AbstractDrinking water and natural gas are essential goods provided by municipal distribution networks. In order for there to be a reliable provision of such commodities, infrastructure managers must regularly dispatch vehicles to perform tasks such as valve and hydrant inspections and meter replacements. As public concerns about sustainability grow, managers must look for ways to reduce carbon emissions. This, however, should be done without jeopardizing service. One conceivable approach is to train staff to perform additional maintenance tasks so that combined work packages can be carried out. This would require resources to be spent in different areas, such as supplemental worker training and organizational restructuring of the utility. Infrastructure managers must thus weigh the expected efficiency gains against the associated costs. The infrastructure asset management process combined with digitalization are powerful tools to assess such questions. Unfortunately, digitalization in the infrastructure sector lags behind other sectors of the economy. To bridge this gap, real-world examples are needed to further spur adoption. This paper addresses this need with a methodology and case study for infrastructure managers of water and gas networks. Specifically, this paper presents a methodology to quantify the resource requirements of operational maintenance programs for a large municipality. As utilities plan maintenance routes differently, four algorithms are used to model the resource requirements of the status quo. The effect of prioritization as well as frequency of inspections/replacements is considered.IntroductionDrinking water and natural gas are vital commodities provided by municipal distribution networks. In order for a reliable and continued provision of such goods, infrastructure managers must regularly plan and execute inspections and interventions (e.g., replacements) on network objects. These activities belong to the larger process of infrastructure asset management (IAM) described in detail for road networks in Adey (2019) and are shown in Fig. 1.The described process is equally applicable to other infrastructure networks including water and gas distribution networks and aims to ensure that infrastructure continues to provide an adequate level of service. Briefly summarized, infrastructure managers must first decide on appropriate goals for the service level provided by the infrastructure and choose indicators for quantifying service (Alegre et al. 2016). Secondly, management strategies (Engelhardt et al. 2000; Shahata and Zayed 2013) must be defined for the different infrastructure assets. This involves considering the deterioration of the asset inventory (Makar et al. 2001; Rajani and Kleiner 2001), predicting future failures (Rajani and Makar 2000; Kleiner and Rajani 2001; Harvey et al. 2014; Scheidegger et al. 2015; Scholten et al. 2013; Yuan 2016; Kerwin et al. 2018; Snider and McBean 2019; Kerwin et al. 2020), modeling the failure consequences (Cromwell 2002), and estimating the costs of executing inspections (Liu et al. 2012) and interventions (Zhao and Rajani 2002). Thirdly, infrastructure managers must determine the exact inspections and interventions to be executed in the upcoming planning period, taking into consideration the relevant constraints (Kerwin and Adey 2017, 2020, 2021). Steps 4 and 5 consist of performing the inspections and executing the interventions and subsequently analyzing their effectiveness. The final step is an analysis of the entire process. The work in this paper addresses specifically Steps 4 and 5 of the IAM process for isolation valves, hydrants, water meters, and gas meters.One of the most impactful developments shaping the IAM process has been digitalization. Digitalization refers to the trend of exploiting digitized data and digital technologies to improve process outputs. Concretely, digitalization allows infrastructure managers to manage large quantities of data from disparate sources and employ efficient algorithms to efficiently extract and communicate useful information. When these capabilities are combined, unexpected synergies appear. This can be observed in the large number of recent research publications focusing on the problem of coordinating and developing common intervention programs for multiple spatially proximate municipal networks (Carey and Lueke 2013; Kielhauser and Adey 2019, 2018; Shahata and Zayed 2015; van Riel et al. 2017; Tscheikner-Gratl et al. 2016; Tscheikner-Gratl 2016). Although digitalization has already significantly influenced IAM over the last several decades, it is lagging compared to other sectors (Sarni et al. 2019; Bose and Kielhauser 2019). This lag in adoption and implementation can be attributed to the long service lives of network components as well as the institutional and professional structures within the water and gas sector, which evolve at a slower pace than technological innovation (Lienert et al. 2006). There is a clear need for real-world examples to highlight the potential of digitalization to improve IAM. This paper makes a contribution toward this end in the following ways: 1.A methodology is provided that can be used to improve Steps 4 and 5 of the IAM process for operational maintenance programs for water and gas networks.2.The methodology is implemented in a real-world example. The example explores the potential gains in efficiency that could be obtained by integrating maintenance tasks such as hydrant and valve inspections, as well as meter replacements into common work packages. Currently, there is no comparable research in the scientific literature.3.The pivotal role of digitalization in improving the IAM process is highlighted, thereby encouraging infrastructure managers to fully embrace this trend.MethodologyThe steps followed in this paper are shown graphically in Fig. 2 using Business Process Model and Notation (Allweyer 2010).Step 1. Identify processThe methodology begins with an adequate description of the specific process to improve. In order for a reliable provision of water and natural gas, infrastructure managers must regularly dispatch vehicles to perform tasks such as valve and hydrant inspections and replace meters. This is a considerable task for large networks, which consist of tens of thousands of spatially distributed objects such as isolation valves, hydrants, and meters. Such objects are crucially important for proper network operation. Valves allow pipe segments to be isolated following an unexpected failure or during repair work and thus minimize service impacts throughout the network (Walski 1993b, a). Fire hydrants are access points to water mains and allow large quantities of water to be rapidly obtained for firefighting purposes. Hydrants are also used to flush water mains and improve water quality at end points of the network with insufficient flow. Meters are devices that measure the quantity of fluid (potable water, natural gas) flowing through the device and are installed at strategic points throughout the distribution network for operation monitoring and at points of consumption to bill clients. To ensure reliable operation of these objects, inspections and interventions are planned at regular intervals based on manufacturer recommendations and the utility’s management strategies. These include valve and hydrant exercising, hydrant flow tests, meter flow measurement tests, and meter replacements (Infraguide 2003; AWWA 2012, 2006b, a; SVGW 2012).These maintenance activities require crews of trained staff to be regularly dispatched throughout the network. The Swiss Association of Gas and Water Utilities (SVGW) recommends that valves be inspected and exercised every 2–5 years and hydrants be inspected and exercised at a minimum every 4 years (SVGW 2012). Water meters for small consumers are replaced on average every 15 years and gas meters every 14 years. Depending on the maintenance strategy followed, the associated resource requirements will differ. For example, increased inspection frequency of valves and hydrants will result in greater reliability and a higher level of provided service but will require more personnel.Step 2. Identify potential improvementReferring back to Steps 4 and 5 of the IAM process, infrastructure managers are interested in estimating the resource requirements of maintenance programs and identifying ways in which they could be lowered. A further driver is the push toward lower carbon emissions. As public concerns about sustainability grow, managers must look for ways to reduce carbon emissions without jeopardizing service. Digitalization can be leveraged for these tasks. Specifically, digitalization facilitates the use of efficient algorithms and dissolves data silos within and between utilities. As these data silos disappear, it becomes apparent that many of these maintenance tasks could theoretically be carried out by the same crew with sufficient training. Conceivably, integration of these tasks into combined work packages would lead to lower carbon emissions because workers would spend more of their day performing maintenance tasks and less time driving. This would have the additional benefit of boosting worker productivity. Currently, there is no comparable research on the potential gains of integrating maintenance tasks such as hydrant inspections, meter replacements, and other potential activities (e.g., water main flushing).Step 3. Model problemIn order to gain useful insights into the problem, a structured approach is needed to reduce the real-world complexity with apt simplifications. The procedure taken to achieve this aim is to list the underlying problem elements, identify key parameters that may influence results, and gather relevant data sources and algorithms needed for the analysis.With regard to evaluating the potential of integrating operational maintenance programs, the underlying problem elements are as follows: 1.Which metric should be used to evaluate performance?2.Which objects require maintenance?3.How to determine the maintenance route?Which Metric Should be Used to Evaluate Performance?Several metrics could be used to estimate carbon emissions and worker productivity under the two examined scenarios. Time-based or carbon-based metrics require numerous assumptions on traffic conditions, data on the utility’s fleet of vehicles, and the time required to perform tasks. Instead, route distance was chosen because it can be easily measured using a GIS database of the road network.Which Objects Require Maintenance?This question requires first determining the management strategies. These represent how objects should ideally be maintained without considering constraints. Strategies require a triggering condition. This can be the object’s age, condition, risk, a specified time interval, or another parameter used by the utility. For meters, valves and hydrants, age or time interval strategies are used. Once the strategies have been determined, the corresponding objects requiring inspections or replacements can be identified.How to Determine the Maintenance Route?Maintenance routing requires knowledge of how the network is divided into work sectors and the routing heuristics used by the utility for planning purposes. For task assignment, large networks are often divided geographically into work sectors. These sectors set the boundaries of the maintenance routes.The planning of routes varies by utility. Some may use sophisticated routing technologies and others may employ simple heuristics. This variability in route planning may affect results and needs to be accounted for in the analysis. It is thus useful to consider the heuristics that exist.The problem of determining the optimal set of routes that a fleet of vehicles should cover in order to make all necessary stops is referred to as the vehicle dispatch problem and was first described by Dantzig and Ramser (1959) for fuel deliveries. Since then, the problem has been extensively investigated and applied to a number of problems in infrastructure management such as utility meter readings (Dong et al. 2007; Shuttleworth et al. 2008) and winter maintenance of roads (Eglese et al. 2015).The problem has often been compared to the traveling salesman problem (TSP) in which a salesman must visit numerous cities and seeks to determine the most efficient route. There are many variants of the problem depending on the specific application and constraints that must be considered and is one of the most widely studied problems in operations research (Helsgaun 2000; Hahsler and Hornik 2007; Mennell 2009). One particularly relevant variant is the hierarchical TSP (HTSP), which determines the route sequentially by the priority level of the different stops (Panchamgam et al. 2013). When planning maintenance routes, an infrastructure manager may assign objects with different priority levels for a number of reasons (Marlow et al. 2012). Examples include prioritizing the inspections of valves and hydrants in areas where these objects are few and far between (Walski 1993b; Liu et al. 2017) or are based on their proximity to critical infrastructure (e.g., hospitals). As the number of objects to inspect increases, a combinatorial explosion of possible routes occurs, which makes large instances of the problem difficult to solve. Thus, industrial applications generally use heuristics to generate satisfactory, near-optimal solutions (Rosenkrantz et al. 1977).In this example, the problem is modeled as an HTSP and four different routing algorithms are defined. The algorithms vary in the routing heuristic and the priority levels of the objects as shown in Table 1. No recommendations are being made that utilities should begin using these algorithms in route planning. They are simply used to represent the variability in route planning that exists among utilities.Table 1. Algorithms used in studyTable 1. Algorithms used in studyAlgorithmRouting heuristicPriority levels1GreedyAll objects are level 12InsertionAll objects are level 13GreedyApproximately 50% level 1, 50% level 24InsertionApproximately 50% level 1, 50% level 2In Algorithms 3 and 4, priority levels are randomly assigned using the binomial distribution such that there is an approximately even distribution. For all routes, Level 1 objects are visited first. Next, the nearest Level 2 object is identified, and the route is recalculated with the remaining Level 2 objects. In reality, specific functional and spatial factors would be defined for prioritization of maintenance tasks. This is outside of the scope of this study. These specific factors differ by utility and have been discussed in other publications (Marlow et al. 2012; Liu et al. 2017).The routing heuristics used are the greedy or nearest neighbor heuristic and the insertion heuristic. The greedy heuristic is simple and analogous to an intuitive approach that a worker might use to inspect or replace objects in the absence of sophisticated routing software. It starts by making a list of objects that must be visited. The order is then determined by first selecting an object on the list at random and then determining its nearest neighbor, which is also on the list. That neighbor’s nearest neighbor is then determined and this process is repeated until all objects on the list have been visited (Hahsler and Hornik 2007; Rosenkrantz et al. 1977).The insertion heuristic has better performance and aims to add objects to an existing route consisting of at least two objects such that the cost of insertion to the total route length is minimized (Hahsler and Hornik 2007; Rosenkrantz et al. 1977). There are different insertion methods such as arbitrary insertion (insert an object at random that is not yet on the route); nearest insertion (insert an object that is nearest to one of the objects already on the route); cheapest insertion (insert an object that have the smallest increase in route length) and farthest insertion (insert an object that is farthest away from any of the objects already on the route).As a visual aid to the reader in understanding the algorithms, Algorithms 1, 2, and 4 are compared by determining the hydrant inspection route in a sector of the case study. Figs. 3–5 compare the two routing heuristics and Figs. 6–8 provide insight into the effect of prioritization. The route determined with Algorithm 3 is not shown for the sake of brevity. The blue segment is the starting point for all routes.The key factors that will influence results are the density of objects to inspect or replace in a given sector and the algorithm used to determine routes. In sectors with a high population density, the density of objects (e.g., meters, hydrants, and valves) will be significantly higher than in rural, low population density areas. Workers will have to travel further on average between objects in rural areas than in urban areas and, thus, the efficiency gains of combining meter replacements and, inspections of valves and hydrants will intuitively be higher in rural areas compared to urban environments.The analysis requires a GIS database of network objects and the road network. The study area is geographically divided into different sectors with different properties (e.g., road network density, population density). This allows the influence of maintenance density on potential efficiency gains to be explored. Furthermore, sectorization reduces the computational complexity of the route calculation. Subsequently, objects needing inspection or replacement are then linked to the road network. This is done in GIS software version 3.16.3 by snapping a set of points (the objects needing inspection or replacement) to the nearest line (the nearest road segment) and counting the number of objects on each road segment. Next, the graph structure of the road network is built using an adjacency matrix. Because a graph is a structure composed of a set of vertices and edges that can be used to describe the connectivity of a network, the adjacency matrix is used to describe how network segments are connected to each other.Step 4. Model Status Quo and Proposed ChangeTo represent the status quo, the routes are determined for each object type separately and repeated for all sectors and each considered algorithm. In the proposed change, all objects requiring maintenance in a sector are combined to determine the corresponding integrated routes. When multiple priority levels are set, Level 1 objects are visited first, followed by Level 2 objects.Case StudyModified data from the distribution networks of the Canton of Geneva are used in this paper. The locations of gas and water meters were estimated using data on service lines. The exact numbers of gas meters, water meters, valves, and hydrants have been altered and do not represent the true quantities found in the Geneva networks.For this study, the Geneva network is broken into seven sectors as shown in Fig. 9. The darkness of shading in Fig. 9 represents the relative population density of the sector. Table 2 summarizes the sector characteristics.Table 2. Characteristics of sectorsTable 2. Characteristics of sectorsSectorArea (km2)Population density (people/km2)Road length (km)Density of road network (km/km2)Number of hydrantsNumber of isolation valvesNumber of water metersNumber of gas metersA32.21,149.3234.57.36691,5423,6591,646B49.01,366.0348.57.18701,6513,9471,698C35.7498.3231.16.55791,4203,4291,477D28.51,671.2206.87.39502,5195,2102,801E37.32,937.5349.79.41,5063,3307,9474,252F40.9461.6282.26.94151,0102,574286G14.014,457.9195.714.02,0751,6157,7565,937Overall237.62,111.71,848.57.87,06413,08734,52218,097Two sets of management strategies were considered to investigate the effect of frequency of inspection or replacement on efficiency gains of integrating maintenance tasks. Maintenance program 1 (Table 3) was determined using the following management strategies: inspect valves and hydrants every four years and replace water and gas meters older than 17 years. Maintenance program 2 (Table 4) was determined using the following management strategies: inspect valves and hydrants every year and replace water and gas meters older than 12 years. These maintenance programs are then executed every year; however, the exact number of objects to replace and inspect will fluctuate somewhat. In practice, if a hydrant or isolation valve is found to be deficient during an inspection, workers will either repair it immediately if possible or make a notification that the object needs to be replaced at the next possible opportunity. In this study, only the initial inspection of isolation valves or hydrants is considered. Throughout this work, the term maintenance density refers to the numbers of objects requiring an inspection or replacement per 10 km of road in the sector.Table 3. Inspections and interventions to execute next year for Maintenance program 1Table 3. Inspections and interventions to execute next year for Maintenance program 1SectorNumber to inspectMaintenance density (objects/10 km road)Number to inspectMaintenance density (objects/10 km road)Number to replaceMaintenance density (objects/10 km road)Number to replaceMaintenance density (objects/10 km road)NumberMaintenance density (objects/10 km road)A1647.039316.82299.8883.887437.3B2176.239211.22086.01093.192626.6C1396.036315.72089.0873.879734.5D23511.461229.630414.71477.11,29862.8E37110.685824.548914.02226.31,94055.5F1033.62549.01475.2150.551918.4G46924.036718.844122.536118.41,63883.7Overall1,6989.23,23917.52,02611.01,0295.67,99243.2Table 4. Inspections and interventions to execute next year for Maintenance program 2Table 4. Inspections and interventions to execute next year for Maintenance program 2SectorNumber to inspectMaintenance density (objects/10 km road)Number to inspectMaintenance density (objects/10 km road)Number to replaceMaintenance density (objects/10 km road)Number to replaceMaintenance density (objects/10 km road)NumberMaintenance density (objects/10 km road)A66428.31,54065.71,29555.254823.44,047172.6B85824.61,62846.71,41640.658816.94,490128.8C57324.81,40560.81,21852.750822.03,704160.3D93445.22,473119.61,81287.695946.46,178298.7E1,58645.33,36696.22,84681.41,55744.59,355267.5F40714.498634.991432.4893.22,39684.9G1,969100.61,47875.52,602133.02,034104.08,083413.1Overall6,99137.812,87669.712,10365.56,28334.038,253206.9GIS data is processed in R using the packages “rgdal” version 1.5-23 (Bivand et al. 2019a) and “maptools” version 1.1-1 (Bivand et al. 2019b). The graph of the road network is built using the packages “igraph” version (Csardi 2015) and “simple features” version 1.0-1 (Pebesma et al. 2019).The greedy heuristic is written in R version 3.5.0 and the insertion heuristic is implemented using the “TSP” version 1.1-10 package (Hahsler and Hornik 2007, 2020). All four insertion methods were used in conjunction with the 2-opt edge improvement heuristic for all considered routes, and the shortest route was chosen. The road network was modeled as an undirected graph; thus, driving restrictions such as one-way roads were ignored.ResultsReferring back to the research question, the goal of this example is to quantify the potential benefits of combining maintenance tasks into integrated work packages and explore the influence of key parameters such as routing algorithm and maintenance density.This is achieved by comparing the distances of maintenance routes needed when tasks are combined versus separated. Utilities plan and carry out inspections differently. Some use sophisticated routing technology along with prioritization, and others do not. To represent this variability, the problem was modeled as a HTSP and four algorithms were defined. To understand the influence of maintenance density on results, the network is split into different sectors and two maintenance programs are defined with differing maintenance requirements. These are denoted as Maintenance programs 1 and 2.Table 5 contains the distances of the routes needed for all object types in all sectors for Maintenance program 1 and Table 6 contains the route distances calculated for Maintenance program 2. Figs. 10 and 11 visualize the calculated route distances versus the maintenance density (Tables 3 and 4) of the routes in both maintenance programs.Table 5. Distances (km) for routes calculated for Maintenance program 1Table 5. Distances (km) for routes calculated for Maintenance program 1SectorHydrantValveWater meterGas meterIntegratedHydrantValveWater meterGas meterIntegratedHydrantValveWater meterGas meterIntegratedHydrantValveWater meterGas meterIntegratedA961239447173829881411401361931407226912616011565222B110121105661919410384561531612031429129914117212989245C7694814412465866937107108156122592239614411058185D111133107691819010795581511402001629030412817914088256E1481631509124212513612780206213250221140406176220197123337F607769612259695889693122112816990110977151G9777967816185688266137142117134114252123104120101214Total6997887024001,1956016665963469899931,2411,0315741,9218801,0909085321,610Table 6. Distances (km) for routes calculated for Maintenance program 2Table 6. Distances (km) for routes calculated for Maintenance program 2SectorHydrantValveWater meterGas meterIntegratedHydrantValveWater meterGas meterIntegratedHydrantValveWater meterGas meterIntegratedHydrantValveWater meterGas meterIntegratedA1661431729020913912613675171243261278140374218232235122315B20216121210925416513816794206317287338175455265236272157379C1341211387816111110511265136217236234121307183192195108248D183170191137227156143158117188290308323226432248262275191357E268211272205337227185224170273410381439320615354326374270504F12612713016172100101104181391831952272830815917318025248G201125183166231165105154142188295206298270405250176255230346Total1,2791,0581,2988011,5911,0649041,0546821,3011,9551,8752,1361,2802,8961,6781,5981,7861,1042,397From Figs. 10 and 11, it is apparent that for every calculated value of maintenance density, there are four data points representing the route distances calculated using the four algorithms. The order of the shortest to longest route is almost always the same. Algorithm 2 generally resulted in the shortest route, followed by Algorithms 1, 4, and 3. The algorithms without prioritization (i.e., Algorithms 1 and 2) always resulted in the shorter route distance, compared to Algorithms 3 and 4, which had two priority levels.Comparing the routing heuristics, the insertion heuristic (Algorithms 2 and 4) resulted in shorter route distances in all but two instances compared to the greedy heuristic (Algorithms 1 and 3). The two instances were both in the F sector for gas meter replacements without prioritization, in which the greedy algorithm produced a slightly shorter route (1.2 and 2.4 km shorter, respectively, for Maintenance programs 1 and 2). These two instances represent the shortest maintenance routes calculated for both maintenance programs and is a potential indication that the insertion heuristic may not be suitable for routing in sectors consisting of very few objects.Fig. 12 compares the distance of the integrated routes with the total distance covered when objects are inspected or replaced separately. Both maintenance programs are included.The difference in route distance between points of the same shape for the same maintenance density represents the efficiency gains that are achieved from integrating maintenance tasks. The gains appear highest when Algorithms 3 and 4 (i.e., two priority levels) are used.Fig. 13 permits a visual comparison of efficiency gains (i.e., decrease in route length) by sector.It is apparent from Fig. 13 that efficiency gains increase with maintenance density in all sectors, but this trend is more apparent in some sectors than others (e.g., sectors B and E compared to G). Normalization is required to compare the gains between sectors. This is done by dividing the gains obtained from routing by the number of objects inspected or replaced in the sector. Fig. 14 shows the normalized gains for both maintenance programs.The influence of maintenance density on efficiency gains is apparent from Fig. 14, and it appears that the differences between sectors observed in Fig. 13 can be attributed to maintenance density and that the relationship appears to follow an exponential decay curve. The influence of the algorithms is also evident. The normalized gains are highest for Algorithms 3 and 4 in which two priority levels are set. This observation is highlighted in Fig. 15.A nonlinear regression was performed with a modified exponential decay model. The regression was done in R using the drc package (Ritz 2016) and the results are shown in Table 7 and Fig. 16. The model has the following parametrization (1) Table 7. Model parameters for fitted curvesTable 7. Model parameters for fitted curvesAlgorithmEstimateStandard errorEstimateStandard errorEstimateStandard error10.057650.016490.264660.0283281.411725.120420.052890.009130.26480.0226463.074112.585530.102950.012380.437560.0381354.794710.721840.087760.011760.42170.0360154.751510.0592Table 8 aggregates the results of the two maintenance programs for all sectors when tasks are done separately versus integrated. To calculate the number of workdays saved, it is assumed that the workday is made up of two components, productive time spent in the work sector and unproductive time (e.g., breaks, time spent travelling to work sector, etc.). A working day is 8.5 h long, of which 6 h are productive. An average travel speed of 30  km/h is used.Table 8. Summary of resultsTable 8. Summary of resultsMaintenance programAlgorithmRouting heuristicNumber of priority levelsTotal distance of all routes when executed separately (km)Total distance of all integrated routes (km)Total gains (km/ year)Work days saved per year (assuming 6  h/day. 30  km/h)11Greedy12,5891,1951,394812Insertion12,2099891,220713Greedy23,8391,9211,9181114Insertion23,4111,6101,8011021Greedy14,4361,5912,8451622Insertion13,7041,3012,4021323Greedy27,2452,8964,3492424Insertion26,1652,3973,76821DiscussionThis study highlights the role of digitalization in improving operational maintenance programs for water and gas networks. Despite the significant progress that has been made in digitizing municipal infrastructure and implementing digital technologies, there is still much untapped potential. Real-world examples are needed to further spur adoption. This paper contributes by presenting a methodology that can be applied to Steps 4 and 5 of the IAM process. The methodology is applied in a case study, in which GIS databases and routing algorithms are leveraged to investigate the potential gains in efficiency from integrating maintenance tasks into common work packages. The problem is modeled as an HTSP and four different routing algorithms are explored to represent the various ways utilities currently plan and carry out maintenance routes. The influence of maintenance density (i.e., the number of objects to inspect or replace in a maintenance sector) on efficiency gains is also considered and shown for a large municipality.The results show that the benefits of integrating maintenance tasks into combined work packages are considerable and vary significantly depending on the algorithm used in route planning and the maintenance density. Referring to Table 8, the efficiency gains amounted to a reduction in total route length of 1,220–4,349  km/year. This reduction translates to a time savings of 7–24 work days per year for a maintenance crew along with the associated avoidance of carbon emissions and other air pollutants. The greatest reduction (4,349  km/year) is observed when Algorithm 3 (i.e., greedy heuristic with two priority levels) is used to determine the routes for Maintenance program 2. The smallest reduction (1,220  km/year) took place with Algorithm 2 (i.e., insertion heuristic with one priority level) for Maintenance program 1.The influence of algorithm selection on efficiency gains is most evident in Fig. 16, where an exponential decay model is fitted to the results by algorithm. The benefits of integration are highest in Algorithm 3, followed by Algorithms 4, 1, and 2. In Algorithms 3 and 4, two priority levels are set, but only one priority level for the others. Based on the results of this case study, it appears that the potential for efficiency gains from maintenance integration is more dependent on how prioritization is used in route planning than the routing heuristic. The influence of prioritization is highlighted in Fig. 15. Comparing the routing heuristics, the gains are consistently larger when the greedy heuristic is used. This can be seen from Fig. 16 by comparing Algorithms 1 and 2 as well as Algorithms 3 and 4. It should be noted that no recommendations are being made for these algorithms. The use of these four algorithms is simply to model the different approaches currently used by utilities for route planning. Two key aspects of how these approaches may differ is in the use of prioritization and the specific routing heuristic. The greedy heuristic represents a simple, intuitive approach for determining a maintenance route, whereas the insertion heuristic is a more sophisticated, better performing one.Maintenance density (i.e., the number of objects to inspect or replace per 10 km of road in the sector) is shown to have an even larger influence on efficiency gains than algorithm selection. In order to compare the results obtained from different sectors, the gains are normalized by dividing the reduction in route length from integration by the number of objects inspected or replaced in the sector (Figs. 13 and 14). The exponential decay models shown in Fig. 16 and Table 7 provide reasonable fits to the normalized gains observed with increasing maintenance density. This model implies that efficiency gains from integration will continue to rise as the density of objects requiring maintenance in a sector increases; however, the rate of gain increase will decrease exponentially and eventually reach a plateau. This implies that improvements from integrated planning will be greatest when the maintenance density of the sector is low, such as in rural, sparsely populated areas or in geographically spacious cities common in North America. Consider the city of Ottawa, Canada, which has an approximate population of 1,000,000, an area of 2,790  km2 and 71,000 valves and hydrants to maintain (Ottawa 2020). Compared to the case study of Geneva, this represents a 3.4-fold increase in the number of objects to inspect spread over an area that is 11.6 times larger. Based on the results of this study, the gains from maintenance integration are likely to be much larger in such a city compared to Geneva.It is possible that a better understanding of the relationship between efficiency gains and maintenance density could be obtained by examining the network properties of the roads in the different sectors. Factors such as the density of intersections and driving restrictions (i.e., one-way roads) could potentially account for some of the variance observed and result in better route calculations and predictions of efficiency gains. These specific factors are not considered in this work.Thus, utilities should consider integration as a viable option to reduce the resources needed for maintenance activities without lowering service provision. The presented approach can be expanded to include other maintenance activities such as hydrant flushing, meter readings, meter calibrations, etc. It is important to emphasize that these efficiency gains are cumulative and would be obtained every year.Infrastructure managers need to weigh these findings against the additional costs of increased training requirements of staff, organizational restructuring, and increased organizational effort required to integrate maintenance tasks. For example, this investment of additional resources might not be worthwhile if worker turnover rate is high or if increased integration leads to more human errors such as workers opening/closing water valves too quickly, which may cause water hammer and subsequent pipe ruptures. On the other hand, greater variability of tasks could lead to higher job satisfaction and a higher retention rate of workers. Readers interested in literature on occupational health and satisfaction are referred to Karasek (1998). When considering such investments, it is essential to remember that these additional costs are incurred only once, whereas the benefits of higher operational efficiency accumulate with time.All utilities are similar in that they must plan and execute maintenance tasks by dispatching vehicles, but they are different in the maintenance strategies they follow and the distances that must be covered to execute these tasks. This paper provides utilities with answers to the benefits that can be expected from integrated planning, the influence of route planning algorithms, and maintenance density so that utilities can produce quantitative estimates for their own networks.ConclusionsThe maintenance of municipal distribution networks often consists of dispatching trained personnel in vehicles to perform inspections and interventions such as the inspection of isolation valves and hydrants and the replacement of water meters and gas meters. Infrastructure managers are interested in accomplishing these tasks as efficiently as possible to reduce emissions and lower costs. Leveraging the potential of digitalization, specifically the use of algorithms and the disappearance of data silos represents a feasible path to achieving this aim. Thus far, researchers have ignored the potential benefits of integrated planning of such maintenance tasks.This study investigated the potential efficiency gains that could be obtained from integrated planning, using a large Swiss municipal network as the case study. The problem is modeled as a hierarchical travelling salesman problem and four algorithms are used to model the current practice of determining maintenance routes. Tangible improvements are demonstrated compared to the status quo. As stated in the discussion, integrated planning produced a reduction in total route length of between 1,220 and 4,349  km/year and a time savings of 7–24 work days per year depending on the routing algorithm and maintenance program.The main findings are that gains from integrated planning are considerable. These gains continue to rise as the maintenance density of a sector increases, but the rate of efficiency gain decreases exponentially with maintenance density. This implies that sparsely-populated, rural areas or geographically spacious cities will benefit the most from integrated planning of maintenance tasks. These benefits need to be weighed against the costs of additional worker training and organizational restructuring; however, it should be noted that these additional costs will only be incurred once while the benefits of improved operational efficiency will be cumulative.Digitalization is permitting infrastructure managers to fundamentally reassess all steps of the IAM process. Increased adoption and implementation of digital technologies will undoubtedly lead to improved management of municipal infrastructure networks. It is likely that this trend will not only influence the IAM process but will also reshape the organizational structures of utilities and the utility sector itself.Data Availability StatementSome or all data, models, or code generated or used during the study are proprietary or confidential in nature and may only be provided with restrictions (e.g., anonymized data) and the written consent of the utility.AcknowledgmentsThe authors gratefully thank the members of the water utility of Geneva (SIG) for their generous financial and logistical contributions, which have made this research possible.References Adey, B. T. 2019. “A road infrastructure asset management process: Gains in efficiency and effectiveness.” Infrastruct. Asset Manage. 6 (1): 2–14. Alegre, H., J. M. Baptista, E. Cabrera, F. Cubillo, P. Duarte, W. Hirner, W. Merkel, and R. Parena. 2016. Performance indicators for water supply services. London: IWA Publishing. Allweyer, T. 2010. BPMN 2.0: Introduction to the standard for business process modeling. Norderstedt, Germany: Books on Demand. AWWA (American Water Works Association). 2006a. Distribution valves: Selection, installation, field testing, and maintenance (M44). Denver: AWWA. AWWA (American Water Works Association). 2006b. Installation, field testing and maintenance of fire hydrants (M17). Denver: AWWA. AWWA (American Water Works Association). 2012. Water meters: Selection, installation, testing, and maintenance (M6). Denver: AWWA. Bose, S., and C. Kielhauser. 2019. Transforming infrastructure: Frameworks for bringing the fourth industrial revolution to infrastructure: Community paper. Geneva: World Economic Forum. Carey, B. D., and J. S. Lueke. 2013. “Optimized holistic municipal right-of-way capital improvement planning.” Can. J. Civ. Eng. 40 (12): 1244–1251. Cromwell, J. 2002. Costs of infrastructure failure. Denver: American Water Works Association. Dong, J., N. Yang, and M. Chen. 2007. “Heuristic approaches for a TSP variant: The automatic meter reading shortest tour problem.” In Extending the horizons: Advances in computing, optimization, and decision technologies, edited by E. K. Baker, A. Joseph, A. Mehrotra, and M. A. Trick (eds) 145–163. Berlin: Springer. Eglese, R., B. Golden, and E. Wasil. 2015. “Chapter 13: Route optimization for meter reading and salt spreading.” In Arc routing: Problems, methods, and applications, edited by A. Corberán and G. Laporte, 313–320. Berlin: Society for Industrial and Applied Mathematics. Engelhardt, M. O., P. J. Skipworth, D. A. Savic, A. J. Saul, and G. A. Walters. 2000. “Rehabilitation strategies for water distribution networks: A literature review with a UK perspective.” Urban Water 2 (2): 153–170. Infraguide. 2003. “Developing a water distribution system renewal plan.” In Best practices: National guide to sustainable municipal infrastructure. Ottawa: Infraguide. Karasek, R. 1998. “Demand/Control model: A social-emotional, and psychological approach to stress risk and active behavior development.” In: Encyclopaedia of occupational health and safety, edited by J. M. Stellman, 34.6–34.14. Geneva: International Labour Office. Kerwin, S., and B. T. Adey. 2017. “Determination of optimal intervention programs for a water distribution network.” In Proc., Leading Edge Sustainable Asset Management of Water and Wastewater Infrastructure Conf. Zürich, Switzerland: ETH Zurich. Kerwin, S., and B. T. Adey. 2021. “Exploiting digitalisation to plan interventions on large water distribution networks.” Infrastruct. Asset Manage. 40 (Feb): 1–16. Kerwin, S., B. Garcia de Soto, B. T. Adey, K. Sampatakaki, and H. Heller. 2020. “Combining recorded failures and expert opinion in the development of ANN pipe failure prediction models.” Sustainable Resilient Infrastruct. 2020 (Jul): 1–23. Kerwin, S., B. García de Soto, and B. T. Adey. 2018. Performance comparison for pipe failure prediction using artificial neural networks. Ghent, Belgium: International Association for Life-Cycle Civil Engineering. Kielhauser, C., and B. T. Adey. 2018. “Determination of intervention programs for multiple municipal infrastructure networks: Considering network operator and service costs.” Sustainable Resilient Infrastruct. 5 (1–2): 49–61. Kielhauser, C., and B. T. Adey. 2019. “A demonstration of the use of a unified service model for urban infrastructure networks.” Infrastruct. Asset Manage. 7 (4): 269–281. Lienert, J., J. Monstadt, and B. Truffer. 2006. “Future scenarios for a sustainable water sector: A case study from Switzerland.” Environ. Sci. Technol. 40 (6): 436–442. Liu, Z., Y. Kleiner, B. Rajani, L. Wang, and W. Condit. 2012. Condition assessment technologies for water transmission and distribution systems. Washington, DC: USEPA. Makar, J., R. Desnoyers, and S. McDonald. 2001. “Failure modes and mechanisms in gray cast iron pipe.” In Underground infrastructure research: Municipal, Industrial and Environmental Applications, Proc. of the Int. Conf. on Underground Infrastructure Research, 1–10. London: CRC Press. Marlow, D. R., D. J. Beale, and J. S. Mashford. 2012. “Risk-based prioritization and its application to inspection of valves in the water sector.” Reliab. Eng. Syst. Saf. 100 (Apr): 67–74. Mennell, W. K. 2009. “Heuristics for solving three routing problems: Close-enough traveling salesman problem, close-enough vehicle routing problem, sequence-dependent team orienteering problem.” Ph.D. dissertation, Business and Management: Decision and Information Technologies, Univ. of Maryland. Rajani, B., and J. Makar. 2000. “A methodology to estimate remaining service life of grey cast iron water mains.” Can. J. Civ. Eng. 27 (6): 1259–1272. Rosenkrantz, D. J., R. E. Stearns, I. Lewis, and M. Philip. 1977. “An analysis of several heuristics for the traveling salesman problem.” SIAM J. Comput. 6 (3): 563–581. Sarni, W., C. White, R. Webb, K. Cross, and R. Glotzbach. 2019. Digital water: Industry leaders chart the transformation journey. Nanjing, China: International Water Association and Xylem Inc. Scheidegger, A., J. P. Leitão, and L. Scholten. 2015. “Statistical failure models for water distribution pipes—A review from a unified perspective.” Water Res. 83 (2): 237–247. Scholten, L., A. Scheidegger, P. Reichert, and M. Maurer. 2013. “Combining expert knowledge and local data for improved service life modeling of water supply networks.” Environ. Modell. Software 42 (1): 1–16. Shuttleworth, R., B. L. Golden, S. Smith, and E. Wasil. 2008. “Advances in meter reading: Heuristic solution of the close enough traveling salesman problem over a street network.” In The vehicle routing problem: Latest advances and new challenges, 487–501. Berlin: Springer. SVGW (Swiss Association of Gas and Water Utilities). 2012. “Standard on water distribution (W4).” In Part 4: Operation and maintenance. Zürich, Switzerland: SVGW. Tscheikner-Gratl, F. 2016. “Integrated approach for multi-utility rehabilitation planning of urban water infrastructure: Focus on small and medium sized municipalities.” Doctoral dissertation, Dept. of Infrastructure, Univ. of Innsbruck. Tscheikner-Gratl, F., R. Sitzenfrei, W. Rauch, and M. Kleidorfer. 2016. “Integrated rehabilitation planning of urban infrastructure systems using a street section priority model.” Urban Water J. 13 (1): 28–40. van Riel, W., J. Post, J. Langeveld, P. Herder, and F. Clemens. 2017. “A gaming approach to networked infrastructure management.” Struct. Infrastruct. Eng. 13 (7): 855–868. Yuan, X.-X. 2016. “Principles and guidelines of deterioration modelling for water and waste water assets.” Infrastruct. Asset Manage. 4 (1): 19–35. Zhao, J. Q., and B. Rajani. 2002. Construction and rehabilitation costs for buried pipe with a focus on trenchless technologies. Ottawa, ON: Institute for Research in Construction.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *