Abstract
One of the major challenges in big cities is planning and implementation of an optimized, integrated solid waste management system. This optimization is crucial if environmental problems are to be prevented and the expenses to be reduced. A solid waste management system consists of many stages including collection, transfer and disposal. In this research, an integrated model was proposed and used to optimize two functional elements of municipal solid waste management (storage and collection systems) in the Ahmadabad neighbourhood located in the City of Mashhad – Iran. The integrated model was performed by modelling and solving the location allocation problem and capacitated vehicle routing problem (CVRP) through Geographic Information Systems (GIS). The results showed that the current collection system is not efficient owing to its incompatibility with the existing urban structure and population distribution. Application of the proposed model could significantly improve the storage and collection system. Based on the results of minimizing facilities analyses, scenarios with 100, 150 and 180 m walking distance were considered to find optimal bin locations for Alamdasht, C-metri and Koohsangi. The total number of daily collection tours was reduced to seven as compared to the eight tours carried out in the current system (12.50% reduction). In addition, the total number of required crews was minimized and reduced by 41.70% (24 crews in the current collection system vs 14 in the system provided by the model). The total collection vehicle routing was also optimized such that the total travelled distances during night and day working shifts was cut back by 53%.
Keywords
Introduction
Population growth in urban areas and increase in living standards have caused a significant surge in municipal solid waste (MSW) generation as well as in its environmental, economic and social impacts. In recent years, researchers have carried out comprehensive studies to reduce the adverse effects of MSW through an integrated waste management system (IWMS) (Agamuthu et al., 2009). This system has a multifaceted structure and covers multidisciplinary activities of MSW from generation to disposal steps (Das and Bhattacharyya, 2015; Gallardo et al., 2015). Implementation of IWM systems in urban areas is a crucial step towards achieving sustainable development.
Of the total amount of money spent on solid waste management (for collection, transport, processing, recycling and disposal), approximately 50 to 70% is related to collection activities; that is, mostly procurement of vehicle fleet, collection crews and fuel costs (Tchobanoglous and Kreith, 2002). In addition, the related greenhouse gas emissions, traffic congestion and its consequences are considered as environmental and social costs of MSW collection (Tavares et al., 2009). As such, any measures taken to improve the collection system efficiencies could lead to significant savings in the overall system costs, and considerable reduction in environmental and social impacts (Tchobanoglous and Kreith, 2002).
A wide variety of optimization models has been established for different functional elements of municipal solid waste management systems (MSWMS). Pires et al. (2011) and Xu et al. (2016) have reviewed and reported various engineering models and assessment tools applied in MSWMS in Europe. Some of these models have their main application in the optimization of solid waste collection and transportation systems. Karadimas et al. (2007), Karadimas et al. (2008) and Xu et al. (2016) have optimized collection routing by using operational research models and applying common heuristic and metaheuristic algorithms such as genetic algorithm, ant colony and simulated annealing. Customized evolutionary algorithms have also been developed to solve the vehicle routing problem (Nuortio et al., 2006). In solving an open vehicle routing problem (OVRP), Tarantilis et al. (2004) provided a decision support system (DSS) employing a metaheuristic algorithm called BoneRout. In solving routing problems, optimization is usually performed by two steps. In the first step the conditions of the case study is simulated in vehicle routing problem (Laporte, 1992), and then the problem is solved by using an intelligent algorithms such as Tabu-Search (Glover, 1989, 1990).
Due to the extensive spatial data required in running routing models, it is recommended to take advantage of remote sensing (RS) and geographical information system (GIS) (Agamuthu et al., 2009; Khan and Samadder, 2016). GIS can handle and monitor spatial and non-spatial data in proper platforms, simultaneously (Keenan, 2008; Tavares et al., 2009). Different functional elements of MSWMS have been optimized by using GIS and other mathematical techniques (Chalkias and Lasaridi, 2011). The main application of such coherent structures was reported in optimization of sites for landfills, incineration plants and transfer stations, which were accomplished by incorporating multiple-criteria decision-making models (MCDM) into GIS (Beskese et al., 2015; Chang et al., 2008; Kara and Doratli, 2012; Moeinaddini et al., 2010; Şener et al., 2010; Wang et al., 2009).
The optimization of waste storage bin locations and vehicle routings is another achievement which can be acquired by combining mathematical algorithms with GIS. Ghose et al. (2006) and Xue et al. (2015) based on GIS, provided a model for MSW collection system with the purpose of optimizing distribution of waste storage bins, load balancing of vehicles and vehicle routing. Zsigraiova et al. (2013) have proposed an innovative methodology in order to control and optimize operating costs related to collection crews, vehicles maintenance, fuel consumption and pollutant emissions. Its innovation lay in the point that it considered not only vehicle routing optimization but also waste collection scheduling. Tavares et al. (2009) have developed a model based on three-dimensional (3D) GIS, in which the relief of terrain and fuel consumption were the main criteria in selecting a suitable cost function to optimize vehicle routing. Zamorano et al. (2009), using GIS, assessed the existing locations of waste storage bins and waste collection routing in Churriana de la Vega (Granada, Spain). Then, through reallocation of bins, they optimized their locations as well as waste collection routing. Khan and Samadder (2016) presented an effective model to allocate solid waste collection bin at appropriate places which are not only with uniform distance but also easily accessible for collection vehicle.
In this article, a GIS-based model is presented to optimize both storage bins locations and vehicle routing system, by solving location allocation and capacitated vehicle routing problem. The model was applied to the Ahmadabad neighbourhood, in the City of Mashhad, with the aims of increasing waste collection efficiencies and reducing collection costs. For this purpose, the potential locations for bins allocation were identified and then, based on population distribution, per capita solid waste generation and road network constraints, the optimum locations were determined. Finally, considering road network and limited capacity of vehicles, the collection distance travelled was optimized.
Description of the study area
The Ahmadabad neighbourhood, located in zone one of region one in the City of Mashhad, was selected as a case study in this research. Mashhad is located in the north-east of Iran and is the second most populated city in the country. The area of the city is approximately 288,650 square kilometres and includes 13 regions, 42 zones and 150 neighbourhoods. Mashhad population is estimated to be 2.7 million. The average annual rate of solid waste production in this city is about 610,000 tonnes. Region one, in the City of Mashhad, has an area of 14,978,674 square meters and includes three zones, 11 neighbourhoods and 176,104 residents. Ahmadabad, with an area of 2,622,101 square meters is the largest neighbourhood located in zone one of region one. The location of this neighbourhood is shown in Figure 1.
Location map of Ahmadabad (Mashhad, Iran).
Current collection system
The current residential solid waste collection system used in the City of Mashhad, is a curb type system. Although solid wastes are collected by rear-loading vehicles, owing to the incompatibility of the waste storage containers (which are usually small to medium sized bags used by dwellers) with the collection vehicles, mechanical collection is rarely used. Two crew members usually accompany each vehicle. They collect and transfer residential wastes to the collection vehicle. Waste collection is accomplished daily by private contractors who are supervised by municipality.
Ahmadabad, the largest neighbourhood of zone one, consists of three operating districts including Alamdasht, C-metri and Koohasngi. A fleet of three mechanized collection vehicles, each with the capacity of 4000 kg (8 m3) provides services to these districts (one vehicle for each district). The number of daily tours performed for each district on a regular basis is two. However, as shown in Table 1, two additional tours are also performed, randomly. Therefore, the total number of tours in the Ahmadabad neighbourhood comes to eight. Each vehicle visits its district twice per night (two tours per night shift from 9:00 p.m. to 7:00 a.m.). But the two random tours occur during the day and by one of the vehicles. After finishing each tour, MSW is transferred to Koodbar landfill located in Mashhad-Neishaboor road.
Current status quo of collection system in the Ahmadabad neighbourhood.
These two daily tours are randomly accomplished and as such allocation of any specific route to them was not possible. Therefore, owing to their daily path variation, the average of their travelled distances was considered in this research.
Methodology
The purpose of this paper is to propose a model which was used to improve some components of MSW collection system (storage bins distribution and vehicle routing) in the Ahmadabad neighbourhood in the City of Mashhad. To reach this aim, first, the potential points for the locations of storage bins were determined and then the optimal locations were selected through solving the location allocation problem. In the following step, considering the optimal locations of the containers, the waste collection routing was modelled and optimized based on the capacitated vehicle routing problem. The research was performed in three succeeding stages (Figure 2) including: (i) collecting necessary spatial and non-spatial data; (ii) processing collected data and importing to spatial database (SDB); (iii) establishing rational relations between all groups of information through network dataset to analyse problems and solve location allocation and vehicle routing problems in order to achieve optimal storage bin locations and tours.
Components of the proposed model and the stages followed in the research. Step 1 collecting necessary data including road network, population and municipal solid waste management systems (MSWMS); step 2 import processed data to spatial database (SDB); and step 3 make systematic relation between information groups through network dataset to analyse location allocation problems and capacitated vehicle routing problems (CVRP) with different approaches.
A network is composed of connected components including nodes, edges and the weights devoted to each edge (Tavares et al., 2009). By creating a network dataset that covers various source information, the problem can be analysed and attributes such as impedances, restrictions and hierarchies can be defined. These network attributes that are specified for each network element, control local traversability. Travelling time, fuel consumption, vehicle restrictions in specific roads, speed limit and one-way streets are some of the examples of such attributes. In network analysis, optimum routes are found by minimizing the objective impedances, e.g. finding the quickest route (minimizing time) or the shortest route (minimizing distance).
Data collection and process
The required information on the current status of solid waste collection system, road maps, district population, etc. was obtained by various means, such as preparing proper questionnaires and making interviews, applying the latest census report and using municipal reports, municipal data bank and AVL (automatic vehicle location) system records. AVL is a type of monitoring equipment which monitors municipal vehicles and records travelled distances, speed of vehicles, vehicle positions and stop times.
Road network data
Data on urban road classifications (highways, major arterial roads, minor arterial roads, collectors and local streets) as well as non-spatial data on street names, speed limits and other restrictions such as one-way streets, were necessary in creating a proper road network system and using them in the models. These data were extracted from the municipal data bank.
Population data
Using equations (1) and (2), the distribution of population between the buildings in the studied area was calculated based on the current information on total residential population, the number of blocks in each district, the number of buildings in each block and the number of stories in each building. Population data were obtained from census reports. The number of stories of each building was assumed to be in proportion to its population
In these equations; Ps represents average population allocated to each story, Pb expresses population of each block and Ns is the total number of stories in each block. The number of stories in each building and the average number of inhabitants in each building is represented by Nsb and Pbu respectively. The procedure through which population information is imported to the model is illustrated in Figure 3.
Procedure of population distribution and allocation to each building. (a) Population block derived from population layer, (b) buildings position in each block, (c) merging information of population layer and building positions layer, (d) demand locations representing population coordination in location allocation problem.
Municipal solid waste management system data
Data of MSWMS including the number of vehicle fleet, depot coordination and disposal site position, start and finish times of each tour, collection routes and travelled distances, the number of tours and the amount of waste collected in each tour were collected by using questionnaire method. The collected data and information then were scrutinized and compared to the AVL records. Eventually, current collection routes were implemented in GIS to determine the daily distances which are travelled by each vehicle in current collection system. The information of current collection routes is elaborated in Table 1.
Table 2 shows some of the characteristics of solid waste generated in the Ahmadabad neighbourhood.
Some characteristics of municipal solid waste (MSW) generated in the study area.
Data were obtained from municipal reports.
Based on the data presented in Table 2 and using equations (3) and (4), the total amount of solid waste generated in each district was calculated
In equation (3); MSWp is the total weight of MSW generated in the district, Pd represents population of the district and ω is the daily per capita solid waste generation. In equation (4); VMSWp indicates the solid waste volume generated, and α1 and α2 express the specific weights of dry and wet fractions of solid waste, respectively. The percent dry and wet fractions of solid waste are presented by β1 and β2 which are about 35 and 65%, respectively. Consequently, Table 3 represents population and MSW generation of each district.
Population and MSW generation of each district.
Network analysis model
To develop a comprehensive model capable of analysing the research problems, ESRI ArcGIS network analysis extension was used. The hierarchy of road network was determined and restrictions consisting of speed limits and one-way streets were implemented and devoted to each spatial element via an attribute table. Because all network elements (roads and intersections) were in the same connectivity group (road network of Mashhad), connectivity policy was used to implement connectivity rules on geometric coincidences of line endpoints, line vertices and points. The connectivity policy used for the highways and major roads sources was ‘End Point’ in order to model crossing objects, such as bridges. However, the local streets source is assigned ‘Any Vertex’ connectivity to allow street features to connect to other street features at coincident vertices. Moreover, a turn feature class was created to impose the movement of vehicles in primary junctions and squares.
Mathematical model
As mentioned, at the time of writing, solid waste is collected daily by the curb collection method. This method is mainly recommended for low-rise detached dwellings, but for the low- and medium-rise apartments, which are the case in this research study area, shared containers (bins) should be used to store residential solid waste (Tchobanoglous and Kreith, 2002). As a result, to improve the current collection system in the study area, storage bins were considered in this research. These storage bins can then be mechanically emptied to the collection vehicles. Potential bin locations were considered in accordance with some recommendations made in the literature, such as allocating bins on the road network (intersections are preferable); installing new bins near existing bins, allowing placement of more than one bin at the same intersection and locating bins only on passable streets (Chalkias and Lasaridi, 2009; Zamorano et al., 2009). The optimization of bin locations and vehicle routings then were performed on the basis of the location allocation problem and capacitated vehicle routing problem, respectively.
Location allocation problem
The location allocation problem is one of the most important problems in operational research. This problem is used to optimize various networks such as transportation and communication, increase operational efficiencies and decline economic costs. The problem is applied to optimize inefficient models by reducing peer-to-peer connections and decentralization by using some hubs instead of one centre. In such models, appropriate nodes are chosen as hubs and other nodes are allocated to them. The chosen node combination that minimizes the operating costs is considered as the optimal solution (Cooper, 1963).
ArcGIS network analysis extension can provide various models to support different objective functions including minimizing impedance (P-Median), maximizing coverage, Minimize Facilities, maximize attendance, etc., with different variables and restrictions. In this study, Minimize Facilities function, calculating the optimum storage bin numbers and locations (based on the defined conditions) was applied.
The Minimize Facilities model was run for each district under investigation, using three different maximum walking distances of the inhabitants to the waste collection bins including 100, 150, and 180 m. This distance is the maximum distance considered between the dwellers’ residency and the storage bins that the residents should cover to dump their waste bags to the storage bins. The first option, 100 m, was applied in accordance with recommended values in previous studies, but the second and third options (150 and 180 m) were chosen based on considerations made on the road network and urban pattern in the study areas.
Capacitated vehicle routing problem
Following optimization of the locations of storage bins and based on the obtained results, waste collection routes were optimized by using a capacitated vehicle routing problem model. The model includes a central depot providing all client demands and basically determines which storage bins should be serviced by each vehicle and in what sequence the orders (bins) should be visited. The main purposes are to service the best orders and to minimize total travelled distance, total collection time and the overall operating costs (e.g. fuel consumption, vehicle maintenance, the number of crews) for the fleet of vehicles. To adapt the model to the real-life situations, constraints such as vehicle capacity limits, total distance driven, the number of hours a driver can work and time window can be imposed.
The literature demonstrates that the amount of savings achieved by the vehicle routing problem, is restricted to how the objective function is defined. Although most studies in literature claim a better result when the time minimizing function (in comparison to travelled distance) is used, estimation of travel time for each road network segment is too complicated. Travel time is also a relative parameter which depends on other parameters such as free-flow speed, traffic conditions and lane capacity Tarh-e-Haftom (2010). Due to the unavailability of the required data necessary to calculate travels times in the study area, travelled distance was chosen as the primary objective to be optimized. The operational research model for such problem was carefully elaborated by Toth and Vigo (2001).
The capacity of each vehicle and maximum stations of each tour were considered 20 m3 and 25 stops, respectively. This transferable volume for each vehicle is calculated according to the amount of non-compacted commingled waste stored in containers by inhabitants. Moreover, because of shortening travelled distance, the restriction of one-way streets is not observed by drivers in the current collection system, which leads to traffic congestion and disruption in some cases. In the proposed model, not only the one-way streets but also other traffic rules such as speed limit are considered.
Result and discussion
The results of this research are presented and discussed in two parts. The first part summarizes the results of the Minimize Facilities model for each district. The second part, however, shows the results of optimizing waste collection routes, obtained by solving the capacitated vehicle routing problem (CVRP).
Minimize Facilities results
As mentioned before, applying all the constraints, the Minimize Facilities model was run for each study area (Alamdasht, C-metri and Koohsangi districts), for three different maximum walking distances of the inhabitants to waste collection bins of 100, 150 and 180 m. The results of the model are shown in the following sections.
Alamdasht district
The number of buildings and the daily solid waste production in this district are 1190 and 27.63 m3 (Table 3), respectively. The number of candidate facilities (bins) considered for this area was 115. The outcomes of the model for each of the three walking distances are presented in Table 4. The results include the number of selected facilities (bins), the number of buildings (demands) and the volume of the waste that are covered. Also, as an example, the number and locations of candidate facilities as well as the optimized number and locations of the facilities for each of the distance alternatives are also shown in Figure 4.
The results of the Minimize Facilities model, Alamdasht for which the number of candidate facilities, total demands (number of buildings and total waste generation) were 115, 1190 and 27.63 m3, respectively.
Results of the Minimizing Facilities model for the Alamdasht district: (a) candidate locations, (b), (c) and (d); the selected number of locations for 100 m, 150 m and 180 m walking distances, respectively.
Based on the results illustrated in Table 4 and Figure 4(a), the worst scenario is 100 m distance. In accordance with 180 m distance, although the results appear to be better than the ones obtained for other distances in regard to the number of demands covered, the walking distance of 150 m was considered the optimal. This is because 26 bin locations associated with this distance can provide a better bin distribution than 18 locations related to 180 m distance. Moreover, the 1.1 m3 capacity of each bin, indicates that to store 27 m3 daily solid waste in the case of 18 stations requires that in most locations more than one bin should be installed to reduce the risk of overloading. However, this cannot be practiced due to the local urban pattern.
C-metri district
As shown in Table 3, municipal solid waste generation in this district is 33.2 m3 day-1 and its total number of building is 1356. The number of candidate facility (bin) allocated to this area was 65. The results of the Minimize Facilities model for this district, based on the three walking distances of 100, 150, and 180 m are presented in Table 5.
The results of the Minimize Facilities model, C-metri for which the number of candidate facilities, total demands (number of buildings and total waste generation) were, 65, 1356 and 33.20 m3, respectively.
Comparing the numbers presented in Table 5 reveals that the 180 m walking distance with the minimum number of bin locations (26) gives the best results in regard to the relative numbers of buildings (1348) and the amount of waste (32.96 m3 day−1) that is covered. Therefore, this distance is selected as the optimum solution. Furthermore, the maximum distance between the bin locations occurs when 180 m distance is considered. This is advantageous for C-metri district because of the existing commercial complexes, serious traffic congestions, special urban pattern and the long distances between the junctions (250 m on an average).
Koohsangi district
The number of buildings in the Koohsangi district is 2336 and the average amount of solid waste production is 64.29 m3 day−1 (Table 3). A total number of 150 candidate facilities were considered for the area. Running the Minimize Facilities model for this district, gave the results which are shown in Table 6. Based on the results illustrated in Table 6, the scenario with 150 m walking distance provides the best results and as such is selected as the optimum choice for the Koohsangi district. This option covers almost 100% of total demand (2334 out of 2336) and covers all the daily waste generation (64.2 m3 day−1). Even though 180 m walking distance also provides acceptable results, but with 1.1 m3 bin capacity and 35 locations, allocation of more than one bin in many locations is unavoidable. Considering the urban pattern of the district, this is very hard to apply.
The results of the Minimize Facilities model, Koohsangi, for which the number of candidate facilities, total demands (number of buildings and total waste generation) were, 150, 2336 and 64.29 m3, respectively.
CVRP result
Subsequent to optimization of bin locations for each district, waste collection routes were optimized by using a CVRP model. Based on the selected optimized bin locations, the CVRP model was developed for each of the three districts under investigation. The model results which cover the optimum number of tours and the tours’ specifications for each district, are shown in Table 7. In addition, a comparison between the tours in the current collection system and the ones obtained by the CVRP model are illustrated in Figure 5 for Alamdasht.
Results of solving capacitated vehicle routing problem (CVRP).
Results of the capacitated vehicle routing problem (CVRP) model for the Alamdasht district as compared to the current collection system: (a) and (b) show tour 1 and tour 2, respectively, in the current collection system, while (c) and (d) present tour 1 and tour 2, resulting from the proposed model.
From Table 7, the results of the models indicate that two of the locations in Koohsangi’s district will not be covered by the tours in this district, due to the fact that the capacities of the vehicles are being fully used up. Because the daily per capita waste generation considered in this study was in its highest range, such a condition is predicted to rarely happen. However, to solve the problem and to ensure the success of the plan, two solutions can be suggested: (1) the allocated collection vehicle to Koohsangi district is equipped with a higher capacity (more than 8 m3 of current vehicle capacity), and (2) these two locations are serviced by either tour 1 of Alamdasht or tour 2 of C-metri districts.
A comparison between the current collection system and the system resulting from the proposed model of this study, in regard to the total distance travelled by the collection vehicles and the total number of required crews for the entire Ahmadabad neighbourhood (Alamdasht, C-metri and Koohsangi districts), is shown in Table 8.
Current collection system compared to the model proposed.
In the current collection system, as mentioned above, each vehicle is accompanied by three crew members (a driver and two labourers), indicating the necessity of providing a total number of 24 crew members for the collection of the daily waste generated in the Ahmadabad neighbourhood. In comparison, in the proposed model, owing to the alteration of the current collection system to a system consisting of storage and collection, two crew members (a driver and a labourer) are sufficient to accompany each vehicle. The labourer transfers the bins to the collection vehicles which are equipped with a mechanical loading mechanism Therefore, the total number of required crew members is reduced to 14, approximately 41.70% reduction in the number of crew members. Also, as the results show (Table 8), a significant reduction in the distances travelled by the collection vehicles occurs in the system optimized by the proposed model. As compared to the current collection system, the reduction in the travelled distances adds up to approximately 48% and 57% during night and daytime shifts, respectively. These remarkable reductions in the number of crew members and the travelled distances result in significant savings concerning collection times and collection costs such as fuel consumption costs, vehicle maintenance costs, crew’s expenses, etc.
Conclusion
This research was carried out to provide an integrated model to optimized waste storage and collection system including; number and locations of storage bins as well as collection vehicle routing, in the Ahmadabad neighbourhood located in the City of Mashhad – Iran. The model which was based on solving location allocation and capacitated vehicle routing problems, was developed and executed in GIS. On the basis of the results, significant improvements and savings were made through proper application of the model. The total number of crew members was reduced from 24 in current collection system to 14 in the improved system offered by the model (a reduction of 41.70%). The integrated model also provided a cut back of 12.50% in the total number of daily collection tours (from eight in the current system to seven in the improved one) as well as 53% in total travelled distances during night and day shifts.
Footnotes
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
