Expert Systems with Applications 37 (2010) 419425

Contents lists available at ScienceDirect

Expert Systems with Applications

journal homepage: www.elsevier.com/locate/eswa

The application of Fuzzy Delphi Method and Fuzzy AHP in lubricant regenerative technology selection
Yu-Lung Hsu a,*, Cheng-Haw Lee a, V.B. Kreng b
a b

Department of Resources Engineering, National Cheng Kung University, No. 1, University Road, Tainan 701, Taiwan Department of Industrial and Information Management, National Cheng Kung University, No. 1, University Road, Tainan 701, Taiwan

a r t i c l e

i n f o

a b s t r a c t
Due to the funding scale and complexity of lubricant regenerative technology, the selection of recycling technology and policy for waste lubricant oil can be viewed as a multiple-attribute decision process that is normally made by a review committee with experts from academia, industry, and the government. This study aims to provide a systematic approach towards the technology selection, in which two phase procedures are proposed. The rst stage utilizes Fuzzy Delphi Method to obtain the critical factors of the regenerative technologies by interviewing the foregoing experts. In the second stage, Fuzzy Analytic Hierarchy Process is applied to nd the importance degree of each criterion as the measurable indices of the regenerative technologies. This study considers eight kinds of regenerative technologies which have already been widely used, and establishes a ranking model that provides decision makers to assessing the prior order of regenerative technologies. The empirical study indicates that the Proper scale is the most important evaluation criterion considered in overall experts. The demonstration of how the prior order of regenerative technologies changes under various domains of experts is addressed as well. 2009 Published by Elsevier Ltd.

Keywords: Lubricant regenerative technology Fuzzy Delphi Method Fuzzy Analytic Hierarchy Process

1. Introduction The efcient recycling of waste lubricant could help reduce both the environmental pollution and gas emission from greenhouses, thus, creating a huge efciency either from environmentallyfriendly or economic levels. Waste lubricant recycling and regeneration not only save the cost of lubricant, but also contribute to environmental protection. The proper management of dispose and recycling of the waste oil becomes critical to the management of environment (Cheng, Lin, Chang, & Huang, 2006/1). Regenerating waste oil into chemical feedstock or fuel oil is one of the preferred recycle methods. At present, there are eight kinds of common lubricant recycling technologies as follows: (1) acid/clay process; (2) distillation process; (3) solvent de-asphalting process; (4) TFE + hydro-nishing; (5) TFE + clay nishing; (6) TFE + solvent nishing; (7) solvent extraction hydro-nishing and (8) TDA + clay nishing and TDA + hydro-nishing. These technologies are different in economic benet, technology maturity and environmental impact, and new technologies have been developed and applied continuously. The government shall be responsible for technology assessment, and combine the views of academia, industrial circles and government
* Corresponding author. E-mail addresses: uf571123@ms5.hinet.net (Y.-L. Hsu), kreng@mail.ncku.edu.tw (V.B. Kreng). 0957-4174/$ - see front matter 2009 Published by Elsevier Ltd. doi:10.1016/j.eswa.2009.05.068

sectors to set up a measuring index for selection of lubricant recycling technology. The traditional Delphi Method, developed by Dalkey and Helmer (1963), has been widely used to obtain a consistent ow of answers through the results of questionnaires (Hwang & Lin, 1987; Reza & Vassilis, 1988). Delphi is an expert opinion survey method with three features: anonymous response, iteration and controlled feedback and nally statistical group response. However, some weaknesses have been exposed, it needs repetitive surveys to allow forecasting values to converge which requires much more time and cost (Hwang & Lin, 1987; Ishikawa et al., 1993). Furthermore, in many real situations, experts judgments can not be properly reected in quantitative terms. Some ambiguity will result due to the differences in the meanings and interpretations of the experts opinions. Since people use linguistic terms, such as good or very good to reect their preferences, the concept of combining fuzzy set theory and Delphi was proposed by Murray, Pipino, and Gigch (1985), and named the Fuzzy Delphi Method (FDM). Lubricant regenerative technology selection is a multiple criteria decision-making problem. Among these, the Fuzzy Analytic Hierarchy Process (FAHP) is one of the most popular (Kahraman, Cebeci, & Ruan, 2004; Teng & Tzeng, 1996; Zhau & Goving, 1991). People often use knowledge that is imprecise rather than precise. The fuzzy set theory approaches could resemble human reasoning in use of approximate information and uncertainty to generate decisions. It was specically designed to mathematically

420

Y.-L. Hsu et al. / Expert Systems with Applications 37 (2010) 419425

represent uncertainty and vagueness and provide formalized tools for dealing with the imprecision intrinsic to many problems (Kahraman et al., 2004; Williams, 2003; Zadeh, 1965). Consequently, to make this study more sensible and gain a more representative description of the decision-making process, this study would apply the FAHP to investigate which evaluation criterion is the most important in overall technical committees. This study contains two stages: the rst stage is to establish the key factors for evaluation of the waste lubricant recycling technologies, and use FDM by consulting experts of academia, industries and government sectors to select a technological selection criterion, in order to nd out the important factors to be considered while selecting a technology; the second stage is based on FAHP, and consults experts of various elds to nd out the importance of various criteria, in order to obtain the measuring index for selecting lubricant recycling technology.

Schematic diagram of Fuzzy Delphi Method threshold is shown in Fig. 1. 2.2. Fuzzy Analytic Hierarchy Process Laarhoven and Pedrycz (1983) proposed the Fuzzy Analytic Hierarchy Process in 1983, which was an application of the combination of Analytic Hierarchy Process (AHP) and Fuzzy Theory. The linguistic scale of traditional AHP method could express the fuzzy uncertainty when a decision maker is making a decision. Therefore, FAHP converts the opinions of experts from previous denite values to fuzzy numbers and membership functions, presents triangular fuzzy numbers in paired comparison of matrices to develop FAHP, thus the opinions of experts approach human thinking model, so as to achieve more reasonable evaluation criteria. As for the experts opinions, this study adopted the Similarity Aggregation Method (SAM) proposed by Hsu and Chen (1996) to integrate experts weight values for various evaluation criteria, the fuzzy weight fraction of criterion of each hierarchy is obtained through the calculating mode of FAHP, and then the sequence of signicance of each criterion is determined based on the hierarchy series connection and defuzzication mode. Laarhoven and Pedrycz (1983) proposed the FAHP, which is to show that many concepts in the real world have fuzziness. Therefore, the opinions of decision makers are converted from previous denite values to fuzzy numbers and membership numbers in FAHP, so as to present in FAHP matrix. The steps of this study based on FAHP method are as follows: 1. Determine problems: Determine the current decision problems to be solved, so as to ensure future analyses correct, this study discussed the evaluation criteria for verication of credit card. 2. Set up hierarchy architecture: Determine the evaluation criteria having indexes to be the criteria layer of FAHP, for the selection of evaluation criteria, relevant criteria and feasible schemes can be found out through reading literatures and collective discussions. This study screened the important factors conforming to target problems through FDM investigating experts opinions, to set up the hierarchy architecture. 3. Set up fuzzy paired comparison matrices: Compare the relative importance between factors given by decision makers in pairs, set up paired comparison matrices, after the denite values are converted to fuzzy numbers according to the denitions in Table 1 and Fig. 2, integrate the fuzzy evaluation values of experts based on the SAM concept proposed by Hsu and Chen (1996). 4. Calculate fuzzy weight value: Obtain the characteristic vector value of fuzzy matrix, namely the weight value of element. This

2. Methodology 2.1. Fuzzy Delphi Method Fuzzy Delphi Method was proposed by Ishikawa et al. (1993), and it was derived from the traditional Delphi technique and fuzzy set theory. Noorderhaben (1995) indicated that applying the Fuzzy Delphi Method to group decision can solve the fuzziness of common understanding of expert opinions. As for the selection of fuzzy membership functions, previous researches were usually based on triangular fuzzy number, trapezoidal fuzzy number and Gaussian fuzzy number. This study applied the triangular membership functions and the fuzzy theory to solving the group decision. This study used FDM for the screening of alternate factors of the rst stage. The fuzziness of common understanding of experts could be solved by using the fuzzy theory, and evaluated on a more exible scale. The efciency and quality of questionnaires could be improved. Thus, more objective evaluation factors could be screened through the statistical results. The FDM steps are as follows: 1. Collect opinions of decision group: Find the evaluation score of each alternate factors signicance given by each expert by using linguistic variables in questionnaires. 2. Set up triangular fuzzy numbers: Calculate the evaluation value of triangular fuzzy number of each alternate factor given by experts, nd out the signicance triangular fuzzy number of the alternate factor. This study used the geometric mean model of mean general model proposed by Klir and Yuan (1995) for FDM to nd out the common understanding of group decision. The computing formula is illustrated as follows: Assuming the evaluation value of the signicance of No. j f element given by No. i expert of n experts is wij aij ; bij ; cij ; f i 1; 2; . . . ; n; j 1; 2; . . . ; m. Then the fuzzy weighting wj of f No. j element is wj aj ; bj ; cj ; j 1; 2; . . . ; m.Among which

aj Minfaij g;
i

bj

n 1X bij ; n i1

cj Maxfcij g
i

3. Defuzzication: Use simple center of gravity method to defuzzify f the fuzzy weight wj of each alternate element to denite value Sj , the followings are obtained:

Sj

aj bj cj ; 3

j 1; 2; . . . ; m

4. Screen evaluation indexes: Finally proper factors can be screened out from numerous factors by setting the threshold a. The principle of screening is as follows: If Sj P a, then No. j factor is the evaluation index. If Sj < a, then delete No. j factor.

Fig. 1. Schematic diagram of Fuzzy Delphi Method threshold.

Y.-L. Hsu et al. / Expert Systems with Applications 37 (2010) 419425 Table 1 The denition of every fuzzy number. Fuzzy number ~ 1 1; 1; 1 ~ 2 1; 2; 3 ~ 3 2; 3; 4 ~ 4 3; 4; 5 ~ 5 4; 5; 6 ~ 6 5; 6; 7 ~ 7 6; 7; 8 ~ 8 7; 8; 9 ~ 9 8; 9; 9 Denition Equally important Judgment values between equally and moderately Moderately more important Judgment values between moderately and strongly Strongly more important Judgment values between strongly and very strongly Very strongly more important Judgment values between very strongly and extremely Extremely more important

421

Fig. 2. Scale of fuzzy numbers.

study calculated these three positive and negative value matrices respectively by using the Column Vector Geometric Mean Method proposed by Buckley.

Z i ai1  ai2   ain 1=n W i Z i ;Z 1 Z 2 Z n

8i

Among which aij : Column i row j of matrix, i; j 1; 2; . . . ; n; Z i : column vector mean value of fuzzy number, i 1; 2; . . . ; n; W i : weight of No. i factor. : multiply fuzzy numbers, e.g. assuming two triangular fuzzy e e numbers A a1 ; b1 ; c1 ; B a2 ; b2 ; c2 ,

generation rate of vehicle waste lubricant and industrial waste lubricant was 90.6% and 55.6%, respectively. According to the Planning and Rate Calculation of Waste Lubricant Recovery formulated by the Environmental Protection Agency (EPA) in 1999, vehicle lubricant accounted for about 59.7%, and industrial lubricant for 40.3% of the Taiwan lubricant market. It is thus estimated that the annual yield of vehicle lubricant was about 240,000 kl (45 59.7% 90.6% = 24.3), and of industrial lubricant about 100,000 kl (45 40.3% 55.6% = 10.0) in 2005. According to the investigation, most of the waste lubricants in Taiwan are recycled and reused as secondary oil and fuel (approx. 94%), but the remaining portion leads to environmental pollution (Table 2). The waste lubricant must meet environmentally-friendly and application criteria for recovery and reuse. Two key indicators include pollution level and viscosity index in this respect. In Europe, the regenerative oils are classied mainly according to the content of chlorides. Since chlorides are harmful to the human body, and complex nishing processes shall be required during the regeneration process, the chlorine content in reclaimed waste oil shall not exceed 50 ppm in EU regulations. A higher viscosity index of waste lubricant means a higher suitability for regeneration into lubricant. Viscosity is the most important consideration in choosing lubricants. The strength of the lubricant lm is approximately proportional to its viscosity, so the higher viscosity indicates the stronger strength of the lubricant lm. The viscosity index (VI) refers to the changing degree of viscosity dependent on temperature: the lower VI means a higher viscosity change in the case of slight temperature change, and vice versa. Thus, in the case of a higher viscosity index, no nishing process shall be additionally required to improve VI, making it more suitable for recovery and reuse with a relatively smaller operating cost. In 2005, there is approx. 340,000 kl of waste lubricant in Taiwan, so the recovery rate is about 4% if the audited statistical recovery yield of 14,000 kl in the same year is divided by 340,000 kl. As compared with European countries, it is found that Luxembourg had a recovery rate of 39%, and the average recovery rate of Europe was 50% back in 2000.

4. Evaluating model application and results 1. Reviewing relevant literature of lubricant regenerative technology and proposing important criteria: More than 17 criteria for lubricant regenerative technology based on reviewing relevant literature (Begum, Siwar, Pereira, & Jaafar, 2006; Emery, Davies, Grifths, & Williams, 2007; Finnveden, 1999; International Maritime Organization, 2004; Lin, Lin, & Jong, 2007) and the current lubricant regenerative technology selection approach are proposed. Denitions of evaluating criteria of the lubricant regenerative technology selection are presented in Table 3. 2. Screen important criteria by Fuzzy Delphi Method: This stage includes three sections. Firstly, it lists three main aspects and 17 items as the key evaluation items of lubricant regenerative technology, and a FDM interview table is set up. The second section is the interview with nine experts from the academic community, lubricant oil producers and competent authority ofcers in Taiwan. Delphi Method mostly aims at easy common understanding of group opinions through twice provision of questionnaires. FDM formed by adding the fuzzy theory in, not only maintains the advantage of Delphi Method, but also reduces the provision times of questionnaires when using traditional Delphi Method as well as the cost. For the third section, the opinions of experts in FDM questionnaires

e e A  B a1 ; b1 ; c1  a2 ; b2 ; c2 a1 a2 ; b1 b2 ; c1 c2 :
;: divide fuzzy numbers, e.g.: assuming two triangular fuzzy e e numbers A a1 ; b1 ; c1 ; B a2 ; b2 ; c2 ,

e e A; B a1 ; b1 ; c1 ;a2 ; b2 ; c2 a1 =a2 ; b1 =b2 ; c1 =c2 :

5. Hierarchy series connection: Connect all hierarchies in series, to obtain all factors weights. 6. Defuzzication: Convert fuzzy numbers to easy-comprehended denite values, this study adopts the center of gravity method to solve fuzzy numbers.

GA

Pn i1 uA xi xi Pn i1 uA xi

7. Sequencing: Sequence defuzzied criteria. 3. Lubricant regenerative technology selection in Taiwan: a background The sales volume in the Taiwan lubricant market was registered separately as 400,000 kl and 450,000 kl in 2004 and 2005, and the

422

Y.-L. Hsu et al. / Expert Systems with Applications 37 (2010) 419425

Table 2 Distribution and hazards of waste lubricant in Taiwan. Distribution Recycling dealer Factories Secondary lubricating in construction Random dumping Common purpose Secondary oil and fuel Furnace fuel De-molding Percentage 79.5 14 2.5 4 Hazards Inferior reclaimed oil, leading to mechanical damage Inclusive of heavy metal, improper incineration leading to serious pollution Pollution transfer Serious environmental pollution

Table 3 Operational type for dening criteria. Aspects Technology Criteria Operating temperature Recovery rate Product quality Development stage Legitimacy of reuse quality specications Economy Water cost Cost of overall demand for energy source Proper scale Cost of equipment demand Return on investment Subsidy Environmental protection Removal of PCB Whether generate acid sludge or not Whether there is residual oil sludge or not. Hazardous chemical substances used in process Depletion-of non-renewable resources IPCC-greenhouse effect The operating type dening During lubricant reclaiming steps, heating shall be used for separation, but different technologies will result in different operating temperatures How much lubricant oil can be extracted from each ton of waste lubricant oil is the calculation principle for recovery rate With reference to standard stipulated by API, there are ve types Refers to on the aspect of application, whether this technology is only for experiments or available for industrial mass production Whether reuse products conform to quality and environmental laws of oil products or not Take the water consumption in regenerating each ton of lubricant oil as the calculation unit, and compare with water rate to gure out water cost Take the energy source consumption in regenerating each ton of lubricant oil as the calculation unit (fuel oil, natural gas, etc.), and compare with energy source price to gure out energy source cost Take annual handling capacity (kt) as base. 210 kt/yr for small scale, 1050 ht/yr for middle scale, 50 kt/yr above for large scale Whether this technical operation needs more equipments which are more precise Twenty year return on investment The government provides subsidies for waste lubricant oil processing Take 50 ppm as the dividing line, there are pcb residuals if the content exceeds 50 ppm The acid sludge means during operation, the mixed precipitate of waste acid and foreign matter derived from removing foreign matter by sulfuric acid Oil sludge is the asphaltic substances which cannot be fractionated at the end of lubricant oil reclaiming process Hazardous substances to environment or human body used in lubricant oil regenerating. Such as sulfuric acid or organic solvents Take the consumption of non-renewable resources in regenerating each ton of lubricant oil as the calculation unit How much greenhouse effect gas will be generated during regenerating each ton of lubricant oil

are converted to triangular fuzzy numbers, and defuzzied values can be gured out after calculation. This stage adopts elements with threshold above 7, and the key evaluation items with threshold below 7 are deleted. The important evaluation items after screening are listed in Table 4. 3. Establish a hierarchical framework: Based on the FDM, a general consensus among experts can be reached to establish a hierarchical structure. The lubricant regenerative technology can be evaluated based on three evaluation aspects and 17 evaluation criteria (Fig. 3).

4. Interview experts of all domains and integrate their opinions:Subject to who ll in AHP questionnaires possess sufcient professional knowledge, so the interviewees are experts from academia, lubricant producers and competent authority ofcers. The evaluation of each factor must go through consistency verication to ensure preferable credibility of results. In order to increase the objectivity of results, there are 17 experts to be interviewed. In the past, the integration of opinions from questionnaires mainly adopted geometric mean method, but the unreasonable integration of group opinions therein would

Table 4 Evaluation criteria after FDM screening. Aspects (code number) Criteria (code number) Score Min Technology (A1) Operating temperature (C1) Recovery rate (C2) Product quality (C3) Development stage (C4) Water cost (C5) Cost of overall demand for energy source (C6) Proper scale (C7) Cost of equipment demand (C8) Removal of PCB (C9) Whether acid sludge is generated or not (C10) Whether there is residual oil sludge or not (C11) Hazardous chemical substances used in process (C12) 2 4 3 4 1 4 4 3 1 2 4 2 Max 10 10 10 10 10 10 10 10 10 10 10 10 Average 7.532 8.132 7.765 8.924 7.454 8.688 8.951 7.623 7.311 8.259 8.358 8.053 De-fuzzy 7.213 7.824 7.437 8.058 7.127 7.953 8.462 7.145 7.102 7.543 7.893 7.347

Economy (A2)

Environmental protection (A3)

Y.-L. Hsu et al. / Expert Systems with Applications 37 (2010) 419425

423

Goal

Aspects

Criteria
Operating temperature Recovery rate

Technology Product quality Development stage

Water cost Cost of overall demand for energy source Economy Proper scale Cost of equipment demand

PCB removal Acid sludge generated residual oil sludge Hazardous chemical substances used

Environmental protection

Fig. 3. The hierarchy model of lubricant regenerative technology selection.

result in incorrect results. Therefore, this study adopts SAM which was proposed by Hsu and Chen (1996), which can integrate group opinions more reasonably, so as to increase the credibility of questionnaires. 5. Calculate the weights of evaluation criteria and weight result of evaluation criteria: The weight values of various elements can be obtained through the opinions of experts resulted from SAM and the FAHP systematic steps. After sequencing, the evaluation criteria have higher signicance, so decision makers can make correct judgments more quickly. Table 5 is the evaluation criteria weight by FAHP, the evaluation criteria weight is obtained based on FAHP questionnaire results of experts, nally the questionnaire results of all experts are integrated to become the overall weight. As for different aspects, experts of various elds pay relatively consistent attention to the operation aspect (A1) and economy aspect (A2), they give quite high weight to these aspects, and the experts from industrial cirTable 5 Evaluation criteria weight of experts from different elds. Aspects Weights of aspects Academic community A1 0.354 Industrial circle 0.432 Government 0.321 Overall 0.395

cles lay stress on the operation aspect in particular. However, the experts from various domains have different opinions on the environmental protection aspect (A3); those from academic community and government sectors think a lot of environmental protection aspect, while the experts from industrial circles think less of environmental protection. For the evaluation criteria contained in various aspects, the experts of various elds are consistent with them, only a few evaluation criteria have difference in opinions. For the operation aspect, the experts of various domains dont have much difference in opinions, most of them lay stress on the evaluation criteria of development stage (C4); for the economy aspect, the experts of various domains have signicant difference in water cost (C5), cost of overall demand for energy source (C6) and equipment demand (C8), among which, experts of industrial circles lay stress on equipment demand which is not stressed by the experts of other two domains, but the experts of industrial circles dont lay stress on cost of overall demand for energy source which is stressed by the other two domains. The industrial circles and the government sectors have much difference in water cost; for the environmental protection aspect, the experts from various domains pay much attention to acid sludge generated (C10), but they have difference in three evaluation criteria such as PCB removal (C9), residual oil sludge (C11) and hazardous chemical substances used (C12), the opinions of experts from industrial circles are different from those of the experts from the other two domains. Fig. 4 shows the weights of lubricant regenerative technology hierarchy model. The maximum weight obtained by combining global priority is technology aspect (0.395) through FAHP, secondly is the economy aspect (0.358) and environmental protection aspect (0.247). As for sequencing the global priority of various evaluation criteria, the weight of the rst four evaluation criteria exceeds 0.1, therein the proper scale (0.1292) in the economy aspect possesses the maximum weight, but the second and third evaluation criteria are on the technology aspect. These four evaluation criteria are paid much attention to in lubricant regenerative technology selection; the result demonstrates technology and economy aspects are main consideration in the evaluation process. 5. Conclusions This study investigates the key factors in lubricant regenerative technology selection by combining FDM, SAM and FAHP, and establishes objective and standardized references. A total of 17 factors inuencing lubricant regenerative technology selection are analyzed through FDM experts opinions investigation, Experts of academia, lubricant oil industry and government sectors were interviewed, and 12 evaluation criteria were obtained as the key factors by interviewed bank experts. SAM and FAHP were used to

Selection for lubricant regenerative technology

A2 A3

Criteria

Weights of criteria Academic community Industrial circle 0.092 0.308 0.312 0.288 0.105 0.192 0.360 0.343 0.125 0.399 0.235 0.241 Government 0.091 0.294 0.279 0.336 0.215 0.283 0.301 0.201 0.261 0.280 0.144 0.315 Overall 0.105 0.295 0.284 0.316 0.153 0.234 0.361 0.252 0.195 0.351 0.168 0.286

C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12

0.135 0.282 0.274 0.309 0.143 0.238 0.377 0.242 0.253 0.294 0.128 0.325

0.324

0.398

0.322

0.358

0.322

0.170

0.357

0.247

424

Y.-L. Hsu et al. / Expert Systems with Applications 37 (2010) 419425

Goal

Aspects

Criteria
Operating temperature Recovery rate

Global Ranking priority 11 3 4 2

Technology Product quality Development stage

integrate experts opinions to obtain the signicance evaluation of various evaluation criteria given by experts in group decision. The results from experts of different elds were compared and analyzed to show the similarities and differences of various experts in lubricant regenerative technology selection. Finally, the results of all experts were used as the evaluation index of lubricant regenerative technology selection. The following conclusions were reached by analyzing the evaluation criteria stressed by experts of various domains when evaluating the lubricant regenerative technology selection based on the demonstration of this study. 1. Experts of various domains lay different emphasis on three main aspects: The experts of academic community lay similar stress on three aspects; only the technology aspect has a slightly higher weight. This is probably because the experts of academic community include those of environmental industry, chemical industry, machinery, electric machinery and so on, therefore the three aspects have relatively mean score on weight; the industrial circles lay emphasis on the technology and economy aspects, the weight of environmental protection aspect is obviously low; the government sectors pay attention to the environmental protection aspect, since they care about whether lubricant regenerative technology will generate secondary pollution or not. Due to various domain experts give quite different weights to different aspects; it is necessary to collect all opinions of different domain experts in the course of lubricant regenerative technology selection, so as to make the evaluation more objective and feasible.

Selection for lubricant regenerative technology

Water cost Cost of overall demand for energy source Economy Proper scale Cost of equipment demand

9 7 1 5

PCB removal acid sludge generated residual oil sludge Hazardous chemical substances used
Fig. 4. The weights of lubricant regenerative technology hierarchy model.

10 6 11 8

Environmental protection

2. Experts of various elds pay quite different attention to evaluation criteria in environmental protection aspect: Although the environmental protection aspect has minimum weight, four evaluation criteria in it make experts from three domains have most difference in their opinions. Due to PCB removal and hazardous chemical substances used, these two evaluation criteria because more severe secondary pollution, they are stressed by experts of academic community and government sectors, but the industrial circles dont. Therefore, the government legislates and sets a bafe plate for the evaluation criteria in environmental protection aspect, and eliminates heavy-pollution technical proposals in advance. 3. Technology aspect is mostly concerned: Due to many technical proposals are still in experimental development stage, or there are a few successful commercial operations, experts of various domains lay stress on the performance of technology aspect, among which, three evaluation criteria such as development stage, recovery rate and product quality rank the second to fourth place in the global priority sequence, its degree of importance is obvious. 4. Proper scale of technology is the most important evaluation criteria: There are quite much waste lubricant oil in Taiwan, however, too many demands of waste lubricant oil recovery processing cannot be satised yet, The waste lubricant oil recovery volume and the scale of potential competitors within the regional extent should be considered in the course of lubricant regenerative technology selection, therefore, it is very important to select proper scale technical proposal.

Y.-L. Hsu et al. / Expert Systems with Applications 37 (2010) 419425

425

References
Begum, R. A., Siwar, C., Pereira, J. J., & Jaafar, A. H. (2006). A benet-cost analysis on the economic feasibility of construction waste minimization: The case of Malaysia. Resources, Conservation and Recycling, 48(1), 8698. Cheng, Y. W., Lin, K. H., Chang, K. H., & Huang, W. R. (2006/1). Schedule of review of waste lubricant recycling system. Environmental Protection Agency, Executive Yuan. Dalkey, N., & Helmer, O. (1963). An experimental application of the Delphi method to the use of experts. Management Science, 9, 458467. Emery, A., Davies, A., Grifths, A., & Williams, K. (2007). Environmental and economic modeling: A case study of municipal solid waste management scenarios in Wales. Resources, Conservation and Recycling, 49(3), 244263. Finnveden, G. (1999). Methodological aspects of life cycle assessment of integrated solid waste management systems. Resources, Conservation and Recycling, 26(34), 173187. Hsu, H. M., & Chen, C. T. (1996). Aggregation of fuzzy opinions under group decision making. Fuzzy Sets and System, 79, 279285. Hwang, C. L., & Lin, M. J. (1987). Group decision making under multiple criteria: Methods and applications. Springer-Verlag. International Maritime Organization (2004). Guidelines on ship recycling. IMO Resolution A 962(23), 736. Ishikawa, A., Amagasa, M., Shiga, T., Tomizawa, G., Tatsuta, R., & Mieno, H. (1993). The maxmin Delphi method and fuzzy Delphi method via fuzzy integration. Fuzzy Sets and Systems, 55, 241253.

Kahraman, C., Cebeci, U., & Ruan, D. (2004). Multi-attribute comparison of catering service companies using fuzzy AHP: The case of Turkey. International Journal Production Economics, 87, 171184. Klir, G. J., & Yuan, B. (1995). Fuzzy sets and fuzzy logic Theory and application. New Jersey: Prentice-Hall Inc. Laarhoven, P. J. M., & Pedrycz, W. (1983). A fuzzy extension of Satis priority theory. Fuzzy Sets and System, 11, 229241. Lin, B., Lin, C.-Y., & Jong, T.-C. (2007). Investigation of strategies to improve the recycling effectiveness of waste oil from shing vessels. Marine Policy, 31(4), 415420. Murray, T. J., Pipino, L. L., & Gigch, J. P. (1985). A pilot study of fuzzy set modication of Delphi. Human Systems Management, 680. Noorderhaben, N. (1995). Strategic decision making. UK: Addison-Wesley. Reza, K., & Vassilis, S. M. (1988). Delphi hierarchy process (DHP): A methodology for priority setting derived from the Delphi method and analytical hierarchy process. European Journal of Operational Research, 137, 347354. Teng, J. Y., & Tzeng, G. H. (1996). Fuzzy multicriteria ranking of urban transportation investment alternative. Transportation Planning and Technology, 20, 1531. Williams, E. (2003). Forecasting material and economic ows in the global production chain for silicon. Technology Forecasting and Social Change, 70, 341357. Zadeh, L. A. (1965). Fuzzy sets. Information Control, 8, 338353. Zhau, R., & Goving, R. (1991). Algebraic characteristics of extended fuzzy numbers. Information Science, 54, 103130.