of Pages 17 Resources, Conservation and Recycling xxx (2012) xxx xxx Contents lists available at SciVerse ScienceDirect Resources, Conservation and Recycling ourna l h o me pa ge: wwwelseviercom/locate/resconrec The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR C-H Hsu a, Fu-Kwun Wang a,, Gwo-Hshiung Tzeng b,c a Department of Industrial Management, National Taiwan University of Science and Technology; No 43, Sec 4, Keelung Rd, Taipei 106, Taiwan b Department of Business and Entrepreneurial Management, and Institute of Proect Management, Kainan University; No 1, Kainan Road, Luchu, Taoyuan 338, Taiwan c Institute of Management of Technology, National Chiao Tung University; 1001, Ta-Hsueh Road, Hsinchu 300, Taiwan a r t i c l e i n f o Keywords: Vendor selection (VS) Recycled materials Multiple criteria decision making (MCDM) Analytic network process (ANP) Supply chain management a b s t r a c t Environmentally conscious manufacturing and product recovery (ECMPRO) has become an obligation of manufacturers, and it has been extended to be the policy and strategy of businesses Producing recyclable products and using recycled materials are optimal strategies for ECMPRO Vendor selection (VS) is one of the multiple criteria decision-making (MCDM) problems in strategic supply chain management The purpose of this article is to propose how the best selection to conduct the recycled materials can be implemented for enhancing and increasing the efficiency of using resources in the manufacturing process through recycled materials VS Aluminum composite panel (ACP) is a global product, and ACP companies in Taiwan use recycled materials in more than 80% for their products on a quantity basis Therefore, we selected the ACP industry of Taiwan as an empirical model to study VS and to reveal methods of improving gaps in each criterion for achieving the aspired levels of performance We use the MCDM model combining DEMATEL-based on ANP (called DANP) with VIKOR to solve the recycled materials VS problems of multiple dimensions and criteria that are interdependent, instead of the independent assumption of an analytic hierarchy process, for mimicking the real-world scenario 2012 Elsevier BV All rights reserved 1 Introduction Environmentally conscious manufacturing is concerned with developing methods for manufacturing new products from conceptual design to final delivery and ultimately to the end-of-life (EOL) disposal such that environmental standards and requirements are satisfied Conversely, product recovery aims to minimize the amount of waste sent to landfills by recovering materials and parts from old or outdated products through recycling and remanufacturing (including reuse of parts and products) (Gungor and Gupta, 1999) The increasing interest in product reuse originates not only from the reinforcement of environmental awareness legislation but also from the fact that the engagement in reuse activities has been proven profitable in many industries (Kannan et al, 2009) So, suppliers face increasing pressure from their customers to improve their environmental performance (Delmas and Montiel, 2009) Mena et al (2011) identified the main root causes of food Corresponding author Tel: +886 2 27376325; fax: +886 2 27376344 E-mail addresses: etctw1234@gmailcom (C-H Hsu), fukwun@mailntustedutw (F-K Wang), ghtzeng@ccnctuedutw, ghtzeng@mailknuedutw (G-H Tzeng) waste in the supplier-retailer interface and compared practices in the UK and Spain Zhang et al (2012) analyzed the demands, possibilities, difficulties and suggestions for waste cooking oil recycling in China For these reasons, green manufacturing, that is, making environmentally sound products through efficient processes, can be good for business and is a current trend in business around the world (Melnyk et al, 2001; Venus, 2011) In the automotive industry, most companies are putting the ability to recycle parts on the same level as safety, fuel economy, and costs when they design new vehicles The 15-nation European Union is considering a rule that would require 85% of a car by weight to be recycled or remanufactured This would increase to 95% by 2015 The source of recycled material is post consumer waste (PCW), of which paper, metal, glass, and plastics are the largest categories (Field and Sroufe, 2007) Olugu et al (2011) developed a set of measures for evaluating the performance of the automobile green supply chain According to a long-term study in the US during 1960 to 1996, the amount of plastics consumed annually have been growing steadily from 05% to 123%, by weight of municipal solid waste (Subramanian, 2000) And the polyethylene, including high density polyethylene and low density polyethylene, forms the largest fraction of plastics in municipal solid waste about 49% For example, aluminum composite panel (ACP) is a multi-layer sheet that is 0921-3449/$ see front matter 2012 Elsevier BV All rights reserved doi:101016/resconrec201202009 Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
G Model RECYCL-2546; No of Pages 17 2 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx Fig 1 The composition structure of ACP produced by laminating two pre-treated and coated aluminum with a fire resistant mineral-filled or polyethylene (PE) core under a continuous high pressure and heat process (Fig 1) ACP is a construction material that has been applied to the decoration of buildings both inside and outside of the wall ACP can be 100% recycled for both aluminum and plastic materials at its EOL To introduce the ACP recycling loop, we modified the product life cycle, recycling states and activities figure of Chen et al (1993) Fig 2 shows the general stages and the associated activities of ACP recycling Although the primary recyclable ACP components include aluminum and plastics, this study focus on recycled plastics, such as, low-density PE (LDPE) and high-density of PE (HDPE) As we know, one of the competencies essential to supply chain success is an effective purchasing function (Cakravastia and Takahashi, 2004; Giunipero and Brand, 1996) Vendor selection (VS), the first step of purchasing function, has a very important role in the supply chain of manufacturing companies Therefore, the purpose of this article is to enhance and increase the efficiency of using resources in the manufacturing process through recycled materials vendor selection (VS) As the demand for environmentally friendly products has grown, the technology for converting PCW into new products has improved, and more recycling programs have been implemented As a result, the demand for recycled Fig 2 ACP life cycle recycling stages and activities materials and the availability and variety of products with recycled content continues to increase (Field and Sroufe, 2007) For example, according to production records of Taiwan s aluminum composite panel (ACP) manufactures they used recycled plastics in their products has grown from 0% to 80% on a quantity basis in the past 10 years Chinese ACP manufacturers use an even higher ratio of recycled plastics Moreover, this tendency will continue to increase following improvements in environmental management systems and recycling technology in the future This is the essential reason for our focus on VS, as regarding recycled materials in the ACP industry expect to support acquisition by companies of environmentally friendly materials with stable quality and quantity, reasonable cost, on-time delivery, and good service As VS is a type of multiple criteria decision-making (MCDM) problem, we propose a hybrid MCDM model combining a decision-making trial and evaluation laboratory (DEMATEL) method with an analytic network process (ANP) and VIseKriteriumska Optimizacia I Kompromisno Resene (VIKOR; translates into multicriteria optimization and compromise solution) method in this study The DEMATEL method (Fontela and Gabus, 1976) was designed to determine the degree of influence of the VS criteria and apply them to normalize the unweighted supermatrix in the ANP The ANP is an extension of the analytic hierarchy process (AHP); indeed, it is the general form of the AHP The ANP handles dependence within a cluster (inner dependence) and among different clusters (outer dependence) The ANP is a nonlinear structure, whereas the AHP is hierarchical and linear with goals at the top and alternatives at lower levels (Saaty, 1999) The ANP has been used successfully in many practical decision-making problems, such as the proect selection, supply chain management, and optimal scheduling problems (Lee and Kim, 2000; Meade and Presley, 2002; Momoh and Zhu, 2003; Sarkis, 2003) A hybrid model combining DEMATEL and ANP (we call this model DEMATEL-based ANP; DANP) has been widely applied to solve a variety of applications in solving MCDM problems, such as e-learning evaluations (Tzeng et al, 2007), airline safety measurements (Liou et al, 2007), and innovation policy portfolios for Taiwan s silicon/semiconductor intellectual property (SIP) Mall (Huang et al, 2007) Strategic management decisions influence the relative importance of the various criteria in the VS process (Weber et al, 2000) The maority of VS models in existing publications ignored the fact that evaluation criteria must be aligned with a firm s environmental strategies (Chou et al, 2007) As mentioned in the previous paragraph, this work is different from previously research in three ways First, we aligned the criteria with the firm s strategy of environmentally conscious green manufacturing to use recycled materials with VS dimensions and criteria Second, we adopted a hybrid MCDM model of DANP to evaluate and improve the performance of vendor s dimensions and criteria, which are interdependent for achieving the best alternative in VS Finally, we combined DANP with VIKOR to evaluate/improve the Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
RECYCL-2546; No of Pages 17 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx 3 aspired level of vendor performance based on the network relation map (NRM) by the DEMATEL technique The best vendor selection for conducting the recycled materials can be achieved The rest of this research is organized as follows In the next section, the relevant literature about vendor selection criteria is reviewed A hybrid MCDM model for solution methodology is developed in Section 3 In Section 4, an empirical case is illustrated to calculate and show the proposed hybrid MCDM model In Section 5, we use the results of Section 4 to discuss and highlight the managerial implications based on the case analysis Finally, conclusions are drawn in Section 6 2 Vendor selection criteria VS has a very important role in the supply chain of manufacturing companies Regarding VS in SCM, many published studies described their evaluation criteria and methods (Chan and Kumar, 2007; Chou et al, 2007; Foster and Ogden, 2008; Ghodsypour and O Brien, 1998, 2001; Govindan et al, 2010; Masella and Rangone, 2000; Shen and Yu, 2009; Yang et al, 2008) But we did not come across any papers discussing VS for conducting the recycled materials in the literature VS decisions are complicated by the fact that various criteria must be considered in the decision-making process The criteria used may vary across different product categories and purchase situations (Shen and Yu, 2009) There may not be a generalized consensus on how to identify suitable criteria because these decisions are highly firm-and situation-specific (Chou et al, 2007; Liu and Hai, 2005; Schmitz and Platts, 2004) With reference to past literature, it can be observed that there has only been limited discussion of virgin material VS dimensions and criteria, which in fact is insufficient for VS for conducting the recycled materials To align the company strategy of environmentally conscious green manufacturing with the use of recycled materials, we must consider critical criteria that are connected with green manufacturing in our study Vachon (2007) described the concept of green supply chain practices as two sets of related yet independent environmental activities: environmental collaboration and environmental monitoring Environmental collaboration can be defined as the planning and development of environmental activities and proects that require direct involvement of an organization, whether with its suppliers or its customers, to ointly develop environmental solutions (Geffen and Rothenberg, 2000; Rao, 2002) According to the above VS criteria review, we modified Chan and Kumar s (2007) VS dimensions which included the overall cost of the product, quality of the product, service performance of supplier, supply risk, delivery, as well as consideration of green supply chain practices for environmental collaboration The problem we studied here has four levels of hierarchy, and different decision dimensions and criteria will be further discussed in Table 1 The overall obective is selecting the best recycled materials vendor for an ACP manufacturing company We denote above dimensions, criteria, and alternatives by D i (i = 1, 2,, 6), C ( = 1, 2,, 17), and V k (k = 1, 2,, n) to form a hierarchy of vendor selection in recycled materials in Fig 3 Based on the literature review, we found the following methods are using recently for evaluation or development suppliers in supply chain management: (1) linear weighting models (Barbarosoglu and Yazgaç, 1997); (2) mathematical programming models (Weber et al, 1991); (3) statistical models (Ronen and Trietsch, 1988; Soukup, 1987); (4) artificial intelligence models (Albino and Gravel, 1998); (5) fuzzy extended AHP (analytic hierarchy process) approach (Chan and Kumar, 2007; Chou et al, 2007; Shen and Yu, 2009); (6) ISM and TOPSIS (Kannan et al, 2009); (7) ISM and fuzzy integral (Yang et al, 2008) It is quite clear that few articles were carried out on the selection of recycled material vendors using the hybrid MCDM model combining with DANP and VIKOR Therefore, we proposed to use DEMATEL combine ANP to determine the degrees of influence among the criteria and VIKOR method for calculating the compromise ranking and gap of the alternatives for alternative improvement 3 Development of solution methodology VS is one of MCDM problems, as any criterion may be interinfluenced, the DEMATEL technique permits us to know the influence structure between the criteria and try to find problems that can be improved DEMATEL technique combined with the ANP method to find the most important criterion that will help to improve VS performance To understand the gap of each criterion and to rank the first important strategy to implement, the VIKOR method will be leveraged for calculating the compromise ranking and gap of the alternatives for alternative improvement In short, the framework of evaluation contains three main phases: (1) constructing the NRM among criteria by the DEMATEL technique, (2) calculating the weights of each criterion by combining the ANP based on the NRM, and (3) ranking and improving the priorities of alternative vendors through the VIKOR The process of this hybrid MCDM model is briefly illustrated in Fig 4 31 The DEMATEL technique for developing NRM The DEMATEL technique has been successfully applied in many situations, such as marketing strategies, e-learning evaluations, control systems and safety problems (Chiu et al, 2006), information security (Ou Yang et al, 2009), financial stock investment (Lee et al, 2009), water resources and environment (Chen et al, 2010), industry technology (Lin and Tzeng, 2009; Lin et al, 2010a c), and portfolio selection based on CAPM (Ho et al, 2011) The methodology can confirm interdependence among variables/criteria and restrict the relationships that reflect characteristics within an essential systemic and developmental trend The method can be summarized as follows (Liou et al, 2007, 2008): Step 1: Calculate the initial average matrix by scores In this step, respondents are asked to indicate the degree of direct influence each factor/element i exerts on each factor/element, as indicated by a i, using an integer scale ranging from 0 to 4 (going from no influence (0), to very high influence (4) ) From any group of direct matrices of respondents, it is possible for experts to derive an average matrix A = [a i ] n n, with each element being the mean of the same elements in the various direct matrices of the respondents The average matrix A is represented as shown in Eq (1) a 11 a 1 a 1n A = a i1 a i a in (1) a n1 a n a nn Step 2: Calculate the initial influence matrix The initial influence matrix X = [x i ] n n is obtained by normalizing the average matrix A (shown by degree, ie, shown by membership and 0 x i < 1, also called the fuzzy cognitive matrix ), in which all principal diagonal elements equal zero Based on X, the initial effect that an element exerts and receives from another is shown The map portrays a contextual relationship among the elements of a system, in which the numeral represents the strength of influence (affected degree) Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
G Model RECYCL-2546; No of Pages 17 4 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx Table 1 The influence dimensions and criteria of comprehensive VS in recycled materials Dimensions Influence criteria Statements of influence criteria Quality of the product (D 1) Ingredient consistency (C 1) The ingredient consistency indicates that the source of recycled materials should be one type of PCW Process capability (C 2) Process capability describes the remanufacturing capability of the vendor s production line Yield rate (C 3) The yield rate of products means the fraction of good products used in the recycled materials in production line Delivery schedule (D 2) Shortest lead time (C 4) Lead time is the prepare time, include delivery time, for the vendor to supply recycled materials to their purchasers Delivery on time rate (C 5) The fraction of the vendor s on time delivery of the total shipments in last one year Serious delivery delay rate (C 6) Supply risk (D 3) Geographical location (C 7) Overall cost of the product (D 4) A long time delays that affect the manufacturer s ability to deliver products to customers in a timely advantage Increasing distance between a vendor and a purchaser increase the risk of delivery delay, as there is an increased risk of issue with long distance transportation Political stability (C 8) For example, a strike by workers in public utilities and transportation will cause maor supply risks for the vendor and purchaser Equipment capacity change (C 9) Recycled material price (C 10) Equipment capacity change indicates that a vendor has sufficient equipment capacity to meet any change in supply situation change The bulk of the cost of ACP is recycled material price, which depends on the ingredient consistency of the recycled materials Handling cost (C 11) The handling cost includes the replenishment costs per unit time and the costs of carrying inventory over a unit time period Process loss cost (C 12) In processing recycled materials, the quality of the material affects the process loss cost much more than the aforementioned causes Services (D 5) Response to demand (C 13) The fast and effective response to change in demand by the vendor is very important Environmental collaboration (D 6) Information acquisition (C 14) After- sales service (C 15) Technology for recycling products and process (C 16) Green manufacturing policy (C 17) Recycled materials users require the marketing information about recycled materials and the situations of their competitors, when devising materials purchasing and products sales strategies There are frequent quality issues regarding recycled plastics materials Therefore, vendor s warranties and claim polices for after-sales service are important criteria for VS A qualified vendor for recycled materials should have the proper technology for recycling products and process, as more recycling technology will stabilize product quality and decrease resource waste in manufacturing A green manufacturing policy is making environmentally friendly products at the design stage and through efficient processes Step 3: Derive the full direct/indirect influence matrix A continuous decrease of the indirect effects of problems can be determined along the powers of X, eg, X 2, X 3,, X h and lim X h = [0] n n, h where X = [x i ] n n, 0 x i < 1 and 0 ix i 1 or 0 x i 1 and at least one column or one row of summation, but not all, equals one If the (i, ) element of matrix A is denoted by a i, the matrix X can be obtained through Eqs (2) and (3), in which all principal diagonal elements are equal to zero X = z A (2) { } 1 1 where z = min n max 1 i n =1 a, n i max 1 i n i=1 a (3) i and lim X h = [0] n n, 0 x i 1 h Step 4: Attainning the total-influence matrix T The total-influence matrix can be obtained through Eq (4), in which I denotes the identity matrix T = X + X 2 + + X h = X(I X) 1 when lim X h = [0] n n (4) h Explanation T = X + X 2 + + X h = X(I + X + X 2 + + X h 1 )(I X)(I X) 1 then, = X(I X h )(I X) 1 T = X(I X) 1, when h Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
RECYCL-2546; No of Pages 17 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx 5 Fig 3 The hierarchy of vendor selection in recycled materials If we define the sum of the rows and the sum of the columns separately expressed as vector r and vector s within the total-influence matrix T through Eqs (5) (6) then T = [t i ], i, = 1, 2,, n, (5) [ n n ] r = [r i ] n 1 = t i, s = [s ] n 1 = t i (6) =1 i=1 n 1 1 n where the superscript denotes transpose If r i denotes the row sum of the ith row in matrix T, then r i shows the sum of direct and indirect effects of factor i on the other factors/criteria If s denotes the column sum of the th column of matrix T, then s shows the sum of direct and indirect effects that factor has received from the other factors Furthermore, when = i (ie the sum of the row and column aggregates), (r i + s i ) provides an index of the strength of influences given and received, that is, (r i + s i ) shows the degree that the factor i plays in the problem In addition, the difference (r i s i ) shows the net effect that factor i contribute to the problem If (r i s i ) is positive, then factor i is affecting other factors, and if (r i s i ) Fig 4 The process of a hybrid MCDM model combined DANP and VIKOR Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
G Model RECYCL-2546; No of Pages 17 6 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx is negative, then factor i is being influenced by other factors (Tzeng et al, 2007) 32 Combining DEMATEL and ANP to find the evaluation weights ANP was published by Saaty (1999); its purpose was to solve the problems of interdependence and feedback between criteria and alternatives in the real world ANP is a mathematical theory that can systematically overcome all types of dependences In ANP procedures, the initial step is to compare the criteria in the entire system to form an unweighted supermatrix by pairwise comparisons Then, the weighted supermatrix is derived by transforming each column to sum exactly to unity (100) Each element in a column is divided by the number of clusters, and thus each column will sum to unity exactly However, using the assumption of equal weight for each cluster to obtain the weighted supermatrix appears irrational because there are different degrees of influence among the criteria (Ou Yang et al, 2008) Thus, we adopted the DEMATEL technique to determine the degrees of influence of these criteria and apply these to normalize the unweighted supermatrix in the ANP to mimic the situation in the real world We named this improved ANP as DANP The improved ANP is divided into the steps as follows: T c 11 = t 11 c11 /d11 t 11 c1 c1 /d11 t 11 /d 11 c1 c1m 1 c1 t 11 ci1 /d11 ci t 11 ci /d11 ci t 11 /d 11 cim 1 ci t 11 cm 1 1 /d11 cm 1 t 11 cm 1 /d11 cm 1 tcm 11 1 m 1 /dcm 11 1 t 11 c11 t 11 c1 t 11 ci1 t 11 ci t 11 cm 1 1 t 11 cm 1 t 11 c1m 1 t 11 cim 1 t 11 cm 1 m 1 = (10) Let the total-influence matrix match and fill into the interdependence clusters It will be called an unweighted supermatrix as shown as Eq (11), which is based on transposing the normalized influence matrix T c by dimensions (clusters), ie W = (T c ) Step 1: Develop an unweighted supermatrix The total-influenced matrix will be obtained from DEMATEL Each column will sum for normalization We call the total-influenced matrix T c = [t i ] n n obtained by criteria and T D = [t D i ] obtained by dimensions m m (clusters) from T c Then, we normalize the supermatrix T c for the ANP weights of dimensions (clusters) by using influence matrix T D Each column will sum for normalization (11) If the matrix W 11 is blank or 0 as shown as Eq (12), this means that the matrix between the clusters or criteria is independent and with no interdependence, and the other W nn value are as above After normalizing the total-influence matrix T c by dimensions (clusters), we will obtain a new matrix T c as shown as Eq (8) (7) W 11 = c 11 c 1 c 1m1 c 11 c 1i c 1m1 t 11 c11 t 11 ci1 t 11 cm 1 1 t c t 11 ci t 11 cm 1 t 11 c1m 1 t 11 cim 1 t 11 cm 1 m 1 (12) Step 2 For obtaining the weighted supermatrix, each column will sum for normalization as show in Eq (13) In addition, an explanation for the normalization T 11 c is shown as Eqs (9) (10), and other T nm c values are as above d 11 ci = m 1 =1 (8) t 11 ci, i = 1, 2,, m 1 (9) t 11 t 1 t 1n D D D T D = t i1 t i t in D D D t n1 t n t nn D D D (13) Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
RECYCL-2546; No of Pages 17 We normalized the total-influence matrix T D, and obtained a new matrix T D, as shown as Eq (14) (where t i D = t i D /d i) t 11 D /d 1 t 1 D /d 1 t 1n D /d 1 T D = t i1 D /d i t i D /d i t in D /d i = t n1 D /d n t n D /d n t nn D /d n t 11 t 1 t 1n D D D t i1 t i t in D D D (14) t n1 t n t nn D D D Let the normalized total-influence matrix T D fill into the unweighted supermatrix to obtain the weighted supermatrix W = T D W = t 11 W 11 t i1 D D W i1 t n1 W n1 D t 1 D W 1 t i D W i t n D W n (15) t 1n W 1n t in D D W in t nn W nn D Step 3 Limit the weighted supermatrix Limit the weighted supermatrix by raising it to a sufficiently large power k, until the supermatrix has converged and become a long-term stable supermatrix to obtain the global priority vectors, called DANP (DEMATEL-based ANP) influential weights, such as lim g (W ) g, where g represents any number of power In brief, the overall weights are calculated by using the above steps to derive a stable limiting supermatrix Therefore, a hybrid model combining the DEMATEL method with ANP methods can deal with the problems of interdependence and feedback 33 The VIKOR method for ranking and improving the alternatives Opricovic and Tzeng (2004) proposed the compromise ranking method (VIKOR) as one applicable technique to implement within MCDM Suppose that the feasible alternatives are represented by V 1, V 2,, V k,, V m The performance scores of alternative V k and the th criterion is denoted by f k ; w is the influential weight (relative importance) of the th criterion, where = 1, 2,, n, and n is the number of criteria Development of the VIKOR method began with the following form of L p metric (Ho et al, 2011): n L p k = w ) ( f 1/p f k (16) =1 ( f f ) where 1 p ; k = 1, 2,, m and influential weight w is derived from the ANP To formulate the ranking and gap measure L p=1 (as k S k ) and L p= (as Q k k ) are used by VIKOR (Tzeng et al, 2002, 2005; Opricovic and Tzeng, 2002, 2004, 2007) n [ ] w S k = L p=1 ( f f k ) = (17) k =1 ( f f ) C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx 7 Q k = L p= = max k { ( f f k ) ( f f ) = 1, 2,, n } (18) The compromise solution min k L p k showed the synthesized gap to be minimized, and it will be selected so that its value will be the closest to the aspired level In addition, the group utility is emphasized when p is small (such as p = 1); on the contrary, if p tends to become infinite, the individual maximal regrets/gaps obtain more importance in prior improvement (Freimer and Yu, 1976) in each dimension/criterion Consequently, min k S k stresses the maximum group utility; however, min k Q k accents on the selecting the minimum from the maximum individual regrets/gaps for shown priority improvement The compromise-ranking algorithm VIKOR has four steps according to the above mentioned factors: Step 1: Obtain an aspired or tolerable level We calculated the best f values (aspired level) and the worst f values (tolerable level) of all criterion functions, = 1, 2,, n Suppose the th function denotes benefits: f = max k f k and f = min k f k or these values can be set by decision makers (ie f is the aspired level and f is the worst value) Furthermore, an original rating matrix can be converted into a normalized weight-rating matrix by using the equation ) ( f f k r k = ( f f ) (19) Step 2 Calculate the mean of group utility and maximal regret The values can be computed by S k = n =1 w r k (the synthesized gap for all criteria) and Q k = max { rk = 1, 2,, n } (the maximal gap in k criterion for priority improvement), respectively Step 3 Calculate the index value The value can be counted by R k = v(s k S )/(S S ) + (1 v)(q k Q )/(Q Q ), where k = 1, 2, m, S * = min i S i or S* = 0 (when all criteria have been achieved to the aspired level) and S = max i S i or S = 1 (when the worst situation); Q* = min i Q i or setting Q* = 0 and Q = max i Q i or setting Q = 1, and v is presented as the weight of the strategy of the maximum group utility Conversely, 1 v is the weight of individual regret Therefore, we also can re-write R k = vs k + (1 v)q k, when S* = 0, S = 1, Q* = 0 and Q = 1 Step 4 Rank or improve the alternatives for a compromise solution Order alternatives decreasingly by the values of S k, Q k and R k Propose as a compromise solution the alternatives V (1), V (2),, V (M) The compromise-ranking method (VIKOR) is applied to determine the compromise solution and the solution is adaptable for decision-makers in that it offers a maximum group utility of the maority (shown by min S), and a maximal regret of minimum individuals of the opponent (shown by min Q) This model utilizes the DEMATEL and ANP processes in Sections 31 and 32 to obtain the influential weights of criteria with dependence and feedback and employs the VIKOR method to acquire the compromise solution 4 Application of the model to empirical case In this section, an empirical study is displayed to illustrate the application of the proposed model to evaluate and find the best vendor for conducting the recycled materials in real world case Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
G Model RECYCL-2546; No of Pages 17 8 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx 41 Background and problem descriptions As an example case, Corporation G has dedicated its efforts since 1966 to develop ACP products for exterior and interior decoration in Taiwan Corporation G is one of the pioneers of ACP manufacturing in Asia The products of Corporation G include fire resistant exterior cladding solution and embossed interior ACP cladding solution To reduce the cost of its products and to be environmentally friendly, Corporation G changed its production line process to use recycled materials instead of virgin raw materials in 1998 Because recycled materials suppliers and vendors could not manage or control the quality of their products consistently and deliver materials on schedule, Corporation G faced critical problems regarding the inconsistent quality of its products and overwhelmed process capability Three alternatives vendors, V 1, V 2, and V 3, could supply the recycled materials to Corporation G We evaluated and improved these vendors and then selected the best one by using the hybrid MCDM model combining DANP with VIKOR as follows 42 Data collection To assess the inter-influence of VS criteria for the DEMETEL calculation, we designed a questionnaire to collect data from experts in the ACP industry (see Appendices A and B) These experts were the vice president, corporation general manager, plant assistant general manager, R&D manager, purchase manager, vice plant manager, and section managers 43 Measuring relationships among dimensions and criteria by DEMATEL In this study, we adopted a DEMATEL decision-making structure and analyzed 6 dimensions of 17 criteria as well as the impact of mutual relationships The ACP experts were thus asked to determine the influential importance of the relationships among the dimensions and criteria The averaged initial direct-relationship 17 17 matrix A (Table 2) was obtained by pairwise comparisons in terms of influences and directions between criteria As matrix A shows, the normalized direct-influence matrix X (Table 3) was calculated from Eqs (1) (3) Then, using Eq (4), the total influence T c (Table 4) and T D (Table 5) were derived, and by using Eq (6), the NRM was constructed by the r and s in the total direct-influence matrix T c and T D (Table 6) as shown in Figs 5 and 6 44 Weighting of each criterion by combining the DEMATEL methods with ANP methods (DANP Technique) In this research, we combined the DEMATEL technique with the ANP method to solve VS problem, and this combination was used to obtain the normalized matrix T c We first normalized the total-influence matrix T By calculating the limiting power of the weighted supermatrix, lim (W ) g is applied until a steady-state g condition is reached (Tables 7 11) By evaluating the VS criteria of Corporation G according to the DEMATEL process, we obtained dynamic relationships between the construction of an important degree of influence-unweighted supermatrix (Table 8), and in accordance with the extent of the impact of various criteria, we achieved a weighted supermatrix (Table 10) Finally, the limit of the supermatrix (Table 11) to confirm the supermatrix has been converged and become a long-term stable supermatrix and to obtain the global and local weights of all criteria and their ranks, as shown in Table 12 Table 2 The initial influence matrix A for criteria Criteria C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16 C17 Sum C1 0 3375 3375 2 2625 275 15 125 3125 275 225 35 2125 2625 2125 3125 3 42 C2 35 0 3375 225 3 2875 1375 125 325 225 2375 35 2375 275 2 25 25 41 C3 3125 3125 0 2125 2875 2375 1125 1 325 2125 225 35 225 25 1875 2625 2375 39 C4 175 1875 2375 0 3375 3125 275 2 2875 2125 2 2 25 2375 175 1375 15 36 C5 175 225 275 275 0 2625 275 1875 275 2 2125 2 25 2375 2125 175 1125 36 C6 2 2125 2875 225 3625 0 2625 2375 2625 1875 2 25 2625 25 25 2125 1875 39 C7 1 075 0625 2375 3125 3 0 275 1125 2125 2125 1875 2625 1625 225 1875 1875 31 C8 0625 05 05 225 2625 2875 3125 0 125 2 1625 175 15 125 2125 125 1125 26 C9 25 275 3 25 3 2625 125 1125 0 225 175 25 3 275 2125 2375 2 38 C10 325 3125 25 2375 2 225 2375 175 2 0 3 25 2375 175 1625 2375 175 37 C11 2 2125 1875 2125 1625 225 1625 1375 1625 2625 0 2125 1875 15 175 225 2 31 C12 2625 2375 2125 2 1625 175 1375 075 175 2875 2125 0 175 175 2 225 2 31 C13 175 2125 2125 2 2125 2375 1875 125 2125 1875 1875 2 0 2125 225 1875 1375 31 C14 225 2625 2875 2875 3 3125 0875 0875 25 2375 1875 2 225 0 2625 2375 15 36 C15 2375 225 2625 225 2375 225 1625 1125 175 1875 1875 2375 2875 25 0 1875 175 34 C16 275 2875 225 125 175 1625 15 1125 25 1875 2 2 2375 1875 2625 0 3 33 C17 25 275 1375 125 1375 1125 15 1125 15 2125 1875 175 2 125 1625 2 0 27 Sum 3575 37 3663 3463 4013 39 2925 23 36 3513 3313 3788 37 335 3338 34 3075 42 Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
RECYCL-2546; No of Pages 17 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx 9 Table 3 The normalized direct-influence matrix X for criteria Criteria C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 11 C 12 C 13 C 14 C 15 C 16 C 17 C 1 0 0081 0081 0 0063 0066 0036 003 0075 0066 0054 0084 0051 0063 0051 0075 0072 C 2 0084 0 0081 0054 0072 0069 0033 003 0078 0054 0057 0084 0057 0066 0048 006 006 C 3 0075 0075 0 0051 0069 0057 0027 0024 0078 0051 0054 0084 0054 006 0045 0063 0057 C 4 0042 0045 0057 0 0081 0075 0066 0048 0069 0051 0048 0048 006 0057 0042 0033 0036 C 5 0042 0054 0066 0066 0 0063 0066 0045 0066 0048 0051 0048 006 0057 0051 0042 0027 C 6 0048 0051 0069 0054 0087 0 0063 0057 0063 0045 0048 006 0063 006 006 0051 0045 C 7 0024 0018 0015 0057 0075 0072 0 0066 0027 0051 0051 0045 0063 0039 0054 0045 0045 C 8 0015 0012 0012 0054 0063 0069 0075 0 003 0048 0039 0042 0036 003 0051 003 0027 C 9 006 0066 0072 006 0072 0063 003 0027 0 0054 0042 006 0072 0066 0051 0057 0048 C 10 0078 0075 006 0057 0048 0054 0057 0042 0048 0 0072 006 0057 0042 0039 0057 0042 C 11 0048 0051 0045 0051 0039 0054 0039 0033 0039 0063 0 0051 0045 0036 0042 0054 0048 C 12 0063 0057 0051 0048 0039 0042 0033 0018 0042 0069 0051 0 0042 0042 0048 0054 0048 C 13 0042 0051 0051 0048 0051 0057 0045 003 0051 0045 0045 0048 0 0051 0054 0045 0033 C 14 0054 0063 0069 0069 0072 0075 0021 0021 006 0057 0045 0048 0054 0 0063 0057 0036 C 15 0057 0054 0063 0054 0057 0054 0039 0027 0042 0045 0045 0057 0069 006 0 0045 0042 C 16 0066 0069 0054 003 0042 0039 0036 0027 006 0045 0048 0048 0057 0045 0063 0 0072 C 17 006 0066 0033 003 0033 0027 0036 0027 0036 0051 0045 0042 0048 003 0039 0048 0 Table 4 The total influence matrix T c for criteria Criteria C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 11 C 12 C 13 C 14 C 15 C 16 C 17 r C 1 0325 0411 0410 0352 0411 0402 0291 0233 0396 0373 0347 0417 0377 0361 0344 0375 0344 1145 C 2 0400 0333 0408 0356 0418 0404 0287 0232 0397 0361 0347 0415 0380 0363 0340 0360 0331 1141 C 3 0373 0384 0313 0335 0394 0373 0266 0214 0378 0340 0327 0395 0358 0340 0320 0345 0313 1070 C 4 0316 0329 0340 0266 0381 0366 0286 0225 0344 0316 0300 0336 0340 0314 0296 0294 0271 1013 C 5 0315 0336 0347 0326 0304 0354 0284 0220 0341 0312 0301 0335 0339 0313 0303 0301 0262 0985 C 6 0341 0354 0370 0335 0406 0315 0298 0244 0358 0329 0317 0367 0362 0334 0329 0328 0295 1056 C 7 0257 0261 0258 0280 0330 0320 0194 0216 0263 0277 0265 0289 0301 0258 0270 0265 0243 0673 C 8 0215 0220 0220 0246 0283 0282 0237 0133 0232 0241 0224 0251 0243 0219 0236 0219 0198 0602 C 9 0351 0367 0372 0336 0389 0371 0264 0213 0297 0334 0309 0365 0367 0338 0318 0332 0297 0774 C 10 0361 0368 0354 0328 0361 0357 0284 0224 0336 0278 0332 0360 0347 0310 0302 0327 0288 0970 C 11 0287 0298 0291 0277 0301 0306 0231 0186 0280 0292 0221 0301 0288 0260 0261 0279 0252 0813 C 12 0307 0310 0303 0279 0306 0301 0228 0174 0289 0302 0274 0258 0291 0271 0271 0284 0257 0835 C 13 0285 0301 0301 0279 0317 0314 0239 0185 0295 0279 0267 0302 0249 0278 0276 0274 0241 0803 C 14 0337 0355 0361 0336 0380 0372 0249 0202 0345 0328 0304 0345 0342 0268 0320 0323 0278 0930 C 15 0318 0325 0333 0303 0344 0332 0249 0195 0308 0298 0285 0331 0334 0305 0243 0293 0266 0881 C 16 0324 0336 0321 0277 0325 0313 0242 0191 0320 0295 0285 0320 0320 0288 0299 0247 0292 0539 C 17 0272 0285 0254 0233 0266 0254 0206 0163 0251 0256 0240 0266 0264 0230 0234 0249 0184 0433 s 1098 1128 1131 0927 1092 1036 0694 0562 0793 0873 0827 0919 0925 0851 0839 0496 0476 Table 5 The total influences matrix T D and influences given/received for dimensions Dimensions D 1 D 2 D 3 D 4 D 5 D 6 r i Dimensions r i s i r i + s i r i s i D 1 Quality 0373 0383 0299 0369 0354 0345 2123 D 1 2123 1935 406 019 D 2 Delivery 0339 0339 0289 0324 0326 0292 1908 D 2 1908 1959 387 005 D 3 Risk 028 0315 0228 0284 0283 0259 1649 D 3 1649 1544 319 011 D 4 Cost 032 0313 0248 0291 0289 0281 1742 D 4 1742 1849 359 011 D 5 Service 0324 0331 0252 0304 0291 0279 1781 D 5 1781 1814 359 003 D 6 Environmental collaboration 0299 0278 0229 0277 0272 0243 1597 D 6 1597 1699 330 010 s 1935 1959 1544 1849 1814 1699 Note: Let i = be r i + s i and r i s i As shown in Tables 7 11, we used the DANP method to obtain the weights and priority of dimensions and criteria of the empirical case of Corporation G According to Table 12, we found that the priority in global weight of the first dimension is delivery (D 2 ), followed by quality (D 1 ), cost (D 4 ), service (D 5 ), environmental collaboration (D 6 ), and risk (D 3 ), in that order In addition, we extended the priority of criteria in each dimension from the local weights in Table 12 For example, delivery (D 2 ) is the first priority in dimensions of a global weight; when extended to local weight, however, we know that the delivery on-time rate (C 5 ) will be the first priority of delivery (D 2 ) All these local and global weights will be helpful in selecting the best alternatives in MCDM problems with VIKOR 45 Using a VIKOR model to calculate the performance value and to select the best alternative vendor for the case corporation There are three vendors to supply plastic recycled materials for the corporation G According to the aforementioned 6 dimensions and 17 criteria, we evaluated the performance of each vendor Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
G Model RECYCL-2546; No of Pages 17 10 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx Table 6 The sum of influences given and received on criteria Criteria r i s i r i + s i r i s i Ingredient consistency C 1 1145 1098 2243 0047 Process capability C 2 1141 1128 2270 0013 Yield rate C 3 1070 1131 2201 0061 Shortest lead time C 4 1013 0927 1940 0086 Delivery on time rate C 5 0985 1092 2077 0107 Serious delivery delay rate C 6 1056 1036 2092 0021 Geographical location C 7 0673 0694 1367 0021 Political stability C 8 0602 0562 1164 0040 Equipment capacity change C 9 0774 0793 1567 0018 Recycled material price C 10 0970 0873 1842 0097 Handling cost C 11 0813 0827 1640 0014 Process loss cost C 12 0835 0919 1754 0083 Response to demand C 13 0803 0925 1728 0122 Information acquisition C 14 0930 0851 1780 0079 After sales service C 15 0881 0839 1721 0042 Technology for recycling products and process C 16 0539 0496 1035 0043 Green manufacturing policy C 17 0433 0476 0909 0043 Note: Let i = be r i + s i and r i s i Table 7 The new matrix T c obtained by normalizing matrix T c in criteria Criteria C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 11 C 12 C 13 C 14 C 15 C 16 C 17 C 1 0283 0359 0358 0302 0353 0345 0316 0254 0430 0328 0305 0367 0348 0334 0318 0522 0478 C 2 0350 0292 0358 0302 0355 0343 0313 0253 0433 0321 0309 0369 0351 0335 0314 0521 0479 C 3 0349 0359 0293 0304 0357 0338 0310 0250 0440 0320 0308 0372 0352 0334 0314 0525 0475 C 4 0321 0334 0345 0262 0376 0361 0335 0263 0403 0332 0315 0353 0358 0331 0311 0520 0480 C 5 0316 0336 0348 0331 0309 0360 0336 0261 0403 0329 0318 0353 0355 0328 0317 0534 0466 C 6 0320 0332 0348 0317 0385 0299 0331 0271 0398 0325 0313 0362 0353 0326 0321 0526 0474 C 7 0332 0336 0332 0302 0355 0344 0288 0321 0391 0333 0319 0348 0363 0312 0325 0521 0479 C 8 0329 0335 0336 0303 0349 0347 0394 0221 0385 0337 0312 0350 0348 0314 0338 0525 0475 C 9 0322 0337 0341 0307 0355 0338 0341 0275 0384 0332 0307 0362 0358 0331 0311 0528 0472 C 10 0333 0340 0327 0313 0345 0341 0337 0265 0398 0287 0342 0371 0362 0323 0315 0532 0468 C 11 0328 0334 0333 0314 0340 0346 0331 0267 0402 0359 0271 0370 0356 0321 0323 0525 0475 C 12 0333 0337 0330 0315 0345 0339 0330 0252 0418 0362 0329 0309 0349 0325 0326 0525 0475 C 13 0321 0339 0340 0307 0348 0345 0332 0258 0410 0329 0315 0356 0310 0346 0344 0532 0468 C 14 0320 0337 0343 0309 0349 0342 0312 0254 0434 0336 0311 0353 0367 0288 0345 0537 0463 C 15 0326 0333 0341 0310 0351 0339 0331 0259 0410 0326 0312 0362 0379 0346 0275 0524 0476 C 16 0330 0342 0327 0303 0355 0342 0321 0254 0425 0328 0316 0356 0353 0317 0330 0458 0542 C 17 0336 0351 0313 0310 0353 0337 0332 0263 0405 0336 0315 0349 0362 0316 0322 0575 0425 Table 8 The unwighted supermatrix W Criteria C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 11 C 12 C 13 C 14 C 15 C 16 C 17 C 1 0283 0350 0349 0321 0316 0320 0332 0329 0322 0333 0328 0333 0321 0320 0326 0330 0336 C 2 0359 0292 0359 0334 0336 0332 0336 0335 0337 034 034 0337 0339 0337 0333 0342 0351 C 3 0358 0358 0293 0345 0348 0348 0332 0336 0341 0327 0333 033 034 0343 0341 0327 0313 C 4 0302 0302 0304 0262 0331 0317 0302 0303 0307 0313 0314 0315 0307 0309 031 0303 031 C 5 0353 0355 0357 0376 0309 0385 0355 0349 0355 0345 034 0345 0348 0349 0351 0355 0353 C 6 0345 0343 0338 0361 036 0299 0344 0347 0338 0341 0346 0339 0345 0342 0339 0342 0337 C 7 0316 0313 031 0335 0336 0331 0288 0394 0341 0337 0331 033 0332 0312 0331 0321 0332 C 8 0254 0253 025 0263 0261 0271 0321 0221 0275 0265 0267 0252 0258 0254 0259 0254 0263 C 9 043 0433 044 0403 0403 0398 0391 0385 0384 0398 0402 0418 041 0434 041 0425 0405 C 10 0328 0321 032 0332 0329 0325 0333 0337 0332 0287 0359 0362 0329 0336 0326 0328 0336 C 11 0305 0309 0308 0315 0318 0313 0319 0312 0307 0342 0271 0329 0315 0311 0312 0316 0315 C 12 0367 0369 0372 0353 0353 0362 0348 035 0362 0371 037 0309 0356 0353 0362 0356 0349 C 13 0348 0351 0352 0358 0355 0353 0363 0348 0358 0362 0356 0349 031 0367 0379 0353 0362 C 14 0334 0335 0334 0331 0328 0326 0312 0314 0331 0323 0321 0325 0346 0288 0346 0318 0316 C 15 0318 0314 0314 0311 0317 0321 0325 0338 0311 0315 0323 0326 0344 0345 0275 033 0322 C 16 0522 0521 0525 052 0534 0526 0521 0525 0528 0532 0525 0525 0532 0537 0524 0458 0575 C 17 0478 0479 0475 048 0466 0474 0479 0475 0472 0468 0475 0475 0468 0463 0476 0542 0425 Note: W = (T c ) 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
RECYCL-2546; No of Pages 17 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx 11 Fig 5 The influential NRM of relation within dimensions Fig 6 The influential NRM of 17 criteria within 6 dimensions according to the opinions of eight experts in ACP manufacturing We evaluated performance on a scale of 0 4, with 0 indicating very bad and 4 indicating the best Then, we used the average performance scores of each vendor and applied the VIKOR model to obtain the performance and aspired level gaps of alternative vendors, as shown in Table 12 5 Results and discussion According to the empirical study in Section 4, our proposed hybrid MCDM model could provide more relevant results For instance, the interdependent and feedback relationship of VS dimensions and criteria can be used as the performance of Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009
G Model RECYCL-2546; No of Pages 17 12 C-H Hsu et al / Resources, Conservation and Recycling xxx (2012) xxx xxx Table 9 The new matrix T D obtained by matrix T D Dimensions D 1 D 2 D 3 D 4 D 5 D 6 D 1 Quality 0176 0177 0170 0184 0182 0187 D 2 Delivery 0180 0178 0191 0180 0186 0174 D 3 Risk 0141 0151 0138 0142 0141 0143 D 4 Cost 0174 0170 0172 0167 0171 0173 D 5 Service 0167 0171 0172 0166 0163 0170 D 6 Environmental collaboration 0162 0153 0157 0161 0157 0152 Table 10 Weighting the unweighted supermatrix based on total-influence normalized matrix W Criteria C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 11 C 12 C 13 C 14 C 15 C 16 C 17 C 1 005 0062 0061 0057 0056 0057 0056 0056 0055 0061 006 0061 0058 0058 0059 0062 0063 C 2 0063 0051 0063 0059 006 0059 0057 0057 0057 0062 0062 0062 0062 0061 0061 0064 0066 C 3 0063 0063 0051 0061 0062 0062 0056 0057 0058 006 0061 0061 0062 0062 0062 0061 0059 C 4 0054 0055 0055 0047 0059 0056 0058 0058 0059 0056 0056 0057 0057 0057 0058 0053 0054 C 5 0064 0064 0064 0067 0055 0068 0068 0067 0068 0062 0061 0062 0065 0065 0065 0062 0061 C 6 0062 0062 0061 0064 0064 0053 0066 0066 0065 0061 0062 0061 0064 0064 0063 006 0059 C 7 0045 0044 0044 0051 0051 005 004 0054 0047 0048 0047 0047 0047 0044 0047 0046 0048 C 8 0036 0036 0035 004 0039 0041 0044 003 0038 0038 0038 0036 0036 0036 0037 0036 0038 C 9 0061 0061 0062 0061 0061 006 0054 0053 0053 0057 0057 0059 0058 0061 0058 0061 0058 C 10 0057 0056 0056 0056 0056 0055 0057 0058 0057 0048 006 006 0056 0057 0056 0057 0058 C 11 0053 0054 0054 0053 0054 0053 0055 0054 0053 0057 0045 0055 0054 0053 0053 0055 0055 C 12 0064 0064 0065 006 006 0061 006 006 0062 0062 0062 0052 0061 006 0062 0062 006 C 13 0058 0058 0059 0061 0061 006 0062 006 0062 006 0059 0058 0051 006 0062 006 0062 C 14 0056 0056 0056 0056 0056 0056 0054 0054 0057 0054 0053 0054 0056 0047 0056 0054 0054 C 15 0053 0052 0052 0053 0054 0055 0056 0058 0053 0052 0054 0054 0056 0056 0045 0056 0055 C 16 0085 0085 0085 008 0082 008 0082 0082 0083 0086 0085 0085 0083 0084 0082 007 0087 C 17 0078 0078 0077 0073 0071 0072 0075 0075 0074 0076 0077 0077 0073 0073 0075 0082 0065 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Note: W = T D W recycled material vendors Also, the results can be used to obtain the aspired level gaps of criteria for improvement Some findings are given as follows: (1) From DEMENTAL model in Fig 5, we can easily understand that six dimensions are influenced each other such as quality of the product (D 1 ) will influence delivery schedule (D 2 ), supply risk (D 3), overall cost of product (D 4 ), service (D 5 ), and environmental collaboration (D 6 ); supply risk (D 3 ) will influence delivery schedule (D 2 ), service (D 5 ), overall cost of the product (D 4 ), and environmental collaboration (D 6 ) These influential relations will help managers to do the decision-making In order to reduce the overall cost of product, managers should request their vendors to improve product quality first Then, managers can refer D 1 from Fig 6 to advise their vendors to improve the ingredient consistency of product for upgrading their product quality (2) According to the results of DANP with VIKOR (see Table 12), we found that the total performance of three vendors (V 1 V 3 ) are 2294, 2192, and 2018 respectively That is, we can treat vendor V 1 as the best vendor to conduct the recycled material In addition, we found that vendor V 1 has very good performance on the quality of the product (D 1 ) with 2752 score and environmental collaboration (D 6 ) with 2434 score These two scores of vendor V 1 are larger than those of other two vendors (3) The traditional VS approaches are only used to select the best vendor Our proposed hybrid MCDM model can be used to select the best vendor and to analyze the gaps of aspired level for Table 11 The stable matrix of ANP when power lim (W ) g g Criteria C 1 C 2 C 3 C 4 C 5 C 6 C 7 C 8 C 9 C 10 C 11 C 12 C 13 C 14 C 15 C 16 C 17 C 1 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 2 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 3 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 4 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 5 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 6 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 7 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 C 8 004 004 004 004 004 004 004 004 004 004 004 004 004 004 004 004 004 C 9 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 10 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 11 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 C 12 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 13 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 14 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 006 C 15 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 005 C 16 008 008 008 008 008 008 008 008 008 008 008 008 008 008 008 008 008 C 17 008 008 008 008 008 008 008 008 008 008 008 008 008 008 008 008 008 Please cite this article in press as: Hsu C-H, et al The best vendor selection for conducting the recycled material based on a hybrid MCDM model combining DANP with VIKOR Resour Conserv Recy (2012), doi:101016/resconrec201202009