In this study, it is aimed to analyze the use of operations research methods in studies to solve the problems examined in the IS. For this purpose, the scope of the study was limited to the Web of Science database and keywords were determined. For the publications research no time interval was imposed. A systematic literature review was conducted with these keywords in order to find answers to the following research questions. The steps of the review procedure are listed in Figure 1.
The studies accessed through the WOS database were examined in order to answer the following research questions. The research questions were determined as follows:
- What are the operations research methods used in solving problems for industrial symbiosis applications?
- What is the relationship between the methods preferred in these studies and the aims of the studies?
- What are the sectors that operations research methods used in industrial symbiosis studies?
In reviewing process, firstly, a search was made from the WOS database with the keyword "industrial symbiosis". 1680 studies were reached in the WOS database with this keyword. This general review reveals that there are quite a number of studies and it shows the necessity of limiting studies. For this reason, the limits of the search were restricted by determining the secondary keywords. The database was searched by adding the terms in Table 2 next to the phrase “industrial symbiosis” in the keywords.
Table 2 The searching terms used in studies
No
|
|
Key term/Search term
|
Number of studies accessed
|
Number of studies included
|
1
|
Deterministic Methods
|
Linear programming/non-linear
|
23
|
11
|
2
|
Mathematical model
|
7
|
6
|
3
|
Integer programming
|
2
|
2
|
4
|
Goal programming
|
1
|
1
|
5
|
Dynamic programming
|
1
|
1
|
6
|
Heuristic Methods
|
Heuristic
|
5
|
4
|
7
|
Meta-heuristic
|
-
|
-
|
8
|
Particle swarm optimization
|
1
|
1
|
9
|
Bee colony or ant colony optimization
|
-
|
-
|
10
|
Genetic algorithm
|
1
|
-
|
11
|
Tabu search
|
-
|
-
|
12
|
Annealing Simulation
|
-
|
-
|
13
|
MCDM Methods
|
Multi criteria decision making
|
6
|
3
|
14
|
Analytic hierarchy process (AHP)
|
6
|
5
|
15
|
Analytic network process (ANP)
|
2
|
2
|
16
|
TOPSIS
|
3
|
1
|
17
|
PROMETHEE
|
-
|
-
|
18
|
VIKOR
|
1
|
-
|
19
|
DEMATEL
|
3
|
2
|
20
|
COPRAS
|
-
|
-
|
21
|
ELECTRE
|
-
|
-
|
22
|
|
Simulation
|
91
|
41
|
Total
|
153
|
80
|
A review procedure was designed with the identified research questions and keywords. Criteria for inclusion and exclusion have been established for the review procedure. These criteria are included in Table 3.
Table 3 Inclusion and exclusion criteria
No
|
Inclusion criteria
|
1
|
The title contains the search terms.
|
2
|
Summary or keywords contain search terms.
|
3
|
Articles, symposium and conference studies are included.
|
4
|
Only English sources are reviewed.
|
|
Exclusion criteria
|
1
|
Unrelated abbreviation matches are excluded.
|
2
|
Collaborations appearing on different search terms are excluded.
|
3
|
Conference, symposium, book, book chapter, and lecture notes are excluded.
|
Operations Research Techniques
Today, service and production systems represent different and complex organizational structures. The differentiation of organizational structures, the increase in specialization and complexity cause decision problems that are difficult to solve. In such an environment, managers need scientific methods for the efficient distribution of resources, taking into account all the criteria that affect decision problems. Because resources are scarce and humans' wants are infinite, the necessity to distribute resources effectively is one of the important reasons for the emergence of operations research (OR) methods (Öztürk, 2011) . Operations research was used during the Second World War to effectively allocate scarce resources to various military movements and services within each movement. The success of OR to develop a system or situation during the war has contributed to the interest of the method in different areas. It is a scientific method that is frequently preferred in decision problems related to the best planning and operation of production and service systems in which resources are shared (Taha, 2000) .
OR is a method that contributes to the optimum operation of the system by modeling deterministic and stochastic (probabilistic) systems and providing solutions to decision problems in these systems. The most important feature of OR is that optimize the system parameters, taking into account the constraints encountered in decision problems (Topuz & Nasuf, 1991) . OR methods were classified by Taha (2000) as deterministic, stochastic and nonlinear models. Linear programming, transportation models, network models, goal programming, integer linear programming, dynamic programming and deterministic stock models are known as deterministic models. Stochastic models include prediction models, decision analysis, game theory, queuing theory, probabilistic dynamic programming, probabilistic stock models, Markov decision processes, and simulation.
Deterministic Methods
Linear programming (LP) is a mathematical modeling method developed to assure optimum use of limited resources. In linear programming, the constraint and objective functions are linear and the decision variables take continuous values. In some decision problems, some or all of the decision variables can take integer values. In this case, integer linear programming (ILP) model is an effective solution technique for decision makers. In ILP, if all the variables take integer values, it is called as integer linear programming, and if some of the variables take integer values, it is called as mixed integer linear programming. Linear programming is concerned with the optimization of a single objective function. If the objectives are more than one and various, the goal programming method, also known as the multi-objective decision-making method, is preferred. Goal programming method is a widely used method among OR techniques. Goal programming offers compromise solutions based on the importance level of each goal, rather than a single solution by optimizing the goals when the system has multiple and conflicting goals. Another method that is widely used in the literature among deterministic OR research techniques is the deterministic dynamic programming method. Dynamic programming is a method for the optimum solution of a problem with n variables, by dividing the problem into n stages and optimizing a single variable subproblem at each stage. One of his important contributions to problem solving is decomposing the problems into stages with the principle of optimum. Since the structure of the stages changes according to the structure of the problem, the calculation details of each stage are designed by the problem solver (Taha, 2000) .
Heuristic Methods
In the literature, it has been seen that heuristic solution methods are used in cases where the structure of the problem does not allow modeling with mathematical models or when the solution times of exact solution methods are not an acceptable for the decision-maker. Heuristic solution methods are examined under two separate titles as special heuristics and meta-heuristics. Special heuristics are the methods in which the solution algorithm is modeled according to the structure of the relevant problem only to solve the problem under consideration.
Meta-heuristic optimization methods are that have search strategies with the ability to solve nonlinear problems with global search processes without initial constraints, avoiding heading towards local optimum points to search the solution space effectively (Merchaoui et al., 2018).
Multi-criteria Decision Making
Multi-criteria decision making (MCDM) is an important tool that helps to solve many problems characterized by multiple alternatives and criteria. MCDM methods are among the operations research methods and frequently used in decision making problems in the literature. MCDM methods are used in decision analysis and there are many types. AHP, ANP, TOPSIS, ELECTRE, PROMETHEE, VIKOR, DEMATEL and COPRAS are mostly used in MCDM methods. The methods can be classified by benefit-based and superiority-based methods. AHP, ANP and TOPSIS methods are known as benefit-based methods, while ELECTRE and PROMETHEE methods are known as superiority-based methods (Ersöz & Kabak, 2010) . AHP and ANP methods compare with pairwise comparison matrices and present the ranking of alternatives according to their importance weights. The ANP method also takes into account the interrelationship and dependency between alternatives and criteria (Ağaç & Baki, 2016) . There are some differences and advantages of their methods compared to each other. For this reason, the decision maker should determine the method to be used according to the structure of the problem.
Simulation
Simulation is a modeling method that imitate the characteristics and behavior of real systems. Simulation method can be used for purposes such as gaining insight and better understanding of a system, comparing various plans and scenarios before implementation, predicting the behavior of a system, aiding decision making, developing new tools for process building and research, and training (Moon, 2017).
In the classification of OR techniques according to solution methods, the basic classification is grouped under three titles: exact solution methods, heuristic methods and simulation. Among the solution methods, one of the methods that decision makers can choose as an alternative in cases where they cannot use exact solution methods and heuristic methods is the simulation method.
The simulation method imitates a real-life system for the efficient use of resources. It is a process that includes designing a model to predict the behavior of the system in a certain period and its real performance features, preparing the experiment, running it, and analyzing the results in order to try different scenarios on this model.
The Operations Research Methods in Industrial Symbiosis Applications
Studies using operations research techniques in industrial symbiosis applications will be summarized in this section. Three subgroups were formed to summarize the studies. According to the solution method, these groups are as follows:
- Studies using multi-criteria decision-making methods
- Studies using deterministic and heuristic methods, and
- Studies using simulation methods
The publication years of the studies are given in Figure 2. The studies using multi-criteria decision-making methods were published in 2015 and later, the studies using deterministic models and heuristic methods were published in 2012 and later, and studies using simulation methods were published in 2008 and later. When the studies are examined according to the publication year, the number of studies has been increased significantly in industrial symbiosis in the last few years. The simulation method has been used more than other methods, especially in the last seven years. It is shown that the interest in simulation method has increased in recent years.
Figure 3 summarizes the number of studies published in journals. When the journals are evaluated according to their subjects, it is seen that environmental journals are preferred more by researchers. The highest number of publications was published in the Journal of Cleaner Production with 29 publications. There are 24 different journals under the “other” title in Figure 3. Journals in which only one study was published about IS are presented under the “other” title. The reviewed 80 studies were published in 34 different journals.
Studies using multi-criteria decision making methods
Studies using multi-criteria decision making methods were searched with the keywords related to MCDM methods in Table 1. As a result of this search, a total of 24 studies were accessed. Since some of these studies were included in two different keywords, duplications and out-of-scope studies were excluded. For example, in the search with the keyword "analytic hierarchy process", one of the studies is found in the keyword "TOPSIS" and the other in the keyword "DEMATEL". When duplicates and out-of-scope studies were excluded, the total number of accessed studies decreased to 13. When the publication years of the studies are examined, there is no study before 2015.
AHP, ANP, DEMATEL, TOPSIS and VIKOR methods are among the methods used as a solution method for multi-criteria decision making methods. Among these methods, the most preferred method is the AHP method. Only one study (Leong et al., 2016) used fuzzy sets. No study is found in which PROMETHEE, COPRAS, and ELECTRE methods were used in the IS.
These are determining the priority values of the criteria affecting the IS network, ranking the alternatives, IS network design, evaluating the effects of IS applications on performance. Under the other title, there are studies carriesd out other than these topics. A summary of the studies reviewed is presented in Table 4. When the sectors of the studies are analyzed, energy (Afshari et al., 2020) , chemical product sector (Teh et al., Eldermann et al., 2017) , water (Leong et al., 2016) and industrial eco-park (Leong et al., 2017) . Studies conducted in the IS with MCDM methods are summarized in Table 4.s
Studies Using Deterministic, Non-Deterministic and Heuristic Methods
In this section, the studies that use the deterministic, non-deterministic and heuristic methods to solve the IS problems are given. The studies are classified under three titles as operational, tactical and strategic decisions in IS applications according to their decision structures. Due to the dynamic nature of the IS application, a significant majority of the studies seek solutions for operational and tactical decision processes. Some studies were included in more than one category. Studies taking into account annual data (Cimren et al., 2011), were evaluated in the group of both operational and tactical decision problems. By-product synergy (BPS) is an industrial ecology practice that involves utilization of industrial by-products as feedstocks for other industrial processes. A novel decision support tool is developed to analyze BPS networks that involve material processing and transport among regional clusters of companies. Mathematical programming techniques are used to determine the optimal network design and the material flows that minimize total cost or environmental impacts. This methodology is incorporated into a graphical software package called Eco-Flow™. The tool has been applied to model and analyze available synergies in an existing BPS network centered in Kansas City, Missouri. A base case, which assumes no synergies, is compared with the optimal BPS solution found by Eco-Flow™. The results for Kansas City suggest that when companies in the network cooperate to optimize the system profitability, up to $15 million per year of savings are possible. The findings also indicate that the BPS approach would result in 29% reduction in total cost, 25.8% reduction in average company cost, 30% reduction in carbon dioxide (CO2) emissions, and 37% reduction in waste to landfill. The modeling approach is being extended to better represent the dynamics of industrial and ecological processes were evaluated in both operational and tactical decision problem groups. There are studies dealing with IS implementation for strategic purposes. For example, Afshari et al. (2020) models IS application for energy efficiency and makes recommendations for energy consumption.
Among the solution methods, linear programming method was preferred more than other methods (Montemanni & Jamal, 2018; Zhou et al., 2012) . Six of the nine studies conducted with linear programming presented a solution to the problem with integer linear programming method. In some of the studies using the integer linear programming method, the problem was modeled using the mixed integer programming method (Afshari et al., 2020; Gonela et al., 2015) . Gonela et al. (2015) modeled the problem as probabilistic with the integer linear programming model, taking into account the probabilistic nature of the problem. In addition to the linear programming method, there are studies in which goal programming (Tiu and Cruz, 2017) and dynamic programming (Suzanne et al., 2020) methods are used. These methods are also known as exact solution methods. Apart from exact solution methods, heuristic methods are also included in the literature. When the studies that use heuristic methods are examined, special heuristics are preferred more than meta-heuristics. In the literature, genetic algorithm, taboo search, bee colony etc. are known as meta-heuristic methods. In the IS, special heuristics are preferred more than meta-heuristics. Particle swarm optimization, which is a meta-heuristic method, was used in only one of the studies conducted with heuristic methods (Ren et al., 2016) . No study was found in which other meta-heuristic methods were used.
When the objective functions of the studies are examined, there are more studies that focus on cost minimization and optimization of environmental effects. Also there are ten studies that aim to profit maximization. Based on the constraints, the supply-demand balance is the main constraint. In addition, there are studies that take into account the constraints that do not allow the purchase and sale of waste materials in the symbiosis network (Montemanni & Jamal, 2018; Yesilkaya et al., 2020) .
Studies Using Simulation Method
As a result of the literature review, the usage of the simulation method is quite common in the IS. When the WOS database was searched with the keyword "simulation", 91 studies were found. When the conference papers and literature review studies were excluded, the number of studies decreased to 60. As a result of the studies, 19 studies that included literature review such as bibliometric analysis and that were not related to the method or industrial symbiosis were excluded. Thus, the number of studies examined was decreased to 41. It has been determined that agent-based simulation, numerical simulation and historical simulation methods are used in studies where simulation method is used. There are studies that also offer a hybrid approach as a solution method. For instance, Yazan et al. (2020) used the simulation and the game theory methods integration.
The aims of the studies are categorized under five classes in Table XX. Studies examining the factors affecting EC such as public tax regulations (Fraccascia et al., 2017) are examined under the title of criterion evaluation (CA). Studies in which a simulation model is established for the installation and execution of the IS network are examined under the network design (ND). Studies examining the changes in the amount of input and output in IS networks and the changes in the production process and resource usage in different scenarios are given in scenario analysis (SA). Studies on profitability and resource efficiency are summarized under the efficiency analysis (EA), and studies investigating the analysis of environmental impacts are summarized under the environmental impacts (EE).
Approximately 30% of the studies examined are for the design of the IS network and the design of industrial areas as eco-industrial parks, 27% for the analysis of the factors affecting the installation process, and 20% for the analysis of environmental effects. When the studies are classified according to sectors, studies on energy and iron and steel sectors are more. Under the energy title, there are IS applications for the coal sector in countries like China where coal resources are abundant, IS applications for heat energy for the recovery of waste heat, and IS applications for energy recovery between urban settlements and industrial areas.
Almost all of the studies that evaluated the criteria in Table 6 make a general assessment of the Eco-industrial parks (EIP) rather than making an assessment according to a specific sector. The reviewed literature is given in Table 6.