Optimal Preventive Maintenance, Repair, and Replacement Program for Catch Basins to Reduce Urban Flooding: Integrating Agent-Based Modeling and Monte Carlo Simulation

Abstract

Urban sprawl has resulted in great losses of vegetation areas, an increase in impervious surfaces, and consequently the direct flow of stormwater into stream channels (i.e., the immediate flow of stormwater into stream channels, in comparison to the indirect flow that is represented by practices aiming to retain stormwater for a certain period of time and treat the polluted stormwater prior to flowing into the stream channels such as detention/retention basins, among others). Stormwater management systems such as catch basins (CBs) are needed to reduce the effect of stormwater runoff. Preventative maintenance, repair, and replacement of CBs are critical to achieve stormwater management best practices. Those practices prevent the blockage of the stormwater system, limit the pollutants in storm sewers, and reduce the risk of flooding. However, no preceding research studies have been conducted to model and simulate the serviceability of CBs and to determine optimal strategies for operating CBs. To that extent, this study establishes a framework to develop and validate an optimal and adaptive maintenance, repair, and overhaul (MRO) strategy for CBs. In relation to that, an agent-based model (ABM) integrated with Monte Carlo simulation was developed for all 560 CBs in New York City’s District 5 and was statistically validated using 99% confidence intervals. The MRO parameters were optimized to minimize the total cost of the system and attain the desired level of serviceability of CBs. Sensitivity analysis was conducted to guide the maintenance planning process of CBs and reveal the effect of the input parameters on the model’s behavior. In addition, ten thousand Monte Carlo iterations were simulated to derive the distributions of the defined parameters. The results proved that in order to minimize the overall cost of repair, maintenance, and replacement of CBs and attain a minimum serviceability threshold of 80%, the following optimal MRO policy needs to be implemented: having seven service crews (where service crews are human resources (i.e., MRO teams) needed to perform the required maintenance, repair, and replacement work), implementing a replacing policy, and replacing CBs after five maintenance periods. The findings revealed that the service crews represent the most critical parameter in affecting the total cost and serviceability of CBs. This research contributes to the existing literature by offering a better knowledge of the management process of CBs and devising optimal MRO strategies for properly operating them. Ultimately, this research helps decision-makers and engineers increase the lifespan of CBs and limit their risks of breakdown, increase their efficiency, and avoid unnecessary costs. The proposed model is flexible and can be implemented to any geographical area and with other model/system parameters, which makes it adaptive for any scenario and area presented by the user. Finally, maintaining stormwater management practices helps in protecting the environment by decreasing the demand on stormwater systems, reducing flooding, protecting people and properties, promoting healthier rivers, and consequently creating more sustainable communities.

1. Introduction

Urbanization is continuing to grow at an extraordinary pace, where it is expected to rise from 77% in year 2000 to 84% in the year 2030 in developed countries [1]. In fact, urban sprawl, industrial growth, and infrastructure development [2] have resulted in great losses of vegetation areas, an increase in impervious surfaces, and consequently the direct flow of stormwater into stream channels [3,4,5,6,7]. Moreover, the development of urban areas leads to a substantial change of land use, where large areas of land are rendered impervious [8], which ultimately affects the hydrological cycle, reduces the amount of rainfall that can infiltrate into the soil, and increases the chance of flooding [4]. From a hydrologic perspective, increases in urbanization contribute to faster runoff and larger peaks, owing to the large reductions in water absorption into the ground [9]. This causes the stormwater runoff generated in urban areas to exceed the capacity of the water infrastructure network (i.e., drainage systems) and ultimately leads to flooding [10]. Therefore, urbanization in flood plain areas increases the risk of flooding due to increased peak discharge and volume as well as decreased time to peak [5]. Flood hazard also increases due to hydrological and hydroclimatological changes caused by the land use and microclimatic changes driven by urbanization [11]. In addition, the flow of stormwater runoff over the impervious surfaces leads to the suspension of pollutants into the main pipe systems [12]. Hence, rapid urban growth also leads to the deterioration of water quality [13].

Moreover, climate change is becoming one of the most serious issues that is receiving considerable interest [14,15,16]. Climate change in the United States has greatly increased the number of extreme disasters over the past years, with an associated cost of over USD 1.75 trillion [17]. For example, climate change over the past few years has led to the increase and the intensification of heavy rainfalls [18]. Heavy rainfalls are being accompanied by urban pluvial flooding that has tremendous environmental and social impacts [19]. Flooding is affecting each city in the United States and has led to greater economic losses compared to any other hazard [20]. For instance, a single flooding event can cause USD 1.8 billion of direct damages [21]. Moreover, stormwater flooding has become a global concern, reflecting the need for well-defined management strategies to address flooding hazards exacerbated by climate change [22]. Furthermore, studies have shown that the current stormwater management infrastructures are incapable of accommodating the expected peak runoff caused by climate change, which might lead to a higher frequency and intensity of floods in urban catchments [23]. Studies have also shown that one of the major causes of urban flooding is due to the fact that the current drainage systems have been designed without the considerations of both climate change and urbanization [24]. Furthermore, a review and survey of the adapted drainage systems has shown that the current practices are inadequate in addressing water quality issues and expected flooding [25]. That being said, there is a need for improving stormwater management practices to reduce flooding hazards that are exacerbated by urbanization and climate change [26].

Several stormwater management practices have been developed and implemented. Examples of these practices include green roofs, rain barrels and cisterns, permeable pavements, bioretention areas, vegetated and dry swales, vegetated filter strips, sand and organic filters, constructed wetlands, riparian buffers, catch basins (CBs), and curb and gutter elimination, among others [27,28]. Some of the stormwater management practices fall under the “sponge city” development or nature-based solutions. Although these practices have a great potential for urban water quality and quantity management, most of them are considered relatively new, and hence many local governments are reluctant to widely implement them because they are unfamiliar with their maintenance requirements and costs [29]. Thus, there are no reliable data that are widely available and that could be used to accurately and precisely study the maintenance, repair, and replacement aspects of such “sponge city” development or nature-based solutions.

Therefore, many municipalities and local governments are relying on traditional stormwater management practices such as CBs. CBs are chambers that have a sediment sump at their base to retain debris underneath the overflow point to the sewer lines [30]. These sumps are considered permanent storage areas for sediments that are not intended to flow to the downstream pipes. CBs are also defined as inlets with grating at the curb, discharge at the bottom connected to the storm drainage system, and a sump that is usually around 0.5 m to 1 m below the outlet [25]. CBs are assigned to separate sediments and contaminants from water that drains into the main sewer line [31]. Moreover, one of the main roles of CBs is to temporarily retain surface runoff before the direct disposal into the main sewer system [17].

While a CB generally has a small storage area, a well-designed network of CBs could play a vital role in flood control. In fact, CBs allow water to quickly drain from streets and thus they can work wonders for removing excess water and preventing flood damages to properties nearby [32]. This is achieved not necessarily by their relatively small storage area but rather by their ability to direct the runoff water through drainage pipes underground and ultimately move excess runoff water to municipal storm drains [33], thus reducing urban flooding. Hence, to improve stormwater management practices, CBs have been used extensively to remove coarse material or sediments from the stormwater runoff, reduce sedimentation in the drainage systems, and reduce the effect of stormwater runoff [25,34,35]. Additionally, it is well documented that one of the serious causes of flooding is blocked CB grates in streets [36]. In fact, blocked CBs could lead to street flooding since intense storms may push leaves and litter into CBs, where they could mold into mats and obstruct the basins, which prevents rainwater from entering the storm sewer, thereby causing street flooding [21]. Ultimately, maintaining the serviceability of CBs is important in improving stormwater management practices and flooding prevention.

Furthermore, the main factors that influence the effectiveness of CBs are their systematic maintenance, their design, and their sump size [37]. In relation to that, the maintenance, repairs, and replacement of CBs are crucial for the avoidance of clogging and potential flooding. While previous research efforts highlighted the significance of CBs in stormwater management practices, few studies have focused on devising optimal maintenance, repair, and replacement strategies for CBs. Thus, this research aims to simulate the life span of CBs and optimize the policies needed for minimizing the overall cost of the system while attaining the required serviceability level of the CBs.

2. Goal and Objectives

The goal of this research is to develop an optimal and adaptive maintenance, repair, and overhaul (MRO) policy for CBs, where MRO refers to the process needed to keep or restore the assets (i.e., CBs) to their functioning state. In relation to that, the accompanying objectives are: (1) model and simulate the failure and maintenance process of CBs while considering the associated uncertainties; (2) determine the best MRO parameters that result in the desired level of serviceability of CBs; and (3) understand the impact of different MRO parameters on the overall cost as well as the serviceability level of CBs.

3. Literature Review

This section reviews the existing scholarly works related to CBs, provides the needed background information on agent-based modeling, and pinpoints the knowledge gap addressed in this research.

3.1. CBs-Related Research Efforts

Multiple previous studies have shed light on the importance of CBs in stormwater management. In relation to that, CBs have been listed as one of the common practices for stormwater management that provide underground storage for rainwater [38]. CBs were also identified as one of the main applications in stormwater management that prevent sediments from reaching the main sewer lines [39]. Moreover, sumps constructed in the CBs were determined capable of minimizing the accumulation of sediments in the drainage system as well as reducing the need for maintenance [40].

Many studies have focused on the hydrological aspects of CBs. Some research efforts developed formulations for calculating the capture rate of CBs [41,42]. The methods of sediment removal by the sumps of CBs have also been the interest of several research studies that tried to establish and study the effectiveness of new techniques [30,43]. Research also has been directed towards factors that affect the efficiency of sediment capture in CBs, where it was concluded that the size of the bed load sediment in CBs is a major factor that can affect sediment deposition [44,45]. Studies have aimed to enhance the knowledge about flow fields in CBs and estimate the sedimentation rate as a function of the sump depth [17]. Experiments were also performed to enhance the knowledge about CBs, where full-scale physical models were developed to assess the scouring potential of sediments in CBs, including (1) scour experiments that measure the particle size distribution, turbidity, and total suspended solids at the outlet for different flow rates and depth of sediments; and (2) hydrodynamic tests that measure velocities for different inlet geometries and flow rates and at varying locations in the control volume of the CB [46]. Models have been developed and implemented to study several hydrological aspects of CBs. For instance, two-dimensional computational fluid dynamics models were implemented to examine the impact of flow rate, sediments’ specific gravity, and overlying water protection depth on the migration of sediments and cleaning or scouring of CBs [47]. Furthermore, simplified models were presented that can estimate the concentration of effluent-suspended sediments captured from CBs [48].

In addition to the hydrological characteristics of CBs, some research efforts were directed towards the study of the environmental aspects of CBs and the methods needed or recommended to limit the associated effects. In relation to that, a norm was developed to assess the efficiency of pesticides used in CBs to control mosquitos, where it was concluded that basins with greater depths hold more organic material and consequently attract more mosquitoes, and thus require more pesticides [49]. The regulations for the reuse of CB cleanings were investigated to help identify environmentally sound, cost-effective reuse alternatives [50]. Other studies focused on the importance of using CB inserts (CBIs), where CBIs are installed within the CB to remove gross pollutants from stormwater [51]. CBIs are responsible for the screening, sedimentation, and adsorption of pollutants as water flows through [52]. In relation to that, the water quality and properties of the solids captured by CBIs were investigated [53]. Additionally, the best installment properties of CBIs and their durability were studied [54], as were the design and construction of CBIs to maximize the rate of pollutant removal [12]. Different types of CBIs have been examined and compared in terms of their sediment removal capacity [55]. Finally, an evaluation of the efficiency of CBIs in the removal of urban stormwater pollutants was also performed in previous studies [56].

In addition to the hydrological and environmental aspects of CBs, other studies highlighted the importance of the frequent cleaning of CBs for their optimal performance and for enhancing their effectiveness [57]. It was concluded that the regular cleaning of CBs could limit the possibility of flooding caused by future storms and decrease the chemical oxygen demand level by up to 56% and the biochemical oxygen demand level by up to 88% [58]. It was also determined that if CBs are not properly and frequently maintained, then clogging and production of pollutants may occur [40], where it was determined that a frequent cleaning of two times a year increases CB efficiency in capturing particles and reduces the total residue [59]. It was also reported that in the absence of frequent cleaning, CBs could become a source of pollutants, which would deteriorate the quality of water and attract a lot of mosquitoes carrying different viruses that would cause health hazards for the community [60].

Moreover, municipalities adapt different practices for maintaining CBs. In general, municipalities are required to develop, update, and implement CB cleaning, maintenance, and inspection programs, where a CB must be cleaned and maintained as frequently as possible to ensure trash, sediment, and other debris are removed and guarantee proper functioning of the CB. Some states or municipalities set predetermined time intervals, frequencies, or rules for inspecting, cleaning, and maintaining CBs, while others also rely on ad-hoc complaints about clogged or non-functioning CBs. Common practices include [61,62]: (1) inspecting or cleaning of CBs four times per year and at the end of foliage and snow-removal seasons; (2) removing organic material, sediment, and hydrocarbons four times per year or whenever half of the depth of the CB is reached; (3) always cleaning out CBs after street sweeping; and (4) more frequent cleaning might be necessary in the case of handling runoff from land uses with higher potential pollutant loads or discharging runoff near or to a critical area. It is worth mentioning that these frequencies or periods generally vary between municipalities. Furthermore, CBs should generally be inspected by state licensed professional engineers who should also certify the conditions of the inspected CBs [63]. Other practices also rely on ad-hoc complaints. For example, New York City (NYC) developed the NYC 311 platform to collect sewer-related complaints from the public or city residents, including CB-related issues such as clogged CBs where the reported issues are referred to the appropriate agency for inspection and follow-up [64]. Moreover, the Department of Environmental Protection (DEP) plays an important role as it is responsible for cleaning, inspecting, and identifying the CBs that have a high likelihood of clogging [65]. Furthermore, the Departments of Transportation might also be responsible for ensuring that the tops of the CBs are clear on major roads, while the Departments of Sanitation could be responsible for clearing the tops of CBs on minor roads [65].

3.2. Agent-Based Modelling (ABM)

ABM is a relatively new modeling technique used for simulating a system with interacting elements called agents [66]. ABM has several strengths in terms of its flexibility and ability to model heterogenous agents with different properties as well as its ability to obtain deeper perceptions of the simulated system as compared to other traditional simulation methods that cannot accurately or precisely capture such perceptions [67,68,69]. Moreover, agent-based models have the ability to construct agents at a distinct degree of characterization and describe non-linear relations and interactions between them [70]. Furthermore, ABM is considered the most efficient simulating method when it comes to large dynamic systems that cannot be solved using a single agent [71]. ABM has also been identified as a computational engine that could result in understanding the behavior of different agents and developing the optimal strategies for the studied system [72,73]. As for its shortcoming, ABM can be computationally expensive for large models and is generally not used for real-time predictions [74,75].

ABM has been used in different fields, including stormwater management practices. For instance, ABM was used in flooding studies to predict the loss of life due to dam breaks [76]. ABM was also used to simulate the evolution of flood risks and vulnerability to establish the needed insurance mechanism after flooding events [77]. ABM was implemented in the flooding risk assessment domain by simulating the evacuation plan for pedestrians after flood events [78]. ABM was leveraged to devise optimal maintenance programs for entire networks of green infrastructure to ensure that they function properly and continue to perform their water quality and flood control functions [79]. ABM was also adapted to investigate the effect of flood protection policies, the citizen’s behavior, and the frequency of floods on communities’ flood risk [80].

ABM has also been implemented in the construction industry, including enhancing the safety management practices on construction sites [81,82], analyzing construction crew performance and related schedule considerations [83,84], improving productivity [84,85], measuring information sharing in construction [86], investigating the mechanism and impacts of goal incongruence during project proposal development in construction [87], and examining vulnerabilities associated with disaster recovery [88,89], among many other applications.

ABM has also gained great momentum in the infrastructure sector. In the transportation domain, ABM was used to study vehicular technologies and mobility-on-demand (MoD) services [90]. ABM also simulated the financial strategies applied in the transportation field to develop optimal financial policies for highway projects [91]. Moreover, ABM was adapted to simulate the usage of shared autonomous vehicles and to establish their environmental benefits [92]. ABM was also used to simulate the performance of transportation systems after earthquakes [93]. ABM was also integrated with urban infrastructure management, where it was used to simulate the influence of the social behavior of individuals by affecting their infrastructure services usages and how social aspects should be taken into consideration during the urban infrastructure management process [94]. Finally, ABM was also used to explore the reactions of relevant stakeholders during a contamination event or response in water distribution systems [95].

3.3. Knowledge Gap

The aforementioned review of the previous related literature shows that while existing studies provided knowledge on the hydrological and environmental aspects of CBs, there is a lack of research work related to devising optimal MRO strategies for the proper functionality of CBs. Although studies have discussed the importance of frequent CB cleaning, no previous study has been conducted to develop models to simulate the functionality of catch basins in order to reveal the optimal parameters needed for operating them and understanding their lifetime cycle. In addition, few studies have focused on preventative maintenance-related considerations as well as the needed replacement and repair of CBs. To that extent, although previous studies modeled the operation and performance of CBs [17,48,96,97,98,99,100,101,102,103,104], the maintenance-related aspects were not extensively modeled. Moreover, the literature review related to the ABM shows that although this modeling and simulation method was implemented in different applications and domains, none of the previous studies used it to model the functionality and maintenance-related aspects of CBs. Thus, to address the limitations of previous studies and address the existing knowledge gap, this paper establishes a framework to develop an optimal and adaptive MRO strategy for CBs. To this end, ABM was used to model, analyze, and optimize the maintenance and serviceability of CBs through minimizing the overall cost of the maintenance, repair, and replacement of CBs while also maintaining a desired serviceability level for CBs.

4. Methodology

A methodology comprised of three main steps was adapted in this study, as summarized in Figure 1. Further details on each step are discussed in the below subsections.

4.1. Step 1: Development of the Agent-Based Model

The developed ABM simulates the states of CBs, tracks the associated costs, monitors the functionality of the CBs, and infers an optimal MRO process. The model was developed using AnyLogic 8.7.12, which is an object-oriented simulation software that supports ABM [105]. The ABM is built according to the defined agents, the collected data, and the state charts of the agents. The following subsections further explain and identify the different elements of the developed model.

4.1.1. Agent Types

ABM is a simulation method that focuses on individuals, known as agents, and associates them with the needed capabilities and inputs [106]. The agents chosen and defined for the simulation of the ABM have a great impact on the model and its results. Agents in the ABM are described as the components that are responsible for making decisions in a composite system and are described by their specific rules and recurring patterns of departments [107]. Moreover, the selected agents are the system’s components that are of interest and for which their behavior must be monitored.

Since maintaining, repairing, and replacing catch basins require service crews (i.e., human resources) to perform the required work, the developed agent-based model has two main agent types: CBs and service crews; where service crews are based in a central location, which is defined as the location of the service company.

4.1.2. Study Area and Data Collection

Study Area

The studied area is in NYC, which is located in the north-east of the United States. NYC has a total area of around 469 square miles [108]. This city is known to be one of the most populated cities in the U.S., with the greatest population density [109]. NYC is composed of five boroughs: Queens, Brooklyn, Manhattan, Staten Island, and Bronx, where Manhattan has the highest population density with a population of 150,000 residents [110]. NYC’s climate is classified as humid subtropical, based on the Köppen-Geiger Climate Subdivisions, with an annual rainfall of around 1270 mm [111,112]. Moreover, the mean annual number of days with precipitation equal to or higher than 0.254 mm is 120 days, while that with precipitation equal to or higher than 25.4 mm is 13–14 days [113,114]. Furthermore, NYC was exposed to 90–102 severe storms from 1960 to 2014, with a subsequent cost ranging from USD 4 million to USD 17 million [114].

As far as land distribution is concerned, NYC has 72% of its land area covered with impervious surfaces [115]. Regarding NYC’s drainage system infrastructure, the sewer connections in NYC are of two types: combined sewer systems and separate storm sewer system [115]. NYC uses CBs to connect the storm water to the sewer system. Moreover, NYC implements different practices for the maintenance and inspection of its CBs. In relation to that, the NYC 311 platform was developed to collect sewer complaints from the public. Until July 2016, the city was evaluating the CBs based on the reported 311 complaints and every 3 years. However, after July 2016, the city started inspecting its CBs on an annual basis. Moreover, the DEP manages a program that is responsible for inspecting, hooding, and maintaining CBs as well as cleaning or repairing CBs [64]. In regard to NYC’s annual budget for CBs, the DEP allocated USD 4.7 million in the year 2022 for hiring staff for the maintenance of CBs, hydrants, and blue belt maintenance [116]. Furthermore, the budget for the inspection and cleaning of CBs was USD 4,184,000 in 2022 and USD 5,443,000 in 2023 [116]. Moreover, NYC’s budget for cleaning arterial highway CBs was USD 512,000 in 2022 and USD 282,000 in 2023 [116].

Data Collection

The developed model was applied to the entire network of CBs in NYC’s Council District 5 (shown in Figure 2) due to the availability of the needed data in this district. This district is mainly located in Manhattan’s upper east side. All 560 CBs located in District 5 were considered in the developed model. However, it is to be noted that the proposed model is flexible and can be implemented in any geographical area and with other model/system parameters, which makes it adaptive for any scenario and area presented by the user.

4.1.3. Agents’ State Charts

State charts are considered the backbone of any agent-based model [84]. These charts are diagrams used to describe and illustrate the states, the transition between a state and another, the triggers that cause a transition, and the actions of agents [117]. The details of the state charts for both agent types, CBs, and service crews, are offered in the below subsections.

4.1.4. CBs

Figure 3 presents the state chart for a CB in the developed ABM. It is worth mentioning that the state chart of the CB shown in Figure 3 represents a continuous loop of state changes, where a CB will either be “operating”, “failed”, “being replaced”, “being maintained”, or “being repaired”. In other words, the CB constantly changes its states throughout the entire simulation period, and thus the state chart shown in Figure 3 does not have an “end” node.

Figure 3 shows that during the simulation of CBs, each CB can take five states, as discussed below:

• Operating: All CBs enter the simulation process through this state in order to consider all the CBs as normally operating basins at the beginning of the simulation.
• Being maintained: The CB can be maintained in two cases:
  • If the period for maintenance, ψ (in days), is due and if the need for replacement is not due yet.
  • If the CB has been repaired and if maintaining the CB is due.

Moreover, μ (in hours) is the time required by the service crews to finish maintaining a CB, where each CB will go back to its operating state after μ.

3.

Failure: The CB reaches its failure state according to the failure rate $ø$ (unitless). The failure rate is defined by three parameters (as shown in Equation (3)): the base failing rate ε (unitless), the lifetime factor LF (represented in Equation (1)), and the factor of maintenance FM (as shown in Equation (2)). The failure of a CB is defined in this paper to be when either water begins to infiltrate from outside the concrete casting or percolates between the asphalt and steel frame of the CB, which then causes the frame and grate to sink under the traffic stress load, resulting in a shifting grate that moves vertically or a complete failure and the formation of a sink hole. Moreover, a CB is considered to be a failed CB that needs replacement if repair is no longer capable of fixing the CB.

$$LF=\max \left(1,\frac{Age of catch basin}{\Psi \ast AF}\right)$$

(1)

where LF refers to the lifetime factor, AF represents the age factor, and ψ is the period for maintenance and is expressed in days.

$$FM=\max \left(1,\frac{Time since last maintenace}{\Psi }\right)$$

(2)

where FM is the factor of maintenance and ψ is the period for maintenance (in days).

$$ø=\epsilon \ast LF\ast FM$$

(3)

where $ø$ is the rate of failure, LF is the lifetime factor, ε is the base failing rate, and FM is the factor of maintenance.

4.

Being repaired: If a CB has failed but replacement is not needed, repair is done. The CB departs its “Being Repaired” state after γ (in days), where γ is defined as the time required for the service crews to repair a CB. After the crew has finished repairing the CB, they check if maintenance is due; if not, the CB goes back to its operating state. The CB repairs considered in this study are structural and range from a simple patching of the asphalt around the perimeter with a quick mortar repair to major or complete structural repairs.

5.

Being Replaced: Under the following two cases, a CB enters the “Being Repaired” state:

• If the CB has failed and, according to the replacement probability θ (unitless), replacing the CB is required or the age of the CB is greater than the time for replacing the CB (as shown in Equation (4)).
• If maintenance is due and the age of the CB is greater than the time for replacement.

$$Time for replacement=\mathsf{\Psi }\ast MP$$

(4)

where ψ (in days) represents the maintenance period, and MP is the frequency of maintenance periods until the CB needs to be replaced (expressed in counts).

Moreover, β (in days) is defined as the time required by the service crews to replace a CB or the time after which the CB can go back to its operating state.

Furthermore, the associated costs for maintaining, repairing, and replacing the CBs are shown in Equations (5)–(7), respectively.

$$C_M=\tau \ast Number of maintence operations$$

(5)

where $C_M$ is the cost for maintaining (in U.S. Dollars) and τ is the cost of a single maintenance task (in U.S. Dollars).

$$C_R=\rho \ast Number of repair operations$$

(6)

where $C_R$ is the cost for repairing (in U.S. Dollars) and $\rho$ is the cost of a single repair task (in U.S. Dollars).

$$C_{REP}=\alpha \ast Number of replacement operations$$

(7)

where $C_{REP}$ is the cost for replacing (in U.S. Dollars) and α is the cost of a single replacement task (in U.S. Dollars).

4.1.5. Service Crews

Service crews are based in a central location, which is defined as the location of the service company. These crews navigate to the specified CB when a request for maintaining, repairing, or replacing a CB is received, to complete the assigned task. When the specified work is done, the crews drive back to the central location. The service crews are agents characterized by mobility and move with a speed V in the overall system. Figure 4 summarizes the state chart of service crews.

Figure 4 shows that service crews have four different states. The service crews could be in their idle state; in their driving state, where the service crews mobilize to a certain CB when a job is received; in the working state; or in the driving back state, where the service crews mobilize back to their central location after accomplishing the assigned task. Moreover, whenever the service crews are in the composite state, the idle or driving back state, or after having completed the needed task (i.e., after leaving the working state), the crews examine if any work requests exist; if requests are assigned, the service crews drive to the associated location and complete the job; however, if there are no requests, the service crews drive back to the central location.

The cost of service crews over the entire planning horizon of 20 years is represented in Equation (8).

$$C_C=SC\ast \lambda \ast 365 days\ast 20 years$$

(8)

where $C_C$ is the total cost of service crews (in U.S. Dollars), SC is the number of service crews, and λ is the daily service crews cost (in U.S. Dollars).

4.2. Step 2: Optimization

The main goal of this study is to optimize the MRO policy of the CBs. Optimization is the process of finding the optimal combination of variables that can yield the best possible solution to the function or equation to be optimized while abiding by the assigned restraints. Thus, after developing the model, the decision-making variables related to the MRO policy of the CBs (shown in

Table 1. Optimization variables.

Variable	Symbol
Number of service crews	SC
Replacement policy (binary) a	RP
Maintenance periods until the CB has to be replaced	MP

^a Replacement Policy: If true (i.e., RP = 1), CBs are replaced after MP; if false (i.e., RP = 0), CBs are only replaced when they cannot be repaired according to the replacement probability θ.

) were optimized by planning over a life-cycle of 20 years.

The objective function minimizes the total incurred cost over the planning horizon or the life-cycle (shown in Equation (9)), subject to a serviceability constraint that ensures a desired fraction of CBs (η) remain operating (shown in Equation (10)).

$$min(Total cost)=\min (C_M+C_R+C_{REP}+C_C)$$

(9)

$$Fraction of operating catch basins=\frac{Number of functioning catch basins after 20 Years}{Total number of catch basins in the system}\times 100\ge \eta \left(\%\right)$$

(10)

The constraint of the optimization is a restriction that is imposed on the possible solutions and is checked at the end of each optimization iteration. If the constraint condition is met, the solution is considered feasible; otherwise, the solution is rejected.

4.3. Step 3: Model Analysis and Validation

The following subsections explain and describe the methods used for analyzing and validating the developed ABM.

4.3.1. Sensitivity Analysis

Sensitivity analysis aims to examine the effect of the variation of the model input on its output [118]. Sensitivity analysis is used to determine the key variables impacting the model’s behavior. Thus, sensitivity analysis was conducted in this paper to investigate the impact of the different decision-making variables related to the MRO policy of CBs (i.e., the variables or factors presented in Table 1) on the overall system cost as well as the fraction of operating CBs. To this end, the following ranges were considered for the different variables in the conducted sensitivity analysis: SC was varied from 5 to 30, RP is a binary variable, and MP was varied between 1 and 10. It is worth mentioning that 30 service crews was not the optimal number of crews that is recommended to be assigned for the CBs (which turned out to be 7 according to the optimization results), but rather 30 service crews was considered the maximum bound for the number of service crews that the optimization model is constrained to. In other words, any other maximum bound could have been chosen as this will not impact the optimization results, but 30 was selected since the study area includes 560 CBs and thus 30 service crews means that each service crew is responsible for 18–20 CBs on average. Finally, we note that all the parameters or variables of the developed model (including the maximum bound for the number of service crews) could be changed according to the user’s preferences.

4.3.2. Monte Carlo Simulation

The Monte Carlo method is a powerful tool for modeling and simulation problems that include uncertainties [119]. Monte Carlo is based on identifying probabilities or statistical distributions for some of the model’s inputs that are uncertain or stochastic [120] and is considered one of the most precise methods for models that include inputs with probabilistic distributions [121]. Monte Carlo simulation is conducted through running the model several times with different combinations of input values that are uncertain or that follow a probability distribution. The output of Monte Carlo is a set of possible model behaviors and probabilities of different outcomes that could occur [122].

To reduce subjectivity and possible biases, a large number of replicated runs is suggested when conducting a Monte Carlo analysis [123], and several studies have recommended to run 10,000 replications [124,125]. That being said, and since the developed ABM in this research includes different uncertainties, the model was integrated with Monte Carlo analysis that was performed for 10,000 runs. More specifically, to account for the uncertainties in the cost related to the operations of CBs (i.e., maintenance, repair, or replacement) since each task might have distinct conditions, a triangular distribution is considered for the cost parameters of CBs’ operations, as shown in

Table 2. Parameters of the model.

Parameter’s Name	Parameter’s Symbol	Parameter’s Value
Time needed for the maintenance of a CB	μ	1.2 h
Time needed for the repair of a CB	γ	1 day
Time needed for the replacement of a CB	β	1 day
Daily crew cost 1	λ	USD 1000
Maintenance period 2	ψ	121 days
Replacement probability 3	θ	0.12
Base failing rate	ε	~Exp (0.009)
Fraction of functioning CBs (in percentage)	η	80%
Maintenance cost	τ	~triangular (USD 200, USD 250, USD 300)
Repair cost	$\rho$	~triangular (USD 1000, USD 3000, USD 5000)
Replacement cost	α	~triangular (USD 2300, USD 7500, USD 15,000)
Speed of service crews	V	50 kph
Age factor 4	AF	3

¹ Daily Crew Cost^: Calculated by assuming each crew is composed of 2 workers. ² Maintenance Period: Maintenance of CB is due every maintenance period. ³ Replacement Probability: The probability that the failed CB cannot be repaired and requires replacement. ⁴ Age factor: Defines the number of maintenance periods until the CB age starts to affect the failure rate.

, where the triangle distribution is often and generally used for representing the uncertainties and variations in costs [126]. Moreover, Table 2 shows other uncertainties considered in the developed ABM, which are the replacement probability and the base failure rate.

In addition to the model’s variables, which are usually the values that the decision-maker is interested in optimizing [127], each simulation model has a set of parameters that represent objects statically and are normally unchanged in a single simulation. In relation to that, Table 2 presents the parameters adapted in the simulated model.

The maintenance cost of CBs was determined based on a review of prices set by contractors for maintaining CBs in the studied area. Additionally, the daily crew cost was determined by adding the labor cost per day and the daily transportation cost needed to perform the maintenance work, where each crew was considered to consist of two laborers working for a standard 8 hours shift. That being said, the labor cost of the service crew is calculated by multiplying the average hourly wage in the studied area by the number of workers in each crew (i.e., two) and the number of work hours per day that each laborer works (i.e., eight). This cost was added to the transportation cost, which was determined from typical costs incurred in the studied area.

The output of the conducted Monte Carlo analysis in this paper provides a distribution of the overall cost, cost related to maintenance activities, cost related to repairing tasks, cost related to replacement, and the fraction of operating CBs as well as the associated variations at the end of the life-cycle.

4.3.3. Validation of the Model

As provided in previous subsections, the developed ABM incorporates uncertainties, probabilities, and randomness. Thus, an assessment using confidence intervals and significant number of runs must be performed to validate the results. Studies have proved that for the validation of complex simulations, 99% confidence intervals and ten thousand samples are needed [128]. That being said, 99% confidence intervals were derived for the overall cost, the cost related to maintenance activities, cost related to repairing tasks, cost related to replacement, and the fraction of operating CBs at the end of the life-cycle, and they were assessed to validate the model. The results showed that all models’ outputs corresponding to the determined optimal MRO policy are within their respective 99% confidence intervals), which validates the proposed model.

5. Results

The following subsections provide the obtained results from the ABM simulation and optimization, sensitivity analysis, Monte Carlo simulation, and validation of the model.

5.1. Sensitivity Analysis

The distribution (i.e., location) of all the CBs located in NYC’s District 5 and that were included in the developed ABM is shown in Figure 5.

The sensitivity analysis was performed to identify the most critical decision-making variables that affect the model’s response (i.e., behavior). In relation to that, Figure 6 shows the tornado charts that resulted from the sensitivity analysis performed in this paper.

Figure 6a shows the tornado chart of the fraction of operating CBs (in %) at the end of the life-cycle. Figure 6a reflects that the number of service crews has the widest band in this tornado chart. Thus, the number of service crews is the most influential variable on the fraction of operating CBs.

Similarly, Figure 6b shows the effect of the decision-making variables on the overall cost at the end of the planned life-cycle. Figure 6b also reflects that the number of service crews has the widest band and thus is the key influential variable on the overall cost.

Moreover, Figure 7 shows how by the end of the life-cycle the overall cost and fraction of operating CBs (in %) vary as the number of service crews changes from 5 to 30.

As shown in Figure 7, the overall cost of the system increases for every increment in the number of service crews. That being said, ensuring a minimal cost requires the selection of the least possible number of service crews. Regardless of minimizing the overall cost, the selection of the least number of service crews reduces the level of serviceability (i.e., having less than 40% of CBs operating). These results contradict the aim of designing an optimal MRO plan. Thus, the decision-maker is hereafter required to choose the minimal required serviceability level of the CBs system. Correspondingly, since the minimal fraction of operating CBs in this study, chosen to be 80%, is an optimization constraint, every solution not achieving this constraint will be assigned an infeasible solution. So, all solutions corresponding to the number of service crews less than seven are rejected (see red circle in Figure 7). That is, keeping more than 80% of the CBs operating with the least cost requires seven service crews, as shown in Figure 7. In other words, seven is the optimal number of service crews. The serviceability criterion was chosen to be 80% since previous studies have reported that 80% is a high operating rate and generally reflects healthy operations [129,130]. Nevertheless, it is worth noting that the serviceability criteria (as well as all parameters and variables of the proposed ABM model) are mutable and can be changed according to the user’s or decision-maker’s preferences.

Moreover, Figure 8 presents the change of the overall cost and fraction of operating CBs (in %) as MP changes in the range of 1–10.

Figure 8 shows that as the MP increases, the overall cost decreases. Thus, the minimal cost requires the selection of the largest MP. However, the serviceability level of 80% operating CBs is obtained for MPs less than or equal to 5 (see red circle in Figure 8). According to Figure 8, keeping at least 80% of the CB operating with the minimal cost is attained with MP = 5. That being said, the optimal value of the MP decision-making variable is 5.

Furthermore, Figure 9 represents the change of the overall cost and fraction of operating CB (in %) with the change of the RP variable.

As shown in Figure 9, the least overall cost is associated with RP = 0 (i.e., RP is False or is not implemented). However, since the minimal fraction of operating CBs in this study, chosen to be 80%, is an optimization constraint, every solution not achieving this constraint will be considered as an infeasible solution. Thus, RP = 0 is rejected and RP = 1 (i.e., RP is True or is implemented) is considered feasible since the fraction of operating CBs is greater than 80%. That being said, the optimal value of the RP decision-making variable is RP = 1.

5.2. Optimal Model Variables

The optimization results determined that when achieving the minimal overall system cost and maintaining at least 80% of the CBs in the operating state at the end of the life-cycle, the decision-maker has to apply the following MRO policy: having seven service crews, following a replacement policy, and replacing CBs after five maintenance periods. It could be concluded that the achieved optimal results complement the sensitivity results discussed in the preceding subsection. The optimal MRO policy will lead to an overall cost of USD 190,100,000 and a fraction of operating CBs of 84.8%.

5.3. Monte Carlo Simulation

Ten thousand Monte Carlo runs were performed. These runs yielded the distributions of the overall cost, cost related to maintenance activities, cost related to repairing tasks, cost related to replacement, and the fraction of operating CBs at the end of the life-cycle. These distributions are shown in Figure 10.

Figure 10 reveals that the model’s output possesses a unimodal behavior since all distributions only have one peak. Moreover, Figure 10 shows that the distributions are more-or-less symmetrical. That being said, the obtained distributions could be categorized as bell-shaped distributions.

Furthermore, Figure 11 presents the graph of the variations in the yearly overall cost over the planning horizon, the yearly cost related to maintenance activities over the planning horizon, the yearly cost related to repair tasks over the planning horizon, and the yearly cost related to replacement over the planning horizon, with their associated error bars.

Figure 11 shows that after the sixth year, the model’s output stabilizes such that the average yearly overall cost is around USD 9,500,000, the average yearly cost related to maintenance activities is around USD 150,000, the average yearly cost related to repair tasks is USD 3,600,000, the average yearly cost related to replacement is USD 3,200,000, and the remaining USD 2,550,000 is the average yearly cost of the service crews.

Furthermore, the large variability seen in years 1 and 2 in Figure 11 is due to the nature of simulation models that require some iterations in order to stabilize. More specifically, the variability in the obtained results is due to the fact that all CBs were assumed to be operating in the beginning of the simulation, and the CBs were constantly changing their states (between operating, failed, being maintained, being repaired, and being replaced) throughout the simulation period; hence, the simulation model needs some iterations to stabilize. Thus, the first few iterations/years were associated with bigger variability as compared to the other iterations/years (as shown in Figure 11). Although some models could generally undertake hot starts to avoid or reduce model instabilities by providing initial values (or guesses) for some of the model parameters or variables, this was not performed in the developed ABM so as to reduce potential biases in choosing these initial values. In relation to that, while this might generally require additional iterations to stabilize the model, the proposed ABM did not need a considerable number of iterations as it was able to converge relatively fast. Furthermore, while the ABM model encountered certain instabilities in the initial iterations, these variations do not affect the optimal maintenance recommendations provided in this paper since the results were reported based on the stabilized behavior of the model.

Moreover,

Table 3. The associated errors (in U.S. Dollars) of the model’s outputs.

Year	Overall Cost	Maintenance Cost	Repair Cost	Replace Cost
1	96,910.5	7528.0	60,316.5	116,860.6
2	105,196.9	3834.5	59,057.9	127,310.7
3	103,877.1	3918.6	60,999.8	127,502.9
4	106,626.1	3971.8	59,727.3	129,165.9
5	102,477.8	3949.7	60,053.1	126,555.9
6	103,500.7	3931.0	61,116.0	128,350.6
7	103,888.3	3873.4	60,178.7	127,243.7
8	104,698.7	3931.3	61,029.3	128,601.6
9	104,960.3	3947.2	60,229.5	128,339.0
10	104,722.7	3950.4	61,361.8	128,433.9
11	105,230.8	3911.9	60,790.4	129,091.4
12	104,264.4	3923.2	60,263.6	129,545.5
13	104,275.7	3928.1	60,663.5	127,937.2
14	104,754.7	3931.2	60,376.0	128,548.2
15	105,478.4	3936.3	59,748.7	128,545.5
16	104,084.0	3929.4	59,973.1	127,827.1
17	104,884.1	3956.0	60,747.2	128,093.4
18	105,563.5	3897.5	59,907.0	129,071.9
19	103,809.4	3935.4	61,358.2	129,150.5
20	104,873.9	3920.2	60,237.1	128,617.2

presents the errors associated with each year with respect to the yearly overall cost over the planning horizon, the yearly cost related to maintenance activities over the planning horizon, the yearly cost related to repair tasks over the planning horizon, and the yearly cost related to replacement over the planning horizon. It is worth mentioning that these errors are also visualized in Figure 11.

The average error associated with the total annual cost is 1.09%, with the annual maintenance cost is 2.72%, with the annual repair is 1.67%, and with the annual replace is 4.04%. These are considered short error bars [131]. These short error bars indicate a low spread of the data, and that the data are clustered around the mean (i.e., low variations in the model’s output and thus more confidence in the obtained results). Furthermore, the short bars indicate the high reliability and accuracy of representing the data by their mean. Thus, the line plots that represent the average annual costs of the 10,000 Monte Carlo simulations are a good representation of the overall data due to the high reliability of the mean.

Finally, although Figure 11 shows a similar trend of error range at most of the positions (i.e., at different years), Table 3 presents the actual values of each error bar and shows that the error values are not the same, but rather, they are close to each other.

5.4. Model Validation

Although the simulated model has been fully analyzed, its validation is still required in order to deem it as an effective, precise, and systematic model. That being said, and as explained in the “Methodology”, the 99% confidence interval validation approach, which has been recommended for such models, was chosen for validating the simulated model [128]. In relation to that,

Table 4. The behavior of the model and the corresponding 99% confidence intervals.

Parameter	Results of the Agent-Based Model	99% Confidence Interval
Overall cost after 20 years	USD 190,100,000	[USD 188,706,728; USD 190,669,288]
Maintenance cost after 20 years	USD 3,017,199	[USD 2,975,766; USD 3,064,700]
Repair cost after 20 years	USD 72,388,368	[USD 71,616,093; USD 72,781,261]
Replace cost after 20 years	USD 63,594,929	[USD 62,371,475; USD 64,414,078]
Fraction of operating CB after 20 years	84.8%	[73.4%; 88%]

presents the results obtained from the ABM model using the optimal MRO plan and the associated 99% confidence intervals.

As shown in Table 4, all models’ outputs corresponding to the determined optimal MRO policy are within their respective 99% confidence intervals. Eventually, these results validate the proposed MRO policy obtained from the simulated ABM.

It is worth noting that the costs presented in Table 4 refer to the costs incurred throughout the entire 20-year planning horizon. For example, the overall cost of USD 190,100,000 after 20 years corresponds to an average of USD 9,500,000 per year. This cost includes the maintenance, repair, and replacement costs of CBs as well as the crew cost (i.e., labor and transportation) that is needed to ensure a minimum operating rate of 80% (actually, this cost corresponds to an operating rate of the CBs of 84.5%, as shown in Table 4). More specifically, the USD 9,500,000 is distributed as such: USD 150,000 per year for maintenance, USD 3,600,000 per year for repair, and USD 3,200,000 per year for replacement, and the remaining cost is for service crew. To better position this cost to current practices, it was compared to the actual budgets specified by NYC as well as other urban areas in the U.S. In regard to NYC’s annual budget for CBs, the DEP allocated USD 4.7 million in year 2022 for hiring staff for the maintenance of CBs, hydrants, and blue belt maintenance [116]. Furthermore, the budget for the inspection and cleaning of CBs was USD 4,184,000 in 2022 and USD 5,443,000 in 2023 [116]. Moreover, NYC’s budget for cleaning arterial highway CBs was USD 512,000 in 2022 and USD 282,000 in 2023 [116]. These reported budgets are mainly for inspecting and cleaning CBs and thus do not necessarily include the repair and replacement costs. Moreover, comparing the obtained costs from the developed model to the budgets of New Orleans (located in Louisiana, U.S.), the cleaning and assessment contract of CBs was USD 7 million for a period of 120 days, while the contract for minor and major repairs of CBs for a period of 12 months was USD 13 million [132]. That being said, the costs associated with the developed model are not only comparable to current budgets both within the study area and in other urban areas of the U.S., but they are also more efficient and lower since the proposed ABM model helps to reduce costs through optimizing maintenance schedules and programs.

Finally, the developed ABM in this paper and the associated recommendations are believed to enhance the current practices used in NYC for managing its network of CBs. The current practices in NYC rely on the NYC 311 platform to collect sewer-related complaints from the public or city residents, including CB-related issues such as clogged CBs, where the reported issues are referred to the appropriate agency for inspection and follow-up [64]. While this platform might be efficient to a certain extent in maintaining and repairing CBs, it follows a “corrective” maintenance approach, where CBs are mainly repaired or restored following failure, as opposed to a “preventive” maintenance approach, which is recommended by the proposed ABM in this paper. A preventive maintenance approach overcomes many of the disadvantages of corrective maintenance by reducing the probability of occurrence of failure, avoiding sudden or unexpected failures, reducing cumulative costs, increasing system operating periods, decreasing unpredictability and uncertainty, and reducing safety concerns or damages [132,133,134,135]. The proposed approach is also superior to current practices because it embraces a data-driven, proactive approach for operating and maintaining CBs to improve efficiency, which has always been the goal of NYC [136]. The proposed approach also ensures that the target operating rate is met throughout the entire planning period and that the entire network of CBs is always above the serviceability criteria that could be specified by the user of the proposed ABM. Additionally, NYC is constantly re-evaluating its regulations related to the inspection and maintenance of CBs. In fact, NYC was traditionally inspecting CBs every three years, and then it changed the CB inspection requirements to be on an annual basis after 2016 due to increased flooding risks, and NYC is set to re-evaluate its current CB inspection program in the next few years in order to further optimize benefits [64]. Hence, the developed ABM could be a potential option for NYC to consider including as part of its CB inspection and maintenance program. Additionally, NYC is prone to great flood risks as the city is highly vulnerable to flooding caused by hurricanes, tropical storms, nor’easters, intense rainstorms, and even extreme high tides [137]. In fact, there are around 400,000 NYC residents that are affected by the 100-year annual chance floodplains, which is higher than the entire population of Cleveland, OH, Tampa, FL, or St. Louis, MO [137]. The flood risk in NYC is even expected to increase in the future due to climate change and increasing sea levels, which can lead to frequent, potentially daily, tidal inundation in some especially low-lying neighborhoods [137]. While the proposed ABM is not expected to solve these increasing flood risks, it could still contribute to addressing some of the flood risks, especially those caused by intense rainstorms. Since it is hard to accurately estimate the degree to which the proposed ABM model could reduce flooding risks in NYC, it is reported that regular inspection and cleaning of CBs in NYC could reduce CB-related complaints or failures by 26 percent and could increase proactive sewer inspections and cleaning by 133 percent [136]. Hence, it could be hypothesized that improvements of such magnitudes could be expected; however, the proposed ABM in this paper would benefit from a better model validation approach by considering real data from municipalities. Hence, the authors recommend future studies to further validate the proposed ABM by using it in real-world applications or case studies and report the associated reductions in CBs failures and flooding risks.

6. Discussion and Contributions

To reflect the significance of this research, this section discusses the obtained results and the paper’s contributions.

Since the improper or non-frequent maintenance of CBs could lead to their clogging and flooding around them, the proposed agent-based model in this paper can be used to devise optimal stormwater management practices that can help prevent blockage of the CBs, limit the pollutants in the storm sewers, and reduce the risk of flooding. Along with the risk of flooding, improperly maintained CBs could cause dangerous sinkholes around them, which lead to severe damages and injuries. Thus, the outcomes of this paper provide great benefits in designing proper maintenance policies and programs for CBs, which are essential for sustaining, restoring, and ensuring the ongoing circulation of CBs.

Minimizing the overall cost of the system is essential for best management practices since authorities mainly allocate limited funding for the maintenance of CBs. That being said, this paper helps in decreasing the operational costs of CBs, which can lead to a more effective use of the available funds and increase the number of operating/functional CBs. Furthermore, setting the fraction of operating CBs, η, as a constraint for the optimization procedure provides unique benefits and insights since it allows decision-makers to define and achieve a desired level of serviceability or efficiency for the entire network of CBs.

The proposed ABM presents an adaptive and optimal framework for maintaining CBs and providing a better understanding of the most critical variables that affect the overall MRO process. The conducted sensitivity analysis identified that the number of service crews is the most critical variable affecting the overall cost and the fraction of operating CBs. This highlights the role and importance of the workforce in maintaining the assets and in optimizing the MRO process. Having the needed service crews or resources is essential for providing the necessary maintenance, repair, or replacement processes that meet budgets as well as desired serviceability levels.

The proposed work in this paper contributes to maintaining the serviceability of CBs because it decreases the number of failed CBs and minimizes the disruptions associated with their failure. Having a well-defined and optimized MRO strategy will decrease the replacement frequency of CBs and thus will decrease the associated costs and increase the number of CBs in service. Ultimately, the proposed periodic preventative maintenance and replacement of CBs before their complete failure can lower the risks of flooding, enhance safety, and increase the efficiency of CBs.

The research conducted in this paper provides a systematic and adaptive approach that allows decision-makers to plan the MRO process of CBs in a way that minimizes the cost and maintains the desired level of serviceability. It is worth mentioning that the proposed research is adaptive and flexible. In other words, all parameters identified in this model and the locations and number of CBs could be changed according to the user’s desire. Hence, the suggested model could be used to optimize the MRO process of any network of CBs that is of interest.

Moreover, maintenance is one of the most critical factors in quality assurance and in ensuring the long-term functional state of any asset of interest. The maintenance of infrastructures is needed for building a safer, more reliable, more resilient, and sustainable society. This paper specifically focused on maintaining CBs. The paper’s findings were directed to address the poor maintenance practices of CBs that generally lead to several breakdowns and high additional and unnecessary costs. By following the preventative MRO policy recommended in this paper, cities and agencies can increase the lifespan of their CBs and keep them in a good condition, which will allow CBs to remain operational for a longer period and lower the cost of the needed maintenance after a breakdown.

Finally, the effective operation and maintenance of stormwater management practices protects the environment by reducing flooding and protecting people and properties, promoting healthier rivers, decreasing the demand on stormwater systems, and consequently creating more sustainable communities. That being said, by developing an effective strategy and plan for maintaining stormwater management practices, this paper provides social, economic, and environmental benefits that ultimately promote sustainability.

7. Conclusions and Future Work

This paper modeled and simulated the failure and maintenance process of CBs, established the best MRO decision-making variables that result in the optimal performance of CBs, and studied the impact of different MRO-related variables on the overall cost and fraction of operating CBs. This research was applied to NYC’s District 5 network of 560 CBs.

The results of the sensitivity analysis showed that the service crews represent the most critical factor that affects the overall cost and the fraction of operating CBs. Furthermore, the conducted Monte Carlo analysis estimated the average of the overall yearly system cost over the planning horizon, the yearly cost related to maintenance activities over the life-cycle, the yearly cost related to repairing tasks over the life-cycle, and the yearly cost related to replacement over the planning life-cycle. The findings showed that in order to minimize the total cost of the system and keep at least 80% of the CBs in their operating state, seven service crews, a replacement policy, and replacing CBs after five maintenance periods are needed. Finally, the 99% confidence intervals of the model’s response/behavior were obtained to validate the proposed ABM.

The proposed model can be adapted by decision-makers and engineers to implement an optimal MRO process that minimizes the associated costs while attaining the required serviceability of CBs. This model is flexible and can be adapted to any geographical areas and for other system parameters, and it allows users to change the parameters of the model to investigate different scenarios affecting the system’s performance and cost. Ultimately, this paper contributes to the existing literature by offering an enhanced understanding of the management process of CBs and devising optimal MRO strategies for properly operating them. Finally, this research helps decision-makers and engineers to increase the lifespan of CBs, limit their risks of breakdown and the associated threats of flooding, increase their efficiency, and avoid unnecessary costs.

Finally, since the developed model does not consider the physical characteristics of the system and the associated impacts on performance (e.g., storm characteristics, sediment size and quality, land use, and street surface conditions) as well as the factors that could affect the system dynamics and the associated water quality, it is highly recommended that future work should address this limitation by developing a hybrid model that combines the developed ABM in this paper with a conceptual, physically based model that considers the factors that affect the system’s dynamics. In relation to that, this hybrid modeling approach would provide insights into how physical characteristics such as sediment size, sediment quality, land use and street surface conditions, and storm characteristics affect the performance and maintenance of CBs as well as the water quality. Furthermore, the developed model should aim to address stormwater management requirements, flooding, and the sustainability of urban areas.