Related Works

The plant location selection problem is normally considered as a part of supply chain network design. To minimize the total cost as well as determine an optimal flow path for a product, previous studies have focused on demand variations, because the quality of decisions can easily vary due to supply and demand uncertainty, ambiguous information, and various social problems in the global business network. It is likely that a stochastic model, rather than a deterministic approach, can be used to express demand uncertainty. Wang et al. used a stochastic programming model that implies uncertain demand to find a location that maximizes business profits. Amin and Zhang also considered the demand and return uncertainty of a product through the stochastic programming model. Moreover, they included environmental factors, such as the use of eco-friendly materials and clean technology, and used the weighted sum method as well as the ε-constraint method for multiobject optimization. Gülpınar et al. proposed two types of demand distributions (i.e., normal distribution and context intended distribution) regarding facility location in a dynamic environment. Besides that, Wagner and Neshat applied the quantification method to assess supply chain vulnerability. Based on the graph theory, their method of quantifying vulnerability can be dynamically adapted, even if the supply chain is frequently redesigned. In short, the quantification of the supply vulnerability of a food production system must consider production variables such as the properties of food raw materials, changes in production quantities during different seasons, and dynamic market changes.

In addition to fuzzy-based research, several optimization models for supporting decision making have been proposed. Jouzdani et al. proposed a fuzzy model that used a triangular membership function to deal with demand uncertainty in a dairy plant. They considered traffic congestion as an essential factor for selecting the location because the dairy industry is very sensitive to demand variations and the localization of the food industry usually affects supply chain costs. Çebi and Otay proposed a fuzzy-based location selection model for a cement plant by considering various qualitative factors such as availability of resources, strategic factors, government policies, and environmental factors. Mirhadi Fard et al. considered environmental, social, and economic impacts as qualitative decision criteria to choose a sustainable plant location. Moreover, they took into account continuously changing geographic information in the service region and specified spatial characteristics such as accessibility of raw materials and proximity to the market. Rezaei and Zarandi proposed a fuzzy model for dealing with dynamic environments at the initial location of a plant. They also developed a simulation model to recognize any changes in the service region. Moreover, applying seasonal parameters is one solution for ensuring the reliability of a decision model for plant location. Ozgen and Gulsun used triangular possibility distribution (a fundamental part of the possibility theory) to deal with supply and demand uncertainty, along with climate as a seasonal parameter. More specifically, they combined the possibility distribution with the fuzzy analytic hierarchy process (AHP) method to handle both the quantitative and qualitative factors in the decision-making process. However, it is difficult to decide the shape of a membership function for representing the aggregation of data set in fuzzy-based decision-making model; hence the specialist interviews are usually required. Fuzzy TOPSIS approaches have been proposed for selecting a plant location by linguistically evaluating the following criteria: availability of skilled workers, expansion possibility, availability of acquirement material, and investment cost. Aqlan and Lam proposed a fuzzy-based method for supply chain risk assessment and quantified aggregate information, such as expert knowledge, historical data, and supply chain structure, to identify potential risks. Deb and Bhattacharyya proposed a distinct decision support system that uses a multifactor fuzzy inference system for facility layout planning. Dweiri and Meier also applied fuzzy decision making to facility layout planning and used the distance between departments and their relationships for scoring the planned layout.

Askin et al. proposed a genetic algorithm-based method for warehouse location selection and determined the best capacity design for the selected warehouse. They also set the objective function to minimize costs due to demand variations, after which the optimal economic order quantity was derived to continuously adapt to the volatile inventory levels. However, metaheuristic optimization methods such as genetic algorithm-based optimization sometimes require a lot of time to find the optimal solution. Novaes et al. used the Voronoi diagram, useful for conducting spatial analysis, to divide an urban region into service districts. It is important to note that the process parameters of a production system, which determine the productivity of a plant, help decision makers improve the quality of their decision regarding location selection. In this regard, Silva and De La Figuera proposed the integrated approach to find the best plant location using both a stochastic model of a manufacturing system and a deterministic location model. Their study examines the arrival time of customers as well as the processing time and capacity planning of the manufacturing system. Gebennini et al. considered production lead time and delayed quantities of a product to determine demand variations and supply uncertainty. Consequently, in order to make more accurate decisions, various uncertain environmental factors need to be assessed by the appropriate quantification methods.

Vulnerability assessments usually underpin supply chain management due to the quantification of uncertain disturbances for mitigating risk. Albino et al. proposed a quantification method to measure the vulnerability of a production system within a multisupplier network and evaluate critical aspects using two factors, i.e., process uncertainty and product mix variability. Petrovic et al. developed the supply chain simulator to analyze the dynamic behaviour of a serial supply chain in an uncertain environment. For this purpose, they proposed discrete fuzzy sets for modeling uncertain situations in customer demand and external supply to determine the negative effects. Vorst et al. identified sources of uncertainty (e.g., decision process time, order lead time, and order sales period) to improve supply chain performance and validated the trends predicted by the simulation model. Vlajic et al. proposed an integrated framework for guiding food companies, in which supply chain robustness was defined to identify various disturbances through the classified sources of supply chain vulnerability, including external and internal sources that are either controllable or uncontrollable. However, their research mainly focused on internal sources of vulnerability to design robust food supply chains.

In the food industry, since fresh products have a limited shelf life, it is particularly difficult to have many goods in stock at all times. Thus, supply chain management and production planning for fresh products should be carefully conducted when the inventory levels are low. It is important to note that the supply failure of raw materials caused by inaccurate demand predictions and tardiness of finished (or semifinished) products are major factors that trigger vulnerability, which can ultimately disrupt production. Furthermore, as the food industry becomes more globalized, the importance of optimal supply chain management has increasingly been emphasised.

Previous studies have seldom considered an integrated approach for selecting the best plant location using both stochastic simulation and vulnerability quantification, even though many studies have addressed simulation-based optimal layout design. Further, most of the studies considered the supply of raw materials to be relatively stable. Therefore, this study proposes an integrated approach that combines a supply vulnerability analysis and statistical simulation to deal with various uncertain factors (e.g., unstable supply of food raw materials) during plant location selection.