Evaluating the Location of Regional Return Centers in Reverse Logistics Through Integration of GIS, AHP and Integer Programming

International Journal of Industrial Engineering, 22(4), 399-411, 2015

ISSN 1943-670X INTERNATIONAL JOURNAL OF INDUSTRIAL ENGINEERING

EVALUATING THE LOCATION OF REGIONAL RETURN CENTERS IN REVERSE LOGISTICS THROUGH INTEGRATION OF GIS, AHP AND

INTEGER PROGRAMMING A.Zafer Acar1, İsmail Önden2,* and Karahan Kara2

1Department of International Logistics and Transportation

Piri Reis University Istanbul, Turkey

2Department of International Logistics

Okan University, Istanbul, Turkey

*Corresponding author’s e-mail: [email protected] Reverse logistics network problems are accepted as sophisticated research areas in the existing literature due to difficulties of the prediction of the material flow in the network and the conflicting objectives of minimizing total cost and energy consumption while maximizing customer satisfaction and control of the pollution. There are different approaches to dealing with the difficulties of the problem, yet these approaches overlook the geographical aspect of the issues. In this paper, a reversed network of a governmental organization is evaluated. A methodology is proposed that integrates Analytic Hierarchy Process, Geographic Information Systems and integer programming to determine the locations of the return centers and to deal with the complicated structure. Population, airport locations, maritime facilities, railroad lines and highway lines are accepted as the decision criteria. Based on these decision criteria, suitability levels of the candidate cities are calculated. A closed-loop supply chain is taken into consideration and reverse functions of the chain are evaluated with the proposed methodology. Finally, the locations of regional return centers are determined. Keywords: reverse logistics, geographic information systems (GIS), spatial analysis, integer programming, analytic hierarchy process (AHP), location analysis

(Received on April 16, 2013; Accepted on June 21, 2015)

1. INTRODUCTION Reverse logistics is a sophisticated research area in the supply chain management and network approaches. Legislative initiatives, increasing awareness from consumers and companies’ perception of new business opportunities (Salema et al., 2005) increase the importance of this new research topic. Also, reverse logistics reflects a new approach to supply chain. Today, supply chains no longer finish at the point where products reach the customers, but now include returns as well. Returned products can range from disposed products to be recycled to products that are socialsent back due to customers’ dissatisfaction. On the other hand, governmental regulations, social factors and economic concerns make reverse logistics even more interesting area to the researchers. These concerns increase government and private organizations’ interest and improve the role of these organizations (Cruz-Rivera et al., 2009). Moreover, increasing competitiveness, after sales support and returning sales, and green marketing techniques, can force the businesses to focus on reversed operations.

Owing to its economic and environmental benefits, the topic has also attracted organizations such as CLM (Council of Logistics Management) and RLEC (Reverse Logistics Executive Council). This growing interest in the topic has resulted in many scientific articles. Even though the concept of reverse logistics has evolved over time, authors converged on the proposal of Rogers and Tibben-Lembke (2001). This most cited definition of the topic defines reverse logistics as “the process of planning, implementing, and controlling the efficient, cost-effective flow of materials, in-process inventory, finished goods, and related information from the point of consumption to the point of origin for the purpose of recapturing value or proper disposal”. Similarly RLEC argued that reverse logistics is “the process of planning, implementing and controlling backward flows of raw materials, in process inventory, packaging and finished goods, from a manufacturing, distribution or use point, to a point of recovery or point of proper disposal”. In other words, it entails moving goods from their place of use, back to their place of manufacturing for re-processing, re-filling, repairs or recycling / waste disposal (Deloitte, 2014).

Acar et al. Evaluating The Location of Regional Return Centers

400

Reverse logistics, in addition to forward logistics, is a subset of the supply chain. In this case, it starts from the end customer and ends at the origin point. In regard to this context, reverse supply chain management functions can be defined as product collection, reverse logistics, examination & separation and re-distribution (sales) (Krikke et al., 2004). Within the whole reverse supply chain management structure, reverse logistics activities include: processing returned merchandise for reasons such as damage, seasonal, restock, salvage, recall, or excess inventory; recycling packaging materials and reusing containers; reconditioning, remanufacturing and refurbishing products; obsolete equipment disposition; hazardous material programs; asset recovery (Ji, 2006). However, Fleischmann and colleagues (2000) have distinguished three classes of reverse logistics networks based on the form of reprocessing: namely remanufacturing, recycling, and reuse-networks.

Generally reverse logistics can be separated into two main streams depending on its context: economically driven and legislation driven. These two streams have the same sub-sets: reuse, remanufacturing and recycling (Fleischmann, 2001). Moreover, Ji (2006) argued for more streams of reverse logistics such as consumer-driven, market-driven and profit-driven reverse logistics. Those three contexts related reverse logistics types can be sum up as economically driven reverse logistics.

Economically driven reverse logistics networks can be seen in the industries that gain cost advantages as a result of the low raw material prices created by recycling. That’s why the number of successful reverse supply chains are based on purely economic drivers. In particular, many OEM firms establish reverse logistics networks in order to maximize the value recovered through directly using used materials. On the other hand, legislation driven reverse logistics networks concerns supply chains established in response to environmental product take-back legislation. Those reverse logistics networks were mostly established for mandated product take-backs. The execution is typically outsourced to logistics service providers and specialized recycling companies. Because of this, costs are charged to the customers either directly or via the price of new products (Fleischmann, 2001).

At the functional level reverse logistics can be either a subset of closed supply chain management or an independent process (Blumberg, 2004). However, it is often a distinct process, plans and networks of reverse and forward logistics are mixed and cause complexity. The complexity of the problem is caused by the uncertainty of demand or supply sources. There are also conflicting objectives—such as minimizing total distribution cost, maximizing the service level, minimizing the distances to the supply nodes, etc.—that are increasing the complexity level of the problem’s structure. Therefore, dealing with the difficulties requires understanding the reverse logistics environment. There should be some preconditions to build an effective reverse supply chain network. These preconditions are as follows: special distribution centers, separation systems, inventory management systems, timing policies, information flow, flexibility and coordination (Gaurang, 2006). Mathematical models are very important tools for optimizing reversed networks. Those models are used to control the movement of the economically dead products or packages in the network. Mathematical models are also used to reach decisions about remanufacturing, recycling and demolishing.

As mentioned above, with the increasing interest of public and academics in this subject, the number of relevant scientific publications has increased considerably. Despite most of works in the literature reports conceptual results of case studies, some of them proposed general models that were to be applied to a wide variety of reverse logistics network problems. Those studies offered different applications of those models for reverse logistics for different objectives and application areas as well.

In the supply chain analyses, there are strategic, tactical and operational decisions (Wei et al., 2009). Establishing a new facility in the network is a strategic decision due to the difficulty of the structure of the reverse logistics network; hence the decision should be made with care. There are different applications in the existing literature that use mathematical models for reverse logistics for different objectives and application areas (Kroon et al., 1995; Spengler et al., 1997; Hirsch et al., 1998; Hu et al., 2002; Prahinski et al., 2006; Du et al., 2008; Cruz-Rivera et al., 2009). Fleischmann (2001) proposed a linear model for the location of reverse logistic facilities. What those researchers suggest is that the reverse chain must be included in the forward chain but, for other cases, it is better to handle them separately. Their study is accepted as the first study that offers a general model that optimizes reverse and forward networks simultaneously. Jayaraman and colleagues (2003) propose a model for the design of reverse distribution networks based on the warehouse location-allocation model (Jayaraman, et al., 2003). Even though the model is fairly general, it does not consider the forward flow of logistics. Salema and colleagues (2005) proposed a general model that considers the forward and reverse networks simultaneously (Salema et al., 2005). All of these models are based on the warehouse location model, assuming as decision variables the fraction of demand/return served to each customer, through a warehouse/disassembly center and from/to a factory (Salema et al., 2007).

Location studies have focused on spatial aspect concerns such as land suitability. Spatial decision making approaches takes different spatial decision making approaches vary by combining the outputs of several decision criteria to reach the suitability values of the areas under consideration (Joerin et al., 2001; Onden et al., 2012). These studies propose the integration of the GIS’ analytic abilities with operation management tools such as integer optimization and multi-criteria decision making techniques. The approach has been applied to different location studies such as energy plant location selection (Aydin et al., 2013; Uyan 2013), health-care facility location (Vahidnia et al.,2009), waste management (Yesilnacar et al. 2013) and others. There are also existing methodological studies about the integration of the multicriteria approaches with the GIS, and literature studies that index the existing location studies conducted based on these approaches (Jankowski et


401

al. 1994; Joerin et al., 2001; Malczewski 1999, 2006). Due to the suitability calculation capability of this approach, spatial suitability levels of the facilities can be measured based on the methodology. With this approach, proximity calculations can be done even for big data considerations via the GIS, and digitized with the proposed statistical concepts for reversed location selection and network design studies.

Additionally, despite the increasing interest on the topic, most publications are focused on private sectors and economic effects of reverse logistics networks. The public sector is neglected. On the other hand, existing studies approach the existing transportation network with a set of objective functions such as minimization of the total travelled distance, cost, green considerations and others. The context of land suitability in reverse supply chain management is thus neglected. Due to the fact that these studies concern the suitability levels of the alternative areas and integrate the findings with an integer model, the relevant decision criteria and our approach to the solution of the research problem carry a novelty.

Significant amounts of various goods are often procured by the public sector. Then, those goods are distributed to numerous points of consumption that spread all over the country. Nevertheless at the end of the forward logistics process, many used, wrongly distributed, and hazardous goods and packaging materials need to be returned to reduce public budget expenditures. Because of the amount of procurement, the population of the public sector and the geographical width of the country, most of the time it is impossible to return those products to a single centralized return center. Thus, this study focuses on reverse logistics network problems of the public sectors by focusing on a governmental organization. An interdisciplinary approach is developed for locating the Distribution Center/Centralized Return Center (DC/CRC) and Regional Return Center (RRC) facilities. In this framework, to benefit from the integration of the advantages of different analytic tools, to build an efficient reverse supply chain network spatial analysis, network analysis, pairwise comparison and integer mathematical model are integrated to reach proper decision alternatives with a novel methodology for reverse logistics studies. AHP, GIS and integer modeling techniques are used to determine the location decision among the alternatives. AHP is used to determine and evaluate the priority levels of the considered decision criteria. Then, spatial importance levels are calculated according to these priorities. Finally decision alternatives are evaluated with an integer programming (IP) model and the most proper locations and secondary preference decisions are found. The methodology contributes to the literature about how spatial decision criteria can be determined and illustrated on plane in reverse logistics studies. In addition, the novelty in the study area, with the integration of the integer programming with the land suitability concept, this paper also takes a novel multi-criteria spatial decision making approach.

As already mentioned, the literature abounds with papers about reverse logistics, due to the increasing interest of its economic and environmental effects. Those studies in the existing literature of reverse logistics systems are generally based on case studies (Salema et al., 2007). Thus, to fill this gap, this paper proposes a unique methodology for closed reverse supply chain systems to determine the location of reversed return centers. In this study, spatial interactions of the geographic decision criteria are taken into account to measure the flow volumes of the city alternatives. Due to the difficulty of the prediction, it is important to determine flow volumes between cities before the systems’ establishment decision. Existing models in the literature also consider flows between nodes in the reversed systems as parameters for optimization and location decisions. A basic model is proposed for ordering the city alternatives that take into consideration the distances between cities and the central return center that also represents transportation costs. The model does not take notice of the capacity constraint (Fleischmann, 2001) due to the company’s response ability to all kinds of the demand volumes. A three-step methodology is proposed as a decision support system for network design. This methodology has flexible structure for implementation of different kind of reversed systems.

To this end, the following section is devoted to explain the orders of the analysis tools and how to combine these approaches for reaching the preference levels of the return centers. In the third section, an application is given and in the light of the given explanations in the methodology section, the application is implemented for a real-world case. Meanwhile, the findings of the GIS, AHP and integer programming analyses of the proposed model are given in the same section. The summarization, final conclusion, and discussion are given in the last section. 2. METHODOLOGY The integration of these three analyses is proposed to take advantages of each analysis tool in order to represent the whole spatial decision environment and overcome the difficulties of the existing mathematical approaches. The GIS produces geographic information such as distances and attractions based on expert judgments which eliminates necessary calculation in the integer model, and makes the size of the model manageable with the existing solvers. On the other hand, integrating different analysis tools can be interesting to researchers due to the each analysis tool abilities. For instance, AHP allows us to reach criteria priorities with pairwise comparisons instead of direct answers for this consideration. Network analysis is capable of represent the existing traffic network, restrictions, u-turn policy, and so on. Spatial analysis is useful to represent land use and suitability levels, and it is possible to combine outputs of the mentioned analysis tools with integer programming models and optimize for the determined objective(s) due to its flexible structure.


402

Minimizing the total distance between facilities is evaluated under the number of facility constraints. To reach that objective, a methodology that involves the integration of different qualitative analyses is proposed. The first technique is AHP developed for multiple criteria decision making applications (Saaty, 1980) that are used in many cases such as government, education, political, sports and so on (Vaidya et al., 2006). AHP makes use of practice pairwise comparisons between factors and criteria in order to prioritize them. AHP is a useful approach that can combine with other methods to form a hybrid model. The second technique is GIS, which enables analysis of spatial and also non-spatial data. GIS has different analyzing abilities—network and spatial analysis. In addition to the first two approaches, integer programming is a powerful tool for optimization of the supply chain networks. It is also possible to conduct scenario analysis and evaluate different scenarios. Scenario analysis is flexible in terms of integrating the output of the different techniques as the parameters. The proposed methodology consists of three different steps and each step becomes the input of the next. The output of the last step, which is the IP model, represents the result step and it needs to be evaluated before sensitivity analysis. The step-wise methodology of the study is expressed in Figure 1. The following parts of this section are devoted to expressing the methodology and theoretical background of the study.

Figure 1. Proposed methodology 2.1. Evaluation of the Study Area & Prioritization of the Criteria Weights The first phase of the study identifies the decision making process. This step is the starting point of the methodology, and has two parts: the first is clustering the demand nodes into groups, and the second is evaluating the decision criteria. The first part expresses the research of the decision environment of the focal research area such as determining the regulations, interactions between transportation service providers and companies, and so on. Then, with this knowledge, sub-regions of the study area should be determined by the research group. The second part includes determination of the decision criteria and prioritization of these criteria. In this step, decision makers that have experience in reverse supply chain management should work in association with the primary researchers to represent the necessary information.

After determination of the decision environment, the first part of the initial phase of Step I of the methodology. Then, the evaluation process for the criteria weight must be determined. For this determination, we propose using the AHP technique due to its strength in priority calculation. To rank the criteria, the AHP technique is used to determine the priorities of each criterion using pairwise comparisons and a 1-9 scale with the group decision approach. To overcome the consistency problems of the comparisons, consistency test should be performed. For the acceptance of the calculation, λ, consistency index (CI), random index (RI) and consistency ratio (CR) values should be calculated and the results of the CI values should be below 0.1 for acceptance of the comparisons. Because the AHP technique is widely known and used in the literature, we did not express the technique in detail. For the details of the calculation of AHP and the steps of the technique see (Saaty, 1980; Saaty, 1990), and for application examples see (Saaty, 2005). For the definition of the 1-9 scale and the applications of the AHP technique see (Saaty, 1990; Al-Harbi, 2001; Saaty, 2008).


403

2.2. GIS – Network and Land Suitability Analysis The second phase of the methodology is the spatial analysis phase. In this phase, GIS is needed to create geographic information for integer programming. Two analysis tools are needed for geographic information creation. The first is the network analysis for creating the exact distance values between cities based on the road transportation network with the consideration of the rules of this network. The second is the land suitability analysis, which is performed via Spatial Analysis of GIS with the calculated priority values in previous step (AHP). Before performing the geographic analysis that includes the necessary geographic analysis, a geographic database is required to perform both network and spatial analysis.

To create the OD distance values, network analysis should be performed. OD-Cost matrix, which is a known ability of the network analysis, is used to measure the distances between cities to regional return centers and cities to central return centers. This reflects the transportation cost parameters. Spatial analysis is a useful tool that can be used with multicriteria approaches (Din et al., 1992; Adrienko et al., 2001; Chackar et al., 2007). The reflection geographic interaction spatial analysis is proposed to measure the potential flows from cities using the Euclidean distances based on the geographic criteria. The calculated distance-based maps should be reclassified for combining process which is overlaying. Jenk’s reclassification rule uses the breaking values of the spatial distributions of the geographic criteria maps (Jenks 1967). Figure 2 illustrates the Euclidean distance analysis and the overlaying analysis. Due to the classification ability, Jenk’s method is widely used in suitability studies.

Plane Euclidean Distance Reclassification

1 1

1,0 0,0 0,0 1,0 2,0 3,0

Old

ValuesNew Values

5 5 5 5 3 2

1 1,4 1,0 0,0 1,0 2,0 3,0 0-1.05 5 4 5 5 5 3 2

Decision Criteria

2,2 1,4 1,0 1,4 2,2 3,2 1.1-1.55 4 3 4 5 4 3 2

2,0 2,2 2,0 2,2 2,8 3,6 1.6-2.05 3 3 2 3 2 2 1

1,0 1,4 2,2 3,2 3,6 4,2 2.05-3.2 2 5 4 2 2 1 1

1 0,0 1,0 2,0 3,0 4,0 5,0 3.25-5 1 5 5 3 2 1 1

Figure 2. Illustration of the Euclidean distance and overlaying analysis (Gumus et al., 2014) 2.3. Integer Programming The last phase of the methodology is integer programming, which combines the output of the GIS analyses and the optimization approach to make a location decision. The objective of integer programming analysis is to minimize the total travelled distance. This calculation comprises two steps; the first is the distance from the end node to RRC, and second is from the RRCs to the CRC.

The GIS outputs are the parameters of the model as the coefficient of the objective function and the distance matrix, calculated with the network analysis that represents the OD values of the alternative nodes. The spatial suitability map creates information gathered from the spatial analysis. These values are needed as coefficient.

The model takes distances and calculated city priorities into consideration and optimizes for the minimum values of the model that shows the best RRC location alternatives. A basic mathematical model is proposed to show how to integrate the outputs of the network and spatial analysis and determine the best location alternative. The model is given in the following topics of this section. 2.3.1. Notations Indices of the parameters are assigned for cities i, RRCs j and DC/CRC k. Based on these indices, notations of the model are defined as follows:

dij :distance between city i and RRC j djk :distance RRC j and CRC k dik :distance between city i and CRC k ai : spatially attractiveness levels of the city


404

2.3.2. Decision Variable 1-0 binary decision variable, which expresses the decision to build or not, is defined for the construction decision on the DC/CRC and RRCs. xi : 1, if the facility builds in the city alternative i ; otherwise 0. 2.3.3. Objective Function

2.3.4. Constraints

3. REAL-LIFE EXAMPLE OF THE PROPOSED SYSTEM Within the study, location selection of new facilities of a closed loop reverse logistics network is taken into account, and a solution approach is proposed with the integration of three analysis approaches: GIS, AHP and integer programming. To reach this aim, a step-wise methodology is proposed which is illustrated in Figure 1. Each step is designed to create inputs for the following step, and the output of the methodology expresses the selection decision. The application is completed within three steps. The first step is the evaluation of the study area and criteria prioritization. After the first step, the second step is the GIS, which produces geographic information for integer programing model with the input of the AHP results. The final step is the integer programing model that finalizes the study and reaches the location decisions. This section aims to represent how an example case can be solved based on the analysis steps above, and following sub-sections describes an example case for a better understanding. 3.1. Evaluation of the Study Area & Prioritization of the Criteria Weights The evaluation process has two parts. The first is the grouping of the sub-areas of the study area, and the second is the evaluation of the decision criteria. Both of these steps are done by selected experts. In the beginning part of phase I, the study area is separated into four sub-areas: northwest region, southwest region, northeast region and southeast region (distinguished by different colors in Figure 3). The border of the regions is determined by the opinions of the decision makers of the company according to the geographic, demographic and socio-economic variables that constitute the material flow between the cities and the regions. Additionally, the city clusters are accepted according to the terms of the corporation’s network structure. The CRC is located in Ankara because it is the capital of Turkey and the center of the logistics network of the company. The other regions are determined according to the experts’ points of view and the sprawling of the other facilities of the corporation on the analysis area.

. ∗ ∗ ∗ ∗ (1)

∑ 4, ∀ (2)

∑ 1, ∀ 1 (3)

∑ 1, ∀ 2 (4)

∑ 1, ∀ 3 (5)

∑ 1, ∀ 4 (6)

∈ 0,1 (7)


405

Figure 3. Study area and regions

As the second part of Phase I, comparisons are done by experts to determine the priority levels of the spatial decision

criteria. In this step, the pair-wise comparisons are done by experts with the assistance of the research group, who helps with any difficulties or misunderstandings during the comparison process. The Saaty’s 1-9 scale is used for comparisons, and the judgments of the experts and consistency of calculation results are expressed in the Table 1.

Table 1. AHP preferences in pairwise comparison matrices for the decision criteria

Criteria comparison, max λ=5.06, CI= 0.02, RI= 0,89, CR =0.023

Airports Railroad Network Road Network Population Density Maritime PrioritiesAirports 1 0.14 0.13 0.11 0.5 0.042 Railroad Network 5 1 0.5 0.2 7 0.128 Road Network 6 2 1 0,5 7 0.260 Population Density 9 5 2 1 8 0.525 Maritime 2 0.14 0.125 0.11 1 0.044

For the acceptance of the evaluations, λ, Consistency Index (CI), Random Index (RI) and Consistency Ratio (CR) values

are calculated. The results show that the comparison is in an acceptable range and that the priorities are proper to use in spatial analysis. In addition to the consistency, the results show that the most important decision criteria are the population density in the area—then road network, railway network, maritime and airports.

3.2. GIS – Network and Land Suitability Analysis The third phase of the methodology is GIS, which combines the results of the AHP analysis with GIS and produces the input for the integer-programing model. In this step, two different GIS analyses are done; one of them is the network analysis that creates the distance matrix between origin and destination nodes for the integer programing model, and the other is the spatial analysis to create the suitability map. The data used in the analysis is illustrated in Figure 4, and results of the priority calculation are expressed in the Table 1.

For the spatial and network analysis, data creation or collection step is a must for performing the analyses. Due to this necessity, spatial data such as airport nodes, city center locations and borders, motorway lines and rail road lines are gathered from the ESRI’s online data source; the geographic database (geodb) is created based on that data set. Population data is gathered from the Turkish Statistical Institute (TUIK-Turkish Statistical Institute, 2013) and location of the seaports are manipulated in created geodb for the research process. Throughout the creation of geodb, Turkey’s data is clipped from the World scale data set and the data set is converted to the ED_1950 coordinate system for analysis. The created and gathered data is illustrated in Figure 4 to express how the geographic decision criteria are sprawled on the study area.


406

Figure 4. Geographic criteria on the study area

The spatial analysis is done with the Spatial Analyst tool of ESRI’s ArcGIS 10 software and the priorities that are calculated via AHP are combined with the Overlay Tool of the Spatial Analysis. The result map is created based on these priorities. In the spatial analysis, each geographic criterion's distance maps are first created based on the Euclidean distance rule. Then the distance maps are reclassified for the overlaying process. The overlay tool is used to combine the reclassified maps according to the criteria priorities. For more information about spatial analysis and spatial multi-criteria decision-making approaches, see (ESRI 2014; McCoy et al., 2001; Onden et al., 2012). Spatial preferences of the cities are determined according to spatial analysis and the result of this analysis is shown in Figure 5. The preference levels are reclassified into 7 groups where the value of 1 is the best value. The aim of this classification is creating proper interval values that created differences between alternatives for the objective function of the integer model.

Figure 5. Reclassified maps of the geographic decision criteria

The output of the overlay tool, which combines the geographic criteria and population density, is illustrated in Figure 6. The result is gathered in microscopic focus.


407

Figure 6. Result of the spatial analysis

The output of the overlay tool should be generalized to city levels to be used as a parameter in the integer model;

accordingly, an overall value for each city should be calculated. The cities’ average attractiveness levels are calculated based on the normalization principles of the values of the grids in each city. For the calculation, first the sums of the grid values are calculated and are divided into the number of the grids and the city levels, as illustrated in Figure 7. The result is gathered on a 1-6 scale where the 1 shows the highest attractiveness level. Despite the fact that some of the grids have the value of 7, none of the city alternatives obtained a value of 7, and these values are used in the following step, the integer programming model.

Figure 7. City Levels

3.4. Integer Programming Model The IP model is built to reach the best alternatives from the alternative set. This model takes into account spatial attractiveness, which is assumed to be the expression of the suitability level of the each city, distances between facilities in the network and city alternatives. The model’s objective is to minimize the total distance; however, the distance is not the only consideration. The model also takes the classified ranks of the cities into account. The notifications, decision variable, objective function and the constraints are expressed in section 2.3.

The model looks at the locations of the return centers in the alternative set. This involves assumptions in the searching process. Ankara is the capital of Turkey and many governmental organizations’ centers are located in this city. The focused corporation’s main center is also located in Ankara; thus, the DC/CRC should be in Ankara. The distance between cities is calculated according to the city centers. The calculated data are prepared as a matrix and used as parameters in the proposed model. Attractiveness levels of the cities are calculated via spatial analysis based on the priorities that are calculated by AHP with the experts’ point of view. The calculated attractiveness levels are used as ai in the objective function that is expressed in equation 1. According to the objective function, the attractiveness parameter affects the optimal solution as well as the distance values. Thus, the optimal site is calculated according to the distance and also the result of the spatial analysis. Facility counts are accepted as the constraints of the model due to the required number of return centers. Each region should have one RRC. The constraints that express that necessity are explained with the formula 3, 4, 5 and 6. The total number of the CRCs should be 4, as explained in equation 2.


408

A two echelon reversed network is built according to the structure of the corporation’s supply chain and evaluated with the proposed methodology. In the network, return functions are taken into account due to the environment of the processes and the decision of the location of the facilities for collection of the undesirable materials. The first echelon takes transportation distance into account between cities (i) and Regional Return Centers (RRC) (j); the second echelon takes into account the distance between RRCs and Distribution Center/Centralized Return Center (DC/CRC) (k). It is also possible to transport material directly from cities to DC/CRC from the neighbor cities that are illustrated in Figure 8.

Figure 8. Reverse logistics network structure

Spreadsheets are used to solve the problem and the results indicate that the model has a feasible solution. According to the results, each region’s best locations are determined. The results of the model are indicated in Figure 9 based on the proposed model that is illustrated in Figure 1. DC/CRC is located in Ankara since the necessity and the distance are calculated according to that assumption. Under this assumption, RRCs are found in Istanbul for Region 1, Izmir for Region 2, Diyarbakir for Region 3, and Kars for Region 4. The final city decision also indicates that attractiveness levels play an important role to determine the city decision, as expected.

Additionally, to understand the spatial convenience, sensitivity analysis must be done (Delgado et al., 2004). What-if analyses are generally used to evaluate the decision environment, to reach the secondary city decisions and also to test the correctness of the model (Panell, 1997; Delgado et al., 2004). The optimal city locations are taken out of the analysis and the model is solved again. The result shows Canakkale, Gaziantep, Erzincan and Afyon are the secondary preferences for the focused regions which are expressed in the result map in Figure 9.

Figure 9. Results of the model


409

4. DISCUSSSION & CONCLUSION Within this study, a location analysis is done for a reverse logistics system. During the analysis, a public organization’s logistics network is considered and the whole system taken into consideration with the cost minimization objective function. Additionally, geographic aspects of the problem environment are considered and integrated in the current research. With this approach, a methodology focused on evaluating the locations of RRCs in the reverse logistics network is studied and the application showed that the methodology is capable of dealing with the location problems.

A novel methodological approach is suggested for the solution of the considered problem. Integration of the different techniques—GIS/spatial analysis, AHP and integer programming—are used for location selection. To reach the location decision, a step-wise flow is discussed in the paper and expressed in Figure 1. In terms of the proposed methodology, the decision about the location of the facilities can be briefly expressed as follows;

1. In the first step, the city alternatives and decision criteria are determined, and the priorities are calculated with the ability of the AHP technique. The pairwise comparisons of the experts and the priority levels of the decision criteria are given in Table 1.

2. In the second step, the distances between cities are calculated with network analysis and suitability levels are gathered from spatial analysis. Overlaying of the reclassified maps, which are accepted as the flow levels or city suitability of the cities, is calculated with the integration of the GIS/spatial analysis and AHP.

3. In the last step, the integer-programming model is used to combine outputs of GIS/Spatial analysis. The results are obtained from the integer model and the location decision is proposed based on the model, as illustrated in Figure 9. The city alternatives are evaluated with the what-if analysis.

When the existing literature on reverse logistics is deeply inspected, it is seen that most of the studies are performed in different private companies. This study suggests convenient cities for the RRCs depending on transportation abilities such as road, railway, seaway and additionally population. Due to this contribution, it can be said that the proposed methodology and the findings of the study can be focused by the executives of the public sector during the re-designing of their logistics network.

This paper focuses on the geographic aspects of the problem that are missed by previous studies. Euclidean distance analysis based spatial analyses are done for measuring different decision criteria, and finally the suitability levels of the candidate cities. With this approach, geographic conveniences are also considered in addition to cost consideration.

This study is conducted in reverse logistics network design of a governmental organization. Therefore used data are collected from the existing data source and also manipulated. Where these sources are not provided, these sources are based on the opinions of experts in the related area. These can be considered a limitation of the study. The paper and the decision criteria represent the focused study area’s structure. However the sensitivity of the results depends on chosen criteria and the group of experts’ background, experience and knowledge. Thus, the results need to be corroborated by additional inter-subjective studies. In the following studies, the decision environment should be evaluated carefully, and additional criteria should be added for problem solving if needed. As another suggestion for the future studies, the boundaries for the suitability analysis of decision criteria can be researched with different aspects such as hot spot analysis, service area considerations and so on. These new approaches can improve the existing reverse location literature. REFERENCES Al-Harbi, K. M. A.-S. (2001). Application of the AHP in project management. International Journal of Project Management, 19(1) : 19–27. Aydin, N.Y., Kentel, E. and Duzgun, H. S. (2013). GIS-Based Site Selection Methodology for Hybrid Renewable Energy Systems: A Case Study from Western Turkey. Energy Conversion and Management 70: 90–106. Blumberg, D. F. (2004). Introduction to management of reverse logistics and closed loop supply chain processes. CRC Press, Boca Raton, US. Chakhar, S. and Mousseau, V. (2007). Spatial multicriteria decision making. Encyclopedia of Geographical Information

Science, Shekhar, S. & Xiong, H. (eds), Springer-Verlag Berlin Heidelberg. Cruz-Rivera, R. and Ertel, J. (2009). Reverse logistics network design for the collection of end-of-life vehicles in Mexico. European Journal of Operational Research, 196(3): 930–939. Delgado, M. G. and Sendra, J. B. (2004). Sensitivity analysis in multicriteria spatial decision-making: A review. Human and Ecological Risk Assessment: An International Journal, 10(6): 1173–1187.


410

Ding, Y. and Fotheringham, A. S. (1992). The integration of spatial analysis and GIS. Computers, Environment, and Urban Systems, 16: 3–19. Du, F. and Evans, G. W. (2008). A bi-objective reverse logistics network analysis for post-sale service. Computers and Operations Research, 35: 2617–2634. ESRI GIS Resource (2014). Euclidean Distance (Spatial Analyst). Retrieved on 30 March, 2013, from http://resources.arcgis.com/en/help/main/10.1/index.html#//009z0000001p000000. Fleischmann, M. (2001). Reverse Logistics Network Structures and Design. ERIM Report Series Research in Management Reference No. ERS-2001-52-LIS. Retrieved on 30 March, 2013, from http://ssrn.com/abstract=370907. Gaurang, S. P. (2006). A Stochastic Production Cost Model for Remanufacturing Systems. Master’s thesis. Industrial Engineering, University of Texas - Pan American. Gumus, I. and Ismail, O. (2014). Using Spatial Information for International Market Alternatives Ordering, Proceeding of The Fifteenth Annual International Business Research Conference, Jacksonville, USA. Hirsch, B. E., Kuhlmann, T. and Schumacher, J. (1998). Logistics simulation of recycling networks. Computers in Industry, 36(1-2): 31–38. Hu, T. L., Sheu, J. B. and Huang, K. (2002). A reverse logistics cost minimization model for the treatment of hazardous wastes. Transportation Research Part E, 38(6): 457–473. Jankowski, P., Andrienko, N. and Andrienko, G. (2001). Map-centred exploratory approach to multiple criteria spatial decision making. International Journal of Geographical Information Science, 15(2): 101–127. Jankowski, P, and Richard, L. (1994). Integration of GIS-Based Suitability Analysis and Multicriteria Evaluation in a Spatial Decision Support System for Route Selection. Environment and Planning B: Planning and Design, 21(3): 323–40. Jayaraman V., Patterson, R.A. and Rolland, E. (2003). The design of reverse distribution networks: models and solution procedures. European Journal of Operation Research, 150(1): 128–149. Jenks, G.F. (1967). The Data Model Concept in Statistical Mapping. International Yearbook of Cartography, 7: 186–90. Joerin, F., Thériault, M. and Musy, A. (2001). Using GIS and Outranking Multicriteria Analysis for Land-Use Suitability Assessment. International Journal of Geographical Information Science, 15(2): 153–74. Krikke, H., Le Blanc, I. and Van De Velde, S. (2004). Product modularity and the design of closed-loop supply chains. California Management Review, 46(2): 23–29. h Kroon, L. and Vrijens, G. (1995). Returnable containers: An example of reverse logistics. International Journal of Physical Distribution and Logistics Management, 25(2): 56–68. Malczewski, J. (1999). GIS and Multicriteria Decision Analysis, John Wiley & Sons, Inc, New York, USA. Malczewski, J. (2006). GIS Based Multicriteria Decision Analysis: A Survey of the Literature. International Journal of Geographical Information Science, 20(7): 703–26. McCoy, J. and Johnston, K. (2001). ArcGIS 9: Using ArcGIS Spatial Analyst. ESRI, Redlands, USA. Onden, I., Güngör, C. and Şen, A. (2012). Integration of integer programming with GIS analyzing abilities for determining the convenience levels of retail stores. Procedia-Social and Behavioral Sciences, 62: 1144–1150. Panell, J. (1997). Sensitivity analysis of normative economic models: Theoretical framework and practical strategies. Agricultural Economics, 16: 139–152.


411

Patel, M. H., Wei, W., Dessouky, Y., Hao, Z. and Pasakdee, R. (2009). Modeling and solving an integrated supply chain system. International Journal of Industrial Engineering: Theory, Applications and Practice, 16(1): 13-22. Prahinski, C. and Kocabasoglu, C. (2006). Empirical research opportunities in reverse supply chain. The International Journal of Management Science, 34: 519–532. Rogers, D. and Tibben-Lembke, R. (2001). An examination of reverse logistics practices. Journal of Business Logistics, 22(2), 129–148. Saaty, T.L. (1990). How to make a decision: The analytic hierarchy process. European Journal of Operational Research, 48(1): 9–26. DOI: 10.1016/0377-2217(90)90057-I. Saaty, T.L. (1980). The Analytic Hierarchy Process, Planning, Priority Setting, Resource Allocation, McGraw-Hill, New York, US. Saaty, T.L. (2005). Theory and Applications of the Analytic Network Process, RWS Publications Pittsburgh, USA. Saaty, T.L. (2008). Decision making with the analytic hierarchy process. International Journal of Services Sciences, 1(1): 83-98. Salema, M.I.G., Barbosa-Povoa, A. P. and Novais, A. Q. (2005). A warehouse-based design model for reverse logistics. Journal of the Operational Research Society, 57: 615–629. Salema, M.I.G., Barbosa-Povoa, A. P. and Novais, A. Q. (2007). An optimization model for the design of a capacitated multi-product reverse logistics network with uncertainty. European Journal of Operation Research, 179(3): 1063–1077. Spengler, T., Püchert, H., Penkuhn, T. and Rentz, O. (1997). Environmental integrated production and recycling management. European Journal of Operational Research, 97(2): 308–326. TUIK (2013). The Results of Address Based Population Registration System. Retrieved on March 30, 2013, from: http://www.tuik.gov.tr Uyan, M. (2013). GIS-Based Solar Farms Site Selection Using Analytic Hierarchy Process (AHP) in Karapinar Region, Konya/Turkey. Renewable and Sustainable Energy Reviews, 28: 11–17. Vaidya, O.S. and Kumar, S. (2006). Analytic hierarchy process: An overview of applications. European Journal of Operational Research, 169(1): 1-29. Yesilnacar, M.I., Suzen, M.L., Kaya, B.Ş. and Doyuran, .V. (2013). Municipal Solid Waste Landfill Site Selection for the City of Sanliurfa-Turkey: An Example Using MCDA Integrated with GIS. International Journal of Digital Earth, 5(2): 147–64.

Evaluating the Location of Regional Return Centers in Reverse Logistics Through Integration of GIS, AHP and Integer Programming

Documents