A hybrid fuzzy multi-criteria decision making model for selecting a sustainable supplier of forklift filters: a case study from the mining industry

Supplier selection has long been a strategic cornerstone for the advancement of companies from the mining sector due to its impact on many sectors as inventory management, production planning, maintenance scheduling, financial resources and environment. In this regard, expert decision-makers should deem a number of performance conflicting criteria and supplier alternatives. Moreover, choosing a suitable Multi-Criteria Decision Making (MCDM) approach has become a key strategic consideration for supplier selection. Hence, this paper presents a combination of powerful MCDM methods to select the most suitable supplier of forklift filters. The Fuzzy Analytic Hierarchy Process (FAHP) is first used to calculate the initial weights of criteria and sub-criteria under uncertainty, followed by the implementation of Fuzzy Decision Making Trial and Evaluation Laboratory (FDEMATEL) to assess the interrelations and feedback between them. The fuzzy theory is used in both AHP and DEMATEL methods to represent the human vagueness when making judgments. Then, FAHP and FDEMATEL are combined to obtain the final contributions of both criteria and sub-criteria on the basis of interrelations and uncertainty. Finally, the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) is used to rank the supplier alternatives. To validate the effectiveness of the proposed framework, a case study is presented. The results evidence that quality criterion is the most crucial aspect when selecting suppliers of forklift filters.


Introduction
Nowadays organizations recognize the importance of sustainability and it has become an integral part of their business strategy. The main idea of sustainability is to maintaining a harmonious balance respecting the environmental and social concerns, while pursuing 1 3 economic objectives (Faisal et al. 2017). In the case of non-renewable resources, like minerals, the task is highly challenging. In fact, it is necessary to meet the needs of the current generation but also to sustain and enhance its other resources or develop new ones, which future generations can rely on. Despite being a fundamental objective to achieve sustainability, Sauer and Seuring (2017) found in their review that research for a sustainable mining industry has been scarce. Whilst there is a pressure from the different stakeholders to develop sustainable strategies (Schneider and Wallenburg 2012), there is also a debate in the ability to adopt the core principles of sustainability by the mining companies (Sauer and Seuring 2017). In fact, the mining industry is known to have a poor environmental reputation (Muduli et al. 2013). This have forced many mining companies to face competitive, regulatory, and community/social pressures causing increased consideration of greening their supply chains (Kusi-Sarpong et al. 2016). One way to improve the supply chain is to select better suppliers (Igarashi et al. 2013). Suppliers impact all the value chain and also the sustainability performance of the company. Therefore, it is indicated to have a structured approach in the selection process (Chai et al. 2013). Moreover, as selecting a supplier is based on several criteria, which are often conflicting, a structured decision-making approach, as in multi-criteria decision making (MCDM) methods, can help managers in this exercise (Ishizaka and Nemery 2013a). MCDM can take into account several conflicting criteria, including sustainable criteria. In this paper, we present a hybrid MCDM model for selecting a sustainable supplier. The Fuzzy Analytic Hierarchy Process (FAHP) is first applied to calculate the initial weights of criteria and sub-criteria under uncertainty, followed by the implementation of Fuzzy Decision Making Trial and Evaluation Laboratory (FDEMATEL) to assess the interrelations and feedback between the decision elements. The fuzzy theory is used in both AHP and DEMATEL methods to represent the human vagueness when making judgments. Then, FAHP and FDEMATEL are combined to obtain the final contributions of both criteria and sub-criteria on the basis of interrelations and uncertainty. Finally, the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) is used to rank the supplier alternatives. A case study is presented to validate the effectiveness of the proposed framework.
The rest of the paper is structured as follows. Section 2 reviews the literature on supplier selection. Section 3 presents the new hybrid method. Section 4 describes the case study. Finally, Sect. 5 concludes the paper.

Literature review
The supplier selection problem represents a very complex problem due to the presence of many decision levels and different aspects to be considered (Sultana et al. 2015). At the same time, the importance in the whole supply chain management and the effects on the performance of the selection should be as much as possible careful and justified (Khan et al. 2018). For these reasons, in the literature several approaches have been proposed to deal with such problems (for a review see e.g., Chai et al. 2013). One of the most common used approaches is the MCDM method that supports the managers to select the suppliers, based on the defined criteria and their appropriate evaluations (Kahraman et al. 2003). Among the several MCDM approaches (Ho et al. 2010), the Fuzzy MCDM approaches (FMCDM) are gaining a much increasing importance, thanks to the possibility of tackling conflicting interests in very complex problems (Mardani et al. 2015).
One of the most used method is the Fuzzy Analytic Hierarchy Process (FAHP) (Luthra et al. 2017) that in addition to the advantages of the AHP, integrates the possibility of considering the fuzziness of the data for deciding the preferences of the different criteria listed (Ishizaka 2014). For example, Kahraman et al. (2003), applied FAHP to the selection of the best supplier companies for a white good manufacturer in Turkey, while Chan and Kumar (2007) developed it for a global supplier selection problem of a manufacturing organization. The fuzzy Analytic Network Process (FANP) has also been used several times in literature for supplier selection (Chai et al. 2013). An example is provided in (Vinodh et al. 2011) where a conceptual model for supplier selection has been developed for an Indian manufacturing company of electronics switches.
In addition, the AHP, and even the ANP, integration with other approach is very much popular given its flexibility, its simplicity to be understood and its capability of adaptation to different contexts with several criteria defined (Govindan et al. 2015;Zavadskas et al. 2016; Barrios et al. 2016). Mainly, in these approaches, FAHP has adopted to assess weights of the listed criteria while a second approach that uses those weights is adopted to ranking the suppliers. For example, Sivrikaya et al. (2015) integrated FAHP with a goal programming approach. Furthermore, Kar (2015) employed FAHP to evaluate the group preferences for an Indian manufacturing company integrating it with the fuzzy Neural Network that maps the supplier using the weights generated by the FAHP approach.
In many cases, FAHP was adopted together with TOPSIS or Fuzzy TOPSIS. In this sense, Kannan et al. (2013) used the combined approach to select the best green suppliers for an Iranian automobile manufacturing company, also used in a similar context by (Shen et al. 2013). Sultana et al. (2015) integrated in addition to FAHP and Fuzzy TOP-SIS also Fuzzy Delphi method to identify the most important criterion for a leading multinational manufacturing company of battery for several vehicles in Bangladesh. Similarly, Prakash and Barua (2016) adopted the combinations of these methods for the selection of the reverse logistics partners of an Indian electronics industry.
These MCDM methodologies can be also supported by the DEMATEL and its fuzzy variant FDEMATEL methods that are well established in the field of the supplier selection (see e.g., Chang et al. 2011). Baykasoğlu et al. (2013) integrated FDEMATEL to evaluate the weights of the criteria with TOPSIS method to choose among the alternative supplier of trucks for a land transportation company. In a different context, in particular for a communication company in Turkey, Uygun et al. (2015) studied the interrelationship among the main criteria through the DEMATEL method while local weights of the sub-criteria were generated by a FANP approach on the basis of cause-effect relationships derived with the DEMATEL method. FDEMATEL has been combined also with other methodologies as in (Mehregan et al. 2014) where for the selection of the supplier of the national Iranian gas company it was combined with interpretive structural modelling, or in (Keskin 2015) where is joint with Fuzzy c-means algorithm for a glass makers company. Recently, Lin et al. (2018) proposed an approximate FDEMATEL and applied it to a can factory in Jordan.
Lately, the combination with more than one MCDM method together with the FDEMA-TEL has been proposed. Büyüközkan and Çifçi (2012) applied a hybrid MCDM approach to evaluate green suppliers for an automotive manufacturing company in Turkey. The combined fuzzy ANP and fuzzy DEMATEL method helped to better understand the interdependent relationships within and among the set of criteria while fuzzy TOPSIS method facilitated the choice of the alternatives for the ideal green supplier.
This topic, i.e., the selection of green suppliers and more in general the implementation of the sustainable supply chain is vastly increasing in the last years (Zimmer et al. 2016).
As highlighted by Tseng et al. (2019) there is a rising tendency of applying MCDM models for supporting informed decisions that address environmental performance. The selection of the suppliers becomes a fundamental step for the creation of the sustainable supply chain (Maditati et al. 2018) and a definition of appropriate criteria is essential to create such a type of supply chain (Guarnieri and Trojan 2019). For this reason, we have stressed out in our paper how the selection of supplier in the mining industry should be as sustainable as possible and, therefore, it is essential to take into account sustainability aspects when evaluating suppliers. Indeed, despite the increasing interest in involving sustainable criteria, the development of structured and participative MCDM methods that can tackle supplier selection problems while taking into account sustainable criteria is still scarce.
In this direction, an attempt can be retrieved in Orji and Wei (2014) that applied Fuzzy logic, DEMATEL and TOPSIS to highlight the interdependencies between sustainability criteria in order to select the best sustainable supplier in fuzzy environment for a gear manufacturing company in China. Similarly, Petrudi et al. (2018) investigated interrelationships among sustainability criteria, weighted them, and then evaluated suppliers according to these criteria adopting FDEMATEL, Fuzzy Preference Programming, ANP and finally Fuzzy TOPSIS.
To be precise, the novelty of our approach consists in using a combination of MCDM methods (FAHP and TOPSIS together with FDEMATEL) that integrates the advantages of fuzzy DEMATEL in establishing interrelationships between several criteria, with the strength in using the fuzzy AHP method to prioritize the criteria and finally TOPSIS to rank the different alternatives for ranking supplier selection according to a variety of criteria including the sustainable ones. The only other paper, at the best of our knowledge, that proposed a similar combination of techniques for the supplier selection problems was proposed by Abdulgader et al. (2018). However, they have explained their methodology through a demo application. Instead, our paper tackles a real case from a major company and takes into account sustainability aspects of the problem according to the last trend in the literature. We also interact in a structured and participative way with several stakeholders of the company. Indeed, our approach is original because it considers a novel combination of MCDM methods that has never been applied in the mining industry for the sustainable selection of the suppliers. We aim to show how it can help the decisions makers in such complex decisions and improve the sustainability of the mining industry still optimizing other conflicting objectives at the same time.

Description of the proposed methodology
Multicriteria techniques can solve four different types of problems (Roy 1981): (a) Choice problem: The goal is to select a single best action or to reduce the group of actions to a subset of equivalent or incomparable actions. (b) Sorting problem or similarity analysis problems: Actions are sorted into ordered predefined categories. (c) Ranking problem: Actions are ordered in a decreasing preference. The order can be complete or partial if we consider incomparable actions.
(d) Description problem: The goal is to help the description of actions and their consequences.
The proposed new methodology solves ranking problems and choice problem if we select only the first ranked alternative. A six-phase methodology ( Fig. 1) is proposed to establish the most suitable supplier of forklift filters in the heavy machine sector: Phase 1: a decision-making group is selected based on their experience for solving the problem. Then, the participants will be involved in the decision-making process through FAHP and FDEMATEL methods.
Phase 2: the decision elements (criteria and sub-criteria) are defined based on the experts' opinion and the pertinent scientific literature.
Phase 3: FAHP is applied to calculate the relative importance of criteria and subcriteria under uncertainty. In this step, the participants are asked to provide pairwise comparisons that are then processed according to the FAHP method as described in Sect. 3.2.
Phase 4: FDEMATEL is implemented to determine the dispatchers and receivers as well as the strength of influence between criteria/sub-criteria. Herein, the experts evaluate the influence of each decision element on the others applying the step-by-step procedure explained in Sect. 3.3.
Phase 5: FAHP and FDEMATEL are combined to obtain the final criteria weights (Sect. 3.4). Phase 6: TOPSIS method is applied to rank the suppliers based on their closeness coefficient (Sect. 3.5) and quickly identify the best option for a company using forklift filters.

Fuzzy analytic hierarchy process (FAHP)
Fuzzy sets were introduced in the development of AHP methodology in order to cope with vagueness derived from human judgments, e.g. in the pairwise comparisons (Cebeci 2009). In this regard, Fuzzy sets provide more accuracy to solve real-world MCDM problems, which inexorably entails some degree of noise in their variables. With this modification, AHP becomes FAHP. The comparisons are described by triangular numbers M which are represented by (a, b, c) and the membership function is defined as follows: a and c are the lower and upper bounds whilst, the strongest grade is represented by parameter b. The fuzzy triangular numbers to be used in FAHP and its corresponding reduced version of Saaty natural scale (Saaty 1977) are shown in Table 1. From the Saaty scale, only three points are adopted to facilitate the engagement of unskilled respondents and then reduce inconsistencies in the decision-making process.
The description of the FAHP algorithm is presented by phases: • Phase 1: perform paired judgments for both criteria and subcriteria by using fuzzy triangular numbers described in Table 1. From this, a fuzzy judgment matrix Ã k a ij can be obtained via applying Eq. 1: a k ij denotes the kth expert's preference of ith criterion/sub-criterion over jth criterion/ sub-criterion.
• Phase 2: If there is a committee of experts, the judgments are then aggregated by using Eq. 2, where K is the number of experts participating in the decision-making process. Then, the initial fuzzy judgment matrix is updated as outlined in Eq. 3. • Phase 3: Calculate the geometric mean of fuzzy comparisons r i for each criterion and sub-criterion based on Eq. 4.
• Phase 4: Obtain the fuzzy weights of each criterion and sub-criterion (w i ) using Eq. 5.
• Phase 5: Defuzzify (w i ) by applying the Centre of Area method (Wang et al. 2009) outlined in Eq. 6 where M i is a non-fuzzy number.

Fuzzy decision making trial and evaluation laboratory
Considering crisp values (experts' responses) do not properly represent the real-world context, FDEMATEL has been proposed as a suitable method to assess the interdependence between uncertain criteria/sub-criteria and subsequently identify causal relationships in complex MCDM networks. Fuzzy logic effectively represents the vagueness of human judgments and address the variability of real-world scenarios (Baykasoğlu et al. 2013). This approach also provides digraphs to classify the MCDM elements into two groups (cause and effect) so that focused improvement strategies can be proposed for increasing the suppliers' performance. The steps of FDEMATEL are described as follows: • Phase 1-Use a fuzzy linguistic scale: To incorporate the uncertainty of human comparisons, five linguistic levels are utilized to represent the "influence" between decision elements: No influence, Weak influence, Medium influence, Strong influence, Very strong influence. The comparison is a fuzzy triangular number (l k ij , m k ij , r k ij ) , which represents the kth expert's preference of ith criterion/sub-criterion over jth, as outlined in Table 2. (2) In case of an expert group, the judgments are averaged using Eqs. 8-10, where K is the number of experts.
• Phase 2-Create the initial fuzzy direct-influence matrix: After gathering the experts' judgments, an initial fuzzy direct-influence matrix D can be derived using Eq. 11.
• Phase 3-Normalize the initial fuzzy direct-influence matrix: D matrix is then normalized to obtain Ñ (refer to Eq. 12) • Phase 4-Obtain the fuzzy total-influence matrix: After computing , and N r = e r ij n×n , the fuzzy total-influence matrix T can be achieved applying Eq. 13. In this formula, I represents the identity matrix.
, when e l ij < e m ij < e r ij for any i, j ∈ {1, 2, … , n}. • Phase 5-Calculate the threshold value (α): This value can be estimated using Eq. 14 so that the cause-and-effect diagram can be generated.
The sum of rows and columns are outlined as separate vectors C i and R j correspondingly, where i = j. Prominence values, represented on the horizontal axis, is achieved by summing the above-mentioned vectors. This variable measures the influence of each criterion/sub-criterion on the other decision elements. Likewise, relation values C i −R j , represented on the vertical axis, categorize the decision elements as dispatchers and receivers. If C i −R j is positive, the element is then categorized as a dispatcher; otherwise, it is considered as a receiver.
The next step is then implementing the CFCS (Converting Fuzzy data into Crisp scores) method outlined in Eqs. 15-23 for defuzzifying C i +R j and C i −R j . After this, the cause-and-effect diagram is achieved given the dataset (i) Normalization where (ii) Calculate left (ls) and right (rs) normalized values: Establish the total normalized crisp value: (iv) Calculate the crisp value: (v) Integrate the crisp values:

The FAHP-FDEMATEL approach
In this work, our primary objective is to generate a robust MCDM model for (i) ranking sustainable supplier alternatives of forklift filters and (ii) identifying the deficiencies that should be tackled by each supplier to effectively underpin sustainability. The weights consider both the short term and long term intervention which is consistent with the time horizons specified in the development plans of companies from the heavy machine building industry. A hybrid FAHP-FDEMATEL approach is then proposed to calculate these weights. This method also addresses the weaknesses of FAHP approach which, in a single way, does not evaluate the fuzzy interrelations often found in models supporting sustainable supplier selection (Phochanikorn and Tan 2019; Chen et al. 2020).
The final global criterion (k = 1, 2, …, m) weight GW k , local sub-criterion (i = 1, 2, …, s) weight LW k i , and global sub-criterion (i = 1, 2,…, s) contribution GW i resulting from FAHP-FDEMATEL can be obtained using Eqs. 24, 25, and 26 correspondingly. Here, m represents the total number of criteria while s denotes the total number of sub-criteria considered in the performance assessment model. Č k +Ř k def and Č k −Ř k def correspond to the prominence and relation values of criterion k (k = 1, 2, …, m) while Č i +Ř i def and Č i −Ř i def describe the prominence and relation of sub-criterion i (i = 1, 2,…, s). Equations 24 and 25 combine the weights provided by F-DEMATEL and F-AHP to estimate the criteria and sub-criteria weights on the basis of interdependence. In particular, F-DEMATEL weights for criteria are calculated according to: between dispatchers and receivers) have different interpretations, their dimensionality is equal and can be therefore mathematically integrated as observed in w j formula. The F-DEMATEL weights are later integrated with FAHP weights represented by vector N (Eq. 7). Finally, the denominator in Eqs. 24 and 25 is included for normalizing the criteria and sub-criteria weights. These are the resulting weights on the basis of interdependence that are used during the decision-making process.

Technique for order of preference by similarity to ideal solution
TOPSIS is an MCDM ranking technique that chooses, from a finite set of options, the alternative with the shortest distance from the ideal solution and the farthest distance from the negative ideal solution (Ashtiani et al. 2009;Chai et al. 2013). The positive ideal solution is comprised of all the best criterion/sub-criterion values (A + ); in contrast, the negative scenario represents the worst attribute values (A − ). TOPSIS then calculate a closeness coefficient for each alternative so that the best alternative can be identified quickly (Olson 2004). This method originally works based on criteria/sub-criteria weights subjectively provided by decision makers; however, in this study, they are provided by the FAHP-FDEMATEL approach. The TOPSIS procedure can be expressed in a series of steps: • Step 1: Establish an initial decision matrix X with "p" suppliers of forklift filters and "n" sub-criteria (Eq. 27). x ij denotes the value of the sub-criterion SF j (j = 1, 2, 3, … , n) in each supplier S i (i = 1, 2, … , p) . To do this, performance indicators need to be defined for each sub-criterion and then gathered for each supplier. • Step 2: Standardize X matrix by converting x ij into normalized measures s ij (Eq. 28). Let n ij be the norm used by TOPSIS (Eq. 29). (24) Step 3: Calculate the weighted normalized decision matrix V (Eq. 30). The set of importance weights (w j ) are provided by the FAHP-FDEMATEL approach. • Step 4: Identify the positive (A + ) and reverse (A − ) extreme performance of each criterion/sub-criterion using Eqs. 31 and 32 correspondingly: Step 5: Set the distance measures of each supplier to both ideal (A + ) and nadir (A − ) using the Euclidean distance as outlined in Eqs. 33 and 34.

Distance to (A + )
Distance to (A − ) • Step 6: For each supplier of forklift filters, determine the closeness coefficient (R i ) applying Eq. 35. If R k = 1 , the supplier performance is equal to (A + ). Therefore, high R i values indicate satisfactory supplier's behaviour. • Step 7: Rank the suppliers of forklift filters by maximizing R i values obtained in Step 6.

Background
Selecting a wrong supplier can have many negative effects (Barthélemy 2001;Stringfellow et al. 2008). A company can face unexpected maintenance costs due to poor performing suppliers. If the supply is unreliable, it has a repercussion of the inventory and the production planning. If the quality or the environmental compliance is not at the required standard, we may face a loss of customers. If the innovation capability of the supplier is lo. we may lose in competitiveness. All these negative effects have a high cost and may even lead to the bankruptcy of the company. These are particular concerns of a company from the heavy machine building industry, which outsources the production of forklift filters, components that are used in heavy-duty-forklift trucks for removing pollen, mud, solid particles, dust and other contaminants that may negatively affect the engine combustion system. Given the continuous operation of these machines in warehouses and stockyards for long hours under extreme weather conditions and environmental contaminants (dust and dirt), the filters have to be averagely changed every 300 h. Hence, selecting a suitable supplier of forklift filters is a critical decision for mining companies. In reply, a practical and managerial MCDM approach was proposed to be implemented in this sector. This MCDM method evaluates several criteria to select the best supplier to minimize the risk of post selection problems. A pilot study was then conducted in a Colombian leading firm with more than 10 years of experience and export tradition in North America and Europe. This company is involved in a continuous improvement plan targeting their supply chain performance, which should satisfactorily respond to customer requirements at the lowest cost.

Creation of the decision-making group
The suppliers play a relevant role and should be therefore rigorously selected. This intervention was previously discussed with several managers from the executive committee of the company in order to obtain feedback and formal approval before the project start. The decision-making process was facilitated by two academic researchers who are co-authors of this article. Three supervisors and two assistants that are usually concerned about the supply of forklift filters in this company participated to the selection process. A brief description of the participants' profile is provided below: • Three managers (Head of Production Department, Head of Purchasing Department and Head of Quality Department) with more than 5 years of experience in heavy machine manufacturing. • Two associate professors from industrial engineering departments with extensive knowledge in decision-making methods, supplier selection, and supply chain management. • Two assistants (Production assistant and Quality assistant) with extensive knowledge in logistics and experience (> 3 years) in heavy machine manufacturing.
In detail, the facilitators designed the MCDM network model for selecting the most suitable supplier of forklift filters. They also taught participants how to perform the pairwise comparisons in AHP and DEMATEL methods. In addition, they created a set of metrics supporting TOPSIS application. On the other hand, the managers were invited to be involved in the expert group since they can better define the critical needs that must be fulfilled by supplier partners in order to reduce the maintenance costs throughout the supply chain. Finally, the assistants were asked to participate in the MCDM process since they have continuously monitored and analysed the performance of different types of forklift filters during several periods.

Design of the MCDM network model
The MCDM model was then discussed during two meetings with the participants in order to determine whether it was suitable and comprehensible. The final version of the model is shown in Fig. 2. In this regard, a set of 9 criteria and 29 sub-criteria was defined to rank four potential suppliers (S 1 , S 2 , S 3 , S 4 ). A description of each criterion can be found in Table 3.
In detail, the sub-criteria "OHSAS 18001 CERTIFICATION" and "ISO 14001 CERTI-FICATION" evaluate whether the supplier of forklift filters are certified under these norms. In fact, suppliers can improve their image and credibility given the fact that companies search for partners complying with the regulations and having health and safety commitment. On the other hand, "GREEN POLICY" verifies whether the candidate supplier has adopted regulations supporting the reduction of the potential environmental impact derived from its operations (Khan et al. 2018;Guarnieri and Trojan 2019). This is also related to the "GREEN IMAGE" sub-criterion which denotes the number of green seals obtained by the forklift filter supplier due to for its commitment to the environment (Shen et al. 2013;Orji and Wei 2014;Petrudi et al. 2018;Guarnieri and Trojan 2019). Another aspect of interest is "POST-SALE SERVICE" which measures the response time to warranty claims (e.g. indemnification requests)."LEAD TIME" is also considered in this model to evaluate the expected time elapsed between the request and delivery dates. Of course, the shortest time, the better. "FLEXIBILITY" represents the types of forklift filter that can be offered by a supplier in order to support customers when addressing design and demand changes. Attention is also paid to "DURATION OF THE GUARANTEE". In particular, companies  Recycle and reuse capability (SF29) In this domain, it is intended to establish whether the main components of forklift filters (i.e. raw material, packaging, and labelling) are ecofriendly, that is to say that low impact can be expected along their life cycle (Kannan et al. 2013;Shen et al. 2013;Orji and Wei 2014;Luthra et al. 2017;Petrudi et al. 2018) from the mining sector are interested in partners providing long warranty periods and giving the choice of extension if needed. This aspect, therefore, becomes a competitive advantage for suppliers during any selection process. "COMMERCIAL TERMS" are also included in the supplier selection model since customers are attracted by partners establishing commercial agreements quickly after request. In addition to the aforementioned sub-criteria, "EXPERIENCE IN THE MARKET" has been regarded in this case for determining how stable a supplier is in the business. Given the growing demand, it is necessary to count on with the support of suppliers that can positively collaborate to reduce the bullwhip effect. On the other side, considering the importance of using technology (e.g. inventory reduction, sales augment, financial returns and positive market penetration) for underpinning fast-and-effective commercial supplier-customer relations and transactions, "E-BUSINESS" was also regarded in the MCDM model (Oliveira and Martins 2010). Likewise, it is necessary to verify whether the suppliers had been previously included in sanction lists (SF12) because of an illegal activity (e.g. terrorism financing, narcotics trafficking, human rights violation and money laundering activities) since it will affect the firm reputation and reliability. Furthermore, it is a good practice to take into consideration whether the supplier is ISO 9001 certified to assure it is working with accepted standards and procedures. On the other hand, "FILTER BRANDS" determines the number of forklift filter brands that can be offered by each supplier. Another aspect of importance is "FILTER MATERIAL" which evaluates the durability of each filter material aiming at identifying which is the most suitable for a specific work environment. The companies from the mining sector have understood that fostering innovation activities requires that suppliers bear more design and development responsibility than before (Wagner 2009). Thus, it is important to determine if the supplier of forklift filters has an R&D department to support the design and production of innovative heavy machine. Additionally, the partner should count on with IT infrastructure to support supplier-customer collaboration through effective decision-making and efficient information management. These aspects are complemented by financial sub-criteria that should be considered in every supplier selection process: PRICE COMPETITIVENESS, PAST FINANCE PER-FORMANCE, and PAYMENT DEADLINE. Specifically, the filter price and payment deadline are critical when companies face different competitors in different markets. Therefore, suppliers offering reasonable prices will be taken into account to increase the competitiveness profile. Besides, it is also relevant to determine the PAST FINANCE PERFOR-MANCE of the potential partner in order to assess whether its cash flow is able to support the forecasted demand. Thereby, backorders and sanctions can be avoided due to lack of forklift filters.
Other sustainability sub-criteria were considered as part of the supplier selection model. Some of them are associated with the manufacturing and logistics processes of forklift filters. For instance, "MANAGEMENT OF WASTE AND POLLUTANTS" evaluates whether the supplier usually implements protocols for the correct management of waste and pollutants derived from the production of forklift filter (Orji and Wei 2014;Luthra et al. 2017;Petrudi et al. 2018;Guarnieri and Trojan 2019). It is also of paramount to determine if the supplier uses ENVIRONMENTAL-FRIENDLY technology to increase energy efficiency while minimizing harmful waste (Khan et al. 2018). Besides, it is necessary to check if the supplier has implemented a REVERSE LOGISTICS system for collecting the used filters so that potential deterioration of the surrounding environment can be reduced (Petrudi et al. 2018). Likewise, it is useful to verify whether the suppliers consider the potential environmental impact of forklift filters as a critical design requirement and can be therefore concluded as promoters of GREEN R&D (Ho et al. 2010). On the other hand, there are sustainability aspects specifically inherent to the main components of forklift filters. A sub-criterion considered in this category is POLLUTION FILTRATION which measures the maximum filtration capacity of forklift filters (Kannan et al. 2013). In addition, managers should check the use of ENVIRONMENTAL-FRIENDLY MATE-RIAL to reduce the negative environmental effects of filters after final disposal (Soleimani et al. 2016;Jindal and Sangwan 2017). The use of ENVIRONMENTAL-FRIENDLY PACKAGING is also of interest when selecting sustainable suppliers as suggested by Khan et al. (2018) and Bai and Sarkis (2010). Indeed, alternatives like using raw 100% recycled materials and creating a circular economy have been proposed to reduce the environmental footprint of packaging. In the meantime, the administrators from the mining sector should also verify the adoption of ECO-FRIENDLY LABELLING in forklift filters. This sub-criterion does not only consider sustainable material choices for labels but eco-friendly practices of label printing companies (Ho et al. 2010). Ultimately, it is also useful to evaluate the RECYCLE and REUSE CAPABILITY of the forklift filters offered by the candidate suppliers. This parameter has been found to be critical when tackling the growing problem posed by waste. In the meantime, as recycle and reuse capability increases, the sustainability and competitiveness of the mining companies also augments considering higher lifecycle of filters and lower replacement costs (Orji and Wei 2014; Lopes de Sousa Jabbour et al. 2018).

Linear dependency
In this study, a data-collection tool was implemented to obtain all the pairwise comparisons required by the FAHP method. The objective of this instrument is to propose an easyeffective way to introduce FAHP to survey respondents who are not skilled in complex mathematics. Thereby, the reliability of the decision-making process can be significantly increased. A survey format (refer to Fig. 3) was designed including the following statement: "How much more important is the element on the left column over the element on the right column?" The survey was filled out by the participants using the three-level   (Table 2). An example of a fuzzy judgment matrix is presented in Table 4, where the data obtained from the survey were added and arranged using Eqs. 1-3. Afterwards, the geometric means of fuzzy judgments were calculated for each decision element and the normalized weight values of factors and sub-factors were determined using Eqs. 4 and 5-7, respectively. The resulting global (GW) and local (LW) weights are shown in Table 5. The consistency ratios (CR) were also calculated to verify the reliability of comparisons (refer to Table 6). Since CR values are not higher than 10%, the estimated weights can be employed to set the priority ranking of forklift filter suppliers. In this respect, the participants were neither inconsistent nor random when comparing the criteria and sub-criteria. The assessment process can be thus regarded as satisfactory as well as the reduced FAHP scale and survey layout can be further proposed for replication in real-world scenarios.

Influence among decision elements
In this phase, a survey format was also designed to obtain the FDEMATEL pairwise comparisons (refer to Fig. 4), which will evidence the influence among factors/sub-factors. For each judgment, it was asked: "how much influence each element on the left exerts over the element on the right?" In this case, the survey respondents answered taking into account the five-point scale shown in Table 2. This process was also repeated until completing all the comparisons. Then, the pairwise fuzzy judgments were computed by implementing Eqs. 8-11. An example of a fuzzy direct-influence matrix D is shown in Table 7. After applying Eq. 12, the normalized fuzzy direct-relation matrix Ñ is presented in Table 8. Finally, the fuzzy matrix is obtained using Eq. 13 shows in Table 9.
The prominence C i +R j and relation C i −R j measures are obtained through C i and R j values illustrated in Table 10. Also, the receivers and dispatchers were indicated in Table 10. The results show that Commercial profile (F4) has the highest positive C + R value (6.14), indicating that is the principal generator of effects and the most influencing factor when selecting the best supplier of forklift filters. Therefore, Commercial profile (F4) should be prioritized by the suppliers to improve its performance. This finding is also stressed out by Foss and Jensen (2019) who concluded that the entire organizational structure of suppliers should be properly aligned to their commercial profile so that effective response can be provided to supply chain partners.
The correlation among sub-factors of each cluster was analysed adopting impactdigraph maps (Figs. 5, 6). The blue arrows represent one-direction influences whilst the orange ones denote feedback relationships. The impact-diagram map for factors is shown in Fig. 5a-i. In this case, the threshold value was set as p = 22.784 9 2 = 0.281 after defuzzifying the corresponding matrix T . It can be mentioned that Quality (F4), Innovation (F5) and Green product design (F9) are receivers and the other criteria are dispatchers. Thus, the potential suppliers have to demonstrate its commitment to quality by providing adequate resources, which is accomplished through investments, effective process control, and suitable monitoring. On a different tack, the presence of F5 and F9 among the first-ranked factors evidence the current need for partnering with suppliers designing, producing, and  delivering innovative and eco-friendly forklift filters so that future sustainability in the market can be fully granted while reducing the carbon footprint associated with the lifecycle of this product. On the other hand, impact-digraph maps for the environmental management certification (Fig. 6a) and commercial profile (Fig. 6b) are shown as examples of how to evaluate interrelations among sub-criteria. In this category, the assigned threshold value was set as p = 12.318 4 2 = 0.770 . As a result, ISO 14001 certification (SF2), Green policy (SF3) and Green image (SF4) were classified as dispatchers; whilst OSHAS 18001 certification (SF1) was considered as a receiver. In this respect, SF2, SF3, and SF4 aim to bring a company into alignment with its environmental policy, a primary goal of OHSAS 18001, so that the consequences on the environment can be at least minimized. A growing consensus among suppliers of forklift filters and mining companies is then needed for greater environmental responsiveness (Fryxell and Szeto 2002). Besides, no causal relationships were identified between (SF1) and the rest of F1 sub-factors.
Also, the (D + R, D-R) data set was mapped for commercial profile sub-criteria (Fig. 6b). The established threshold value was set as p = 4.508 4 2 = 0.282 . In this case, Commercial terms (SF9), Experience in the market (SF10) and E-business (SF11) were classified as dispatchers and Inclusion in sanction lists (SF12) was considered as a receiver. In particular, E-business has been found to streamline processes and integrate the supply chain, which ends up reducing logistic costs (Gunasekaran et al. 2008). On the other side, experienced suppliers provide stability to the company since they fully know the market demands and could effectively contribute to the design of customer-oriented heavy machinery. Besides, it can be viewed that (SF10) is affected by (SF9) and vice versa. This relation was also observed between SF11 and SF9.

Final criteria and sub-criteria weights: integrating FAHP and FDEMATEL
Using Eqs. 24-26, the relative weights of criteria and sub-criteria on the basis of interdependence were calculated. Table 11 summarizes the global and local contributions of each decision element after integrating FAHP and FDEMATEL methods.
According to FAHP-FDEMATEL results, Innovation was the criterion with the highest priority (GW = 0.17) when selecting providers of forklift filters. In addition, there is not a significant difference (0.01) between Innovation (1st place) and Commercial profile (2nd place). Hence, these suppliers need to focus on improving these criteria primarily in order to be attractive to manufacturing companies from the mining industry. This is directly Fig. 6 Impact-diagram map for a Environmental management certification and b Commercial profile related to the high-quality standards required by final customers (e.g. construction firms) when acquiring this type of equipment. Of course, given the multi-criteria nature of this decision, the providers cannot discard the rest of the criteria, which should be continuously evaluated for continuous revision and improvement. In Quality cluster, the most important sub-criterion was ISO 9001 (LW = 0.5). This certification was found to be a credible public signal of effective quality management practices in addition to supporting compliance with regulatory and customer requirements (Castka 2018;Blind et al. 2018). Besides, ISO 9001 is considered valuable for standardizing suppliers' processes, which can reduce logistics costs and augment the number of forklift filters conform to specifications. In Innovation cluster, the most relevant sub-criterion was IT infrastructure (LW = 0.87). In this respect, the suppliers should be prepared to effectively contribute to joint process engineering and the design of new heavy machinery through continuous innovation processes responding to the current market needs. It is therefore important that suppliers of forklift filters invest in IT infrastructure to achieve better product performance, reduced development costs and shorter time to market (Lau et al. 2010). The result of efficient supplier-firm integration is then a set of best practices that can be rapidly deployed within the company using suitable IT infrastructure ensuring information sharing and product co-development.
In the Service level domain, Post-sale service was the most crucial element (LW = 0.69). Companies from the mining industry need suppliers providing support services during the life cycle of forklift filters. This implies collecting the returned filters and repairing/replacing the defective units. Inefficient post-sale service may contribute to higher work-in-process inventory, increased number of backorders and cost overruns. For this purpose, firms from this industry are very interested in collaborating with providers having a Post-sale service that can continuously support their operations. Thereby, their supply chain can be more efficient when facing seasonal demand peaks and increasing sales.
Regarding Commercial profile cluster, Experience in the market (LW = 0.89) was the most representative sub-criterion. Mining companies usually export to European, American and Asian markets, which are increasingly demanding with respect to consumption volume and the compliance of quality requirements. To gain a competitive advantage in these markets, experienced suppliers of forklift filters are highly required. These providers play a pivotal role when ensuring the promised machine's lifetime since they fully know the earthwork operations it will perform. They can then design filters effectively removing the contaminants that can be found in the above-mentioned markets so that the engine combustion system can be protected. The reputation of these companies is also acknowledged by international customers and has become a key aspect for deal closing. Finally, although other sub-criteria were taken into account in the supplier selection model, their global contribution to the final decision was not significant (GW < 10%).

TOPSIS method
This phase presents the implementation of the TOPSIS method whose primary aim is to rank the potential suppliers of forklift filters. In addition, the sub-criteria most contributing to the total separation of each supplier can be identified so that improvement strategies can be effectively designed and implemented. First, a metric (Table 12) was defined per each sub-criterion (also for Supplier selection criterion since it has no sub-elements). Then, using Eq. 27, TOPSIS decision matrix D was achieved (Table 13) where suppliers of forklift filters (S 1 , S 2 , S 3 , S 4 ) were related to the sub-factors. Sub-factors' values were computed taking into account the description shown in Table 12.
Positive (A + ) and reverse (A − ) extreme performances were also defined in Table 13a and b using Eqs. 31 and 32 respectively. The normalized TOPSIS decision matrix N is shown in Table 14a and b applying Eqs. 28 and 29. The weighted and normalized TOPSIS decision matrix T (Table 15a and b) was achieved using Eq. 30. The sub-factors' contributions  (0) Environmental friendly packaging Eco-friendly packaging If eco-friendly material is used for filter packaging (1), otherwise (0) Eco-friendly labelling Environmental-friendly labelling If eco-friendly material is used for filter labelling (1), otherwise (0) Recycle and reuse capability Recycle/Reuse capability If the filter can be recycled or reused (2), otherwise (1)   were determined through the hybrid FAHP-FDEMATEL method. Whilst, the separation of each supplier alternative (S 1 , S 2 , S 3 , S 4 ) from the positive ideal solution (S i + ) is inserted in Table 16 after applying Eq. 33. In addition, the contribution of each sub-factor to the total separation from PIS is specified in this table. Similarly, the separation of each supplier from the negative ideal scenario (S i − ) is calculated using Eq. 34 (refer to Table 17). The rank of suppliers of forklift filters and closeness coefficients CC * i is presented in Fig. 7. These values were calculated using Eq. 35. Based on the results, it can be observed that all the suppliers are not good candidates for effectively providing forklift filters according to the minimum score set by the company (CC* < 0.75). In particular, S2 achieved first place with 0.609 while S3 obtained the lowest score (0.401). However, a distance between the current status and the ideal scenario (both positive and negative) can be appreciated in each potential supplier through Tables 16 and Table 17 correspondingly. The separation measures are useful to identify the strengths (sub-criteria/criteria whose distance from A + is equal to 0) and weaknesses of each potential supplier (sub-criteria/criteria whose distance from A + is higher than zero || distance from A − is equal to 0). We particularly focused on detecting the weaknesses of each provider to support continuous improvement and propel suppliers with highest potential for meeting a firm's needs consistently. In detail, S4 does not comply with international OHSAS 14001 regulations (SF2), which may represent significant delays for manufacturers when implementing environmental policies and targeting reduced carbon footprint. Furthermore, it can be appreciated that SF10 (Experience in the market) is equal to NIS since this company is less experienced compared to the others. On a different note, it is noted that S4 neither use eco-friendly technology (SF22; Separation from NIS = 0) in the production process of its forklift filters nor consider environmental design requirements (SF24; Separation from NIS = 0). This highly compromises its eligibility by mining companies targeting reduced carbon footprint. Ultimately, S4 does not have any evidence of the adoption of eco-friendly packaging (SF27; Separation from NIS = 0) and labelling (SF28; Separation from NIS = 0), which end up hindering its access to demanding international markets. It is then of paramount importance to detect the current gaps of its environmental management system so that future sustainability can be further ensured.
On the other hand, S3 offers the longest supply lead-time-SF6 (6 days) which is not attractive for mining companies seeking for reducing the inventory levels and costs throughout the supply chain. In addition, S3 is the provider with the farthest location (SF18 = 56,436 km; Separation from NIS = 0) which may increase its lead-time variability and subsequently augment the idle time in the manufacturing process (Ortiz et al. 2016). Similar to S4, S3 should upgrade the performance of its green management system for effectively contributing to the environmental policy of the mining industry. In this regard, it is advised to adopt sustainable packaging (i.e. biodegradable plastics and plant-based plastics), while reducing packaging needs without compromising the filter protection (SF27; Separation from NIS = 0). Moreover, it is recommended using green stickers reducing the negative environmental impact caused by the current filter design which uses contaminants derived from printing processes (SF28; Separation from NIS = 0). Regarding S2, the main disadvantage that has been detected is the low number of filter brands (SF14 = 3; Separation from NIS = 0) managed by this supplier. In this regard, S2 should establish new branding alliances to provide a wider offer of forklift filters and subsequently support the product portfolio of the manufacturers. Additionally, S2, in spite of its first place in this study, should diminish the response time (SF5 = 3 days; Separation from NIS = 0) for satisfactorily closing a complaint. A delayed feedback or no follow-up may cause possible sanctions, cost overruns, and unsatisfied customers. Moreover, it should foster the implementation of an E-business platform (SF11) so that supplier-customer can be better administered. Finally, S1 should work on the implementation of an R&D department (SF16) in parallel with a CRM software (SF11) enabling the production of forklift filters with better performance, durability and competitive price; thereby, a sustainable supplier-manufacturer relation can be ensured. Moreover, its location (SF18 = 53,478 km; Separation from PIS = 0.00476) may represent a disadvantage for the heavy-machinery manufacturers since the delivery performance could decrease while the logistics and inventory costs may rise. Aside from these aspects, weaknesses were also detected in environment-related sub-criteria. In this respect, it is recommended: (i) creating a green business budget, (ii) setting a green team supporting the design and implementation of green practices within the company (iii) using biodegradable labeling along with recyclable raw materials, and (iv) install energy efficiency systems in the technology used for the production of forklift filters.

Conclusion
Sustainability practices are a necessity for the survival of the planet. If not adopted, the live on earth is in danger of extinction. All sectors are affected and therefore the adoption of sustainable practices has become an important consideration for business organizations, especially with regard to their supply chains. These practices can help them to growth sustainably in the long term. One core practice is the supplier selection, which is key for the improvement of the supply chain performance. Thus, a structured and sustainability focused evaluation approach for supplier selection is needed among business organizations.
In this paper, we have introduced an integrated hybrid model with Fuzzy AHP, Fuzzy DEMATEL, and TOPSIS to evaluate the supplier selection decision with the recommendation of using sustainability factors. An 'hybrid model' is a procedure where the output of one technique becomes the input of a subsequent technique (Stephen and Labib 2018). This is different from 'hybrid modelling', where independent techniques are used to study the same problem in different ways. Hence the paper contributes to hybrid model and complements the application of other hybrid models in other applications (Kumar et al. 2019;López and Ishizaka 2019;Abastante et al. 2018;Krejčí and Ishizaka 2018;Ihrig et al. 2017;Ishizaka and Pereira 2016;Corrente et al. 2016;Ishizaka and Nemery 2013b).
The hybrid model can be extended to all industries irrespective of their nature and location. These models can be applied to supplier evaluation and selection systems along with other selection problems. For companies to attain overall sustainability, incorporating these sustainable suppliers in their supply chains will change the way business is done, which can lead to long time sustainability. However, all the suppliers should be periodically evaluated and measured on sustainable criteria and rated on a performance rating scale. In a future work, we aim to develop this performance evaluation system that could be used to identify and expose poor sustainable performers in the supply chain and help the companies develop better sustainable suppliers.