رویکرد منطق فازی برای ارزیابی تهیه کننده برای توسعه

Decision making techniques used to help evaluate current suppliers should aim at classifying performance of individual suppliers against desired levels of performance so as to devise suitable action plans to increase suppliers׳ performance and capabilities. Moreover, decision making related to what course of action to take for a particular supplier depends on the evaluation of short and long term factors of performance, as well as on the type of item to be supplied. However, most of the propositions found in the literature do not consider the type of supplied item and are more suitable for ordering suppliers rather than categorizing them. To deal with this limitation, this paper presents a new approach based on fuzzy inference combined with the simple fuzzy grid method to help decision making in the supplier evaluation for development. This approach follows a procedure for pattern classification based on decision rules to categorize supplier performance according to the item category so as to indicate strengths and weaknesses of current suppliers, helping decision makers review supplier development action plans. Applying the method to a company in the automotive sector shows that it brings objectivity and consistency to supplier evaluation, supporting consensus building through the decision making process. Critical items can be identified which aim at proposing directives for managing and developing suppliers for leverage, bottleneck and strategic items. It also helps to identify suppliers in need of attention or suppliers that should be replaced.

Nowadays, manufacturing companies rely heavily on suppliers for providing materials and components used in finished products. Some authors say that approximately 50–70% of production costs are spent on purchased materials and components (Prajogo et al., 2012). Purchasing decisions affect important activities such as inventory management and production planning and control (Katsikeas et al., 2004 and Govindan et al., 2010) and have a significant influence on the cost, quality and delivery of products of the buying company (Talluri and Sarkis, 2002). Thus, managing the performance of suppliers and supporting their continuous improvement has become very critical for managing organizations and supply chains (Schoenherr et al., 2012). Managing buyer–supplier relationships includes activities such as supplier selection and development (Park et al., 2010, Chen, 2011 and Inemek and Matthyssens, 2011). Supplier evaluation helps to make decisions about supplier selection and development (Schmitz and Platts, 2004). Supplier development is commonly defined as any effort or set of practices of a buying company with its supplier aiming at increasing the performance and capabilities of the supplier so as to better meet the buying firm׳s supply needs (Govindan et al., 2010 and Bai and Sarkis, 2011). There are many supplier development practices that may be used (Krause, 1997, Govindan et al., 2010, Bai and Sarkis, 2011, Blome et al., 2013, Dekkers et al., 2013 and He et al., 2013). Choosing what type of supplier development practice or what course of action to deploy to a particular supplier first of all depends on the supplier׳s evaluation. There are variety of models proposed in the literature aimed at evaluating and segmenting the base of suppliers based on the evaluation of the suppliers related to several factors such as quality, delivery, financial health and technical capabilities, among others (Olsen and Ellram, 1997, Araz and Ozkarahan, 2007, Sarkar and Mohapatra, 2006, Omurca, 2013, Rezaei and Ortt, 2013a and Rezaei and Ortt, 2013b). Most of them are two dimensional models and the supplier base segmentation process is based on dimensions related to supplier performance, such as attractiveness of the supplier and intensity of the relationship (Olsen and Ellram, 1997), short-term performance and long-term capability (Sarkar and Mohapatra, 2006) and willingness and capabilities (Rezaei and Ortt, 2013a and Rezaei and Ortt, 2013b). However, decision making related to what type of supplier development practice or what course of action to take regarding a particular supplier depends not only on the categorization of the supply based on its evaluation of performance. The type of item to be supplied and what implications it may have on supply management should also be considered. A much cited item classification model was proposed by Kraljic (1983), which classifies items into four categories: strategic; bottleneck; leverage and noncritical. Kraljic (1983) proposes that each of these categories demands a distinctive purchasing strategy. According to a study carried out by Nellore and Söderquist (2000) in the automotive industry, leverage, bottleneck and strategic items all require increasing degrees of collaboration in the specification process. Consequently, the higher the evaluation of the potential for partnership of a particular supplier, the higher the chance of developing a strategic partner will be. Another important issue to be considered refers to the techniques used in the decision making process. Different decision making techniques are proposed in the literature to deal with the process of supplier evaluation, especially in supplier selection (De Boer et al., 2001, Wu and Barnes, 2011, Ho et al., 2010 and Chai et al., 2013). Evaluation for the purpose of supplier development differs from the case of supplier selection, in the sense that the latter seeks to define an order of preference among potential suppliers while the former aims to categorize suppliers (De Boer et al., 2001, Keskin et al., 2010 and Omurca, 2013). However, the techniques proposed by most of the studies on supplier evaluation found in the literature are more adequate for ordering suppliers (Chen et al., 2006, Sarkar and Mohapatra, 2006, Araz and Ozkarahan, 2007, Çelebi and Bayraktar, 2008, Wang, 2008, Lee et al., 2009, Lin, 2009, Park et al., 2010, Chen, 2011, Zeydan et al., 2011, Baskarahan et al., 2012, Pitchipoo et al., 2013 and Rezaei and Ortt, 2013a). Another limitation regards the use of techniques based on comparison between suppliers (Olsen and Ellram, 1997, Sarkar and Mohapatra, 2006 and Araz and Ozkarahan, 2007; Tuzkaya et al., 2009; Lee et al., 2009, Park et al., 2010, Shirinfar and Haleh, 2011, Zeydan et al., 2011 and Rezaei and Ortt, 2013a). Since the main aim of evaluation for supplier development is to classify individual suppliers based on gaps between real and desired performance, techniques that yield relative performance evaluations are not the most adequate ones. On the other hand, fuzzy rule-based classification methods (Ishibuchi et al., 1992, Nozaki et al., 1996, Castellano and Fanelli, 1999, Nguyen et al., 2012 and Lima et al., 2013) are especially useful for categorizing alternatives, as is the case of segmenting products or suppliers in purchasing models. However, none of the proposals found in the literature dealing with supplier evaluation and segmentation adopts a procedure for fuzzy pattern classification based on decision rules. Therefore, this paper proposes a new approach for evaluation of suppliers for development purposes. Categorization of suppliers for development is dependent on the evaluation of the suppliers, as well as on the categorization of the supplied items. Items are categorized according to the dimension complexity of item and complexity of supply market. Evaluation of suppliers is made on the basis of short-term delivery performance and long-term potential for partnership. Fuzzy inference system combined with the simple fuzzy grid method (Ishibuchi et al., 1992) is also proposed in a procedure for pattern classification so as to categorize items and suppliers. In doing so, it is possible to categorize supplier performance according to the item category so as to indicate strengths and weaknesses of current suppliers and to aid decision making concerning action planning for supplier development. Representation of classes of supplier performance and items by fuzzy numbers allows for subjectivity of the decision makers. Also, the base of decision rules of a fuzzy inference system is designed grounded on if–then scenarios devised by specialists, therefore modeling human reasoning. A descriptive quantitative approach was adopted as a research method (Bertrand and Fransoo, 2002). The fuzzy inference systems were implemented in FuzzyTech® and MATLAB® and applied to a case in the automobile industry. A 3k factorial design was used to test the consistency and sensitivity of the inference systems. This paper is organized as follows: Section 2 briefly revises the subject of supplier management, presenting the contributions from the literature on supplier evaluation. Section 3 presents some fundamental concepts regarding the fuzzy set theory used in the proposition. The proposed fuzzy inference systems combined with the fuzzy grid method are described in detail in Section 4. Section 5 presents the application case and the sensitivity analysis. Final remarks and conclusion about this research are made in Section 6.

This paper presented a method for evaluation of suppliers for development planning that bases its decision process on analyzing the gap between real and expected supplier performance according to the item category. The method is able to identify critical items aiming at proposing directives for managing and developing suppliers for leverage, bottleneck and strategic items. Evaluation of suppliers is made on the basis of short-term delivery performance and long-term potential for partnership. Unlike the previous studies, both of these dimensions of supplier evaluation contribute to identify potential partners in the development or co-development of critical items. It also helps to identify suppliers in need of attention or suppliers that should be replaced. The fuzzy set theory is suitable for dealing with the vagueness intrinsic to qualitative factors of suppliers׳ evaluation, as well as imprecise weighing of different factors by different decision makers. The proposed method applies fuzzy inference combined with the simple fuzzy grid method as a procedure for pattern classification so as to categorize items and suppliers. Unlike other techniques used for supplier evaluation, this is particularly useful for supplier categorization using linguistic classes. Thus, the proposed techniques bring several advantages to the evaluation process, as follows: • Unlike other approaches that combine fuzzy set theory with multicriteria decision making methods, the base of rules of a fuzzy inference system is designed grounded on if–then scenarios devised by specialists, therefore modeling human reasoning. This knowledge about the problem domain is then captured and kept in the system. • Classes of supplier performance and items can be represented by fuzzy numbers, allowing for subjectivity of the decision makers. • The number of potential suppliers simultaneously evaluated is unlimited, unlike comparative approaches such as AHP, ANP, fuzzy AHP and fuzzy ANP, in which the number of alternatives is limited by the human ability to simultaneous comparative judgment. • Using linguistic terms as an indication of supplier performance is more appropriate to give feedback to suppliers in the evaluation process. • The possibility of choosing different operators such as t-norms, s-norms and defuzzification operators brings flexibility to the system. In this particular pilot application, it validated some decisions already made by the decision makers as well as it helped them in reviewing some supplier development action plans. For instance, suppliers S4, S6 and S10 were identified as potential partners in the development or co-development of critical items. On the other hand, the evaluation process indicated the need of replacing supplier S19. In addition, the decision makers involved in the pilot application reported that the adoption of the proposed method brings objectivity and consistency to the decision process, supporting discussion and consensus building through the decision making process. Although there is an effort to incorporate into the system the knowledge and experience of the specialist, once this is done, there is no need to switch between crisp numbers and linguistic terms. The decision maker will only assess the suppliers using crisp values, which will then be fuzzified. The sensitivity analysis using factorial design has shown no significant interaction effects between the criteria within the inference systems. Analysis has also revealed the relative importance of the criteria implicit in the decision rules. The pilot application of the proposed method has also shown that it is a learning process that is dependent on factors such as • the choices of criteria for item and supplier classification in the evaluation models; • parameters of the inference system such as linguistic terms and inference rules; and • knowledge of the decision makers regarding evaluation of items and suppliers. A limitation of the proposed method is that the inclusion of a new criterion increases exponentially the number of decision rules of an inference system. In the pilot application case, the first, third and fourth inference systems were built with four criteria, leading to 81 rules each. The inclusion of a new criterion in any of these systems would generate a set of 243 rules, which would imply in a greater effort and degree of difficulty to define the classes of the consequents of the decision rules. However, it is in general a good practice to work with a limited number of criteria since it facilitates the process of collecting and maintaining the supplier data base. Otherwise, when there is a need to work with a large number of criteria, it is possible to organize the inference systems in a cascade mode, so as to reduce the number of criteria in each inference system. Further research can explore other fuzzy operators, as well as the use of approaches such as 2-tuple representation (Herrera and Martínez, 2000). Other further research can explore application of the Takagi–Sugeno (Pedrycz and Gomide, 2007) technique to define the base of rules from past experiences of supplier evaluation. It can be combined with a fuzzy rule generation procedure proposed by Nozaki et al. (1996) to define the consequent class using numerical data.