Analyzing the Effect of Pressure Management on Infrastructure Leakage Index in Distribution Systems based on Field Data


 In this study, the relationship between pressure management (PM) and infrastructure leakage index (ILI) in distribution systems was investigated based on the field data. Before PM, non-revenue water (NRW) rates were calculated as 48.44%, 76.49% and 36.57% and ILI indicators were determined as 16.97, 22.90 and 26.88, in three DMAs. The leakage volume is also calculated by the FAVAD equation and compared the field data. With the implementation of PM, although NRW ratios decreased, the ILI did not improve in the same rate. ILI class was improved in 3 regions, ILI class dropped from A to B in one region. Although the ILI class did not change in the 5 regions, the loss rates decreased. Using the ILI alone in regions where PM is used can create misleading results in performance analysis. It is thought that it would be more accurate to evaluate losses with performance indicators in PM areas.


Introduction And Theoretical Background
The most serious problem encountered in WDSs is high leakages. Based on increasing water demands and leakages and decreasing water resources, leakage reduction strategies should be developed for the continuation of sustainable water services (Liemberger et al. 2006;Vairavamoorthy 2011). The basic methods in leakage management are PM, improving the repair speed and quality, material management and active leak control (ALC) ( Schwaller & van Zyl 2015). Although the relationship between leakage and pressure was tested and tried in many eld applications, it is mainly explained by the ori ce equation (May, 1994). Liemberger and McKenzie (2005) developed the targets for the daily leakage level per connection in various pressure classes for developed and developing countries. Limit values de ned for developed countries are 5 l/connection/day/ pressure, 150 l/connection/day at 20-40 meters pressure. In developing countries, two times of these values are directly targeted. (Tabesh et al. 2008) assessed leakages for different pressure levels based on the water balance and minimum night ow (MNF), the network hydraulic model and showed that leakage reduction can be achieved with PM. Nicolini and Zovatto (2009) addressed the problem of optimal pressure management in water distribution systems through the introduction and regulation of pressure reducing valves by applying the multiobjective genetic algorithms. Wegelin & McKenzie (2010) developed a PM and reduction strategy to reduce leaks in the city of Cape Town by using pressure reducing valve (PRV) and analyzed the changes in MNF rates. According to eld measurements, signi cant reductions in MNF rates have occurred and signi cant gains were achieved in preventing leaks by controlling the pressure. Thornton (2011) stated that pressure zones should be created rst and the most suitable PRV should be selected to obtain the expected bene ts from PM in the network. Page et al. (2017) proposed the remote real-time pressure control methodology to reduce the failures, leakages and decrease the excessive consumption. Reducing pressure in a water distribution system leads to a decrease in water leakage, decreased cracks in pipes, and consumption decreases. Creaco and Walski (2017) presented an economic analysis of PM method to reduce leaks and burst reduction and applied conventional and real-time controlled PRVs. Authors emphasized that there is no need for PM in areas with low leakage levels and low operating costs. Fontana et al. (2018) applied the real-time control of pressure for leakage reduction in WDSs by regulating the pressure level as nearly constant at critical point. The results showed that pressure controller is effective in optimizing and managing the pressure over the entire WDS. Lati et al. (2018) expressed that the number of failures, leakage rate and consumption increase in distribution systems by the effect of high pressure. It was stated that the pressure management is an effective strategy for improving the system operating condition. Monsef et al. (2018) denoted that the pressure management in urban water distribution networks is one of the options that can signi cantly reduce water loss. The results showed that by applying PM, the network background leakage and the energy consumption have been reduced by 41.72% and 28.4%, respectively, compared to a non-management mode. Haider et al. (2019) stated that water loss control actions, including ALC, passive leakage control, pressure management, and infrastructure asset management, are performed by the municipalities up to the service connections until the cost of these actions becomes equal to the cost of the water lost. Moslehi et al. (2019) proposed a methodology for estimating the short-run economic leakage level with respect to ALC by considering the eld data.
The results revealed that the economic leakage level is signi cantly affected by the operating pressure and infrastructure condition. Garcia et al. (2020) examined the in uence, extent, and impacts of pressure on the physical integrity of water mains. Authors reported that stronger correlations between two variables, which re ects the role of pressure on increased failure rates Areas with a consistently high failure rates.
As can be seen, bene ts such as reduction of existing leaks and failures are obtained from PM. However, signi cant costs such as room construction, device and equipment selection and placement, and automation system are incurred (Charalambous and Kanellopoulou 2010). Therefore, cost-bene t analysis should be done by considering the data obtained by analyzing the cost components (Kanakoudis and Gonelas 2014). The physical, operational and environmental factors are effective in the occurrence of new leakages in WDSs (Farley et al. 2008). Pressure can be at high levels depending on the topographic conditions and the location of the water tanks. In such cases, PM is applied to regulate the pressure and reduce uctuations (Creaco and  (for varied area leakages). In these pipes, the crack expands with the effect of pressure and the amount of water lost per unit time increases. Similarly, as the crack will narrow more by decreasing the pressure, the leakage will decrease more. On the other hand, in distribution systems with pipes with hard material, the N1 is determined between 0.5 and 1 (for xed area leakages). In these systems, there will be no change (expansion or contraction) in the crack due to the pressure change. It has been suggested that the N1 will be taken as 1 in systems with unknown pipe material type or with mixed materials (Lambert and Thornton 2012). Considering the main line and service connections in the pilot areas, mixed pipe material is available in DMA1 (mostly Cast type), DMA5 (predominantly Steel) and DMA6 (predominantly Cast type). Therefore, N1 coe cient was chosen as 1 in these regions.
The effect of pressure change on leakage has been explained in detail in the literature. However, in the implementation of PM, there are many cost items such as eld manufacturing, device supply, labor and data transfer. Therefore, it is very important for e ciency to calculate the expected bene ts correctly and evaluate the leakage level according to the correct indicators before applying PM. In the following sections, the effect of PM on leakage and performance change has been analyzed based on real eld data.
The most appropriate performance analysis should be carried out to test the effectiveness of leakage prevention methods and monitor e ciency by using applicable indicators (Lambert et al. 1999;Liemberger et al. 2006). The indicators based on the percentage of system input volume, network length and number of service connections is used (Lambert 2002;Lambert et al. 1999). However, these indicators do not consider the pressure or network physical known effects on leakage. The NRW rate (although the leakage volume remains constant) changes depending on the regular measurement of the inlet volume and authorized consumptions. Therefore, using an indicator that considers the data representing the system and operating conditions will provide a more accurate assessment. Lambert (1997) emphasized the need for a performance indicator that will allow the network to be evaluated according to many variables and to be compared at an international level. Moreover, the ILI which is the ratio of "Current Annual Real Losses (CARL)" and " Unavoidable Annual Real Losses (UARL)", is frequently used indicator for evaluating performance and comparing systems (equation 2) (Lambert et al. 1999). UARL (liter/day) representing the technically lowest leakage level in a system, is calculated by equation (3) (Lambert et al. 1999). UARL is sensitive to system operating pressure and network characteristics and is directly affected by changes in pressure.
P: the average pressure (m), Lm: the main length (km), Nc: the number of the service connection, Lp: the length of service connection on private property (km).
Since UARL considers the physical and operational data of the system, the ILI offers a more objective evaluation than other indicators ( 2021) stated that ILI accommodates the fact that real losses will always exists, even in the very best and well managed system. As the current pressure regime may not be optimal, ILI should always be interpreted with some measure of pressure and only used for tracking progress if all justi able pressure management has already been completed. Especially in systems where PM is applied, an indicator that provides objective evaluation of the change in system performance should be used. For this reason, ILI indicator should be analyzed according to eld data in PM applied systems. That is, the applicability of these indicators alone in PM should be tested according to eld data. Thus, it will be possible to evaluate the PM performance more accurately in leakage management. In this study, the relationship between PM and ILI in WDSs was investigated and the effectiveness of ILI in monitoring system performance was analyzed. For this purpose, PM was applied in 3 pilot areas in the application area, process indicators and ILI were calculated based on eld data in conditions before and after PM to analyze the effect of PM on the ILI. In this way, the positive or negative effect of the decrease in leakage on the ILI due to PM was evaluated. Moreover, a total of 6 DMAs where PM is not applied were selected and possible changes in ILI and other indicators in case of PM application in these regions were analyzed. Thus, reference information was generated in testing the applicability of PM by analyzing possible bene ts before applying PM. In addition, the effect of the selection of the N1 according to different pipe material types in the FAVAD equation, which reveals the effect of PM on leaks, on the results was also analyzed and discussed.

Study Area And Data
In this study, Malatya (Turkey) WDS with length of approximately 2,000 km network and number of customers 350000 was chosen for eld applications (MASKI, 2020). ALC activities DMA design were carried out between years 2016-2020 and water balance, ow pressure monitoring and MNF monitoring are performed in DMAs. A total of 9 DMAs were taken into consideration and the areas with and without PM were determined (Fig. 1, Table 1). The network lengths, customer densities, water production / operation costs and water loss amounts differ were selected. Inlet ow rates are regularly measured and monitored instantly with the SCADA system in DMAs. Network length, number of customers, service connection length and total consumption were determined for each region by using customer management and GIS databases. The average pressure of the system was obtained by measuring with pressure gauges placed in each isolated area at the average zone point. Since all isolated zones are integrated into the SCADA system, instantaneous pressure changes are monitored. Monthly data were taken into account in establishing the water balance in isolated regions.  In this study, the necessary data was obtained by using the SCADA and customer management systems. Data of the network length, number of the service connections, average length of service connection on private property and average pipe type in DMAs were obtained by using GIS database. In addition, pressure management was implemented by placing pressure control valves at the inlets of the isolated zone. Inlet ow rates and pressure data in each isolated zone are regularly measured and monitored with a SCADA system. In this study, xed output PRV for DMA1, and ow sensitive PRV for DMA2 and DMA3 are used. Flow rates and pressures occurring before and after PM in DMA are monitored by SCADA (Fig. 2).
Before applying PM in DMA1, the inlet ow was measured as 47.2 l/s and the pressure was 65 m (Fig. 2). Considering the leakage-pressure relationship, the high pressure in the region is expected to cause new leakages or increase in leakages in existing failures. In DMA2, the initial pressure was 51 m and the inlet ow rate was about 50 l/s. Finally, the pressure was 50 m and the inlet ow rate was 36.48 l/s before PM in DMA3. Although the pressure in DMA2 and DMA3 is lower than DMA1, the possible effects of pressure on leakage/failure should be monitored. PM was applied in DMAs to reduce the effect of pressure on existing and new leakage. As a result of eld measurements, NRW rate, UARL, CARL and ILI were calculated for three regions before and after PM.
NRW rates in DMAs before PM were calculated as 48.44%, 76.49% and 36.57%, respectively. It is seen that these rates are higher than the limit values (> 25%) recommended in the international literature. As it is known, it is not enough to evaluate the NRW percentage alone, so UARL and ILI were calculated for each region. The ILI in DMA1, DMA2 and DMA3 were determined as 16.97 (D), 22.90 (D) and 26.88 (D), respectively. It can be said that the NRW rates and ILI classes in each region are at a very poor level. For this reason, it seems that the system should be intervened in order to reduce leakage and improve system performance in the regions. With PM, the pressures were reduced from 65 m to 36 m in DMA1, from 50.8 m to 40 m in DMA2 and from 50.1 m to 40 m in DMA3 (Fig. 3). ILI changes in the regions where PM is applied are shown in Fig. 3.
Although there were signi cant reductions in NRW rates (average 21.75%), the ILI did not improve at the same rate. In DMA1, although a gain of approximately 10 Berardi et al. (2015) analyzed that how hydraulic models are relevant to support pressure control strategies at both planning and operation stages on the real WDN. Moreover, the effectiveness and changes of the ILI indicator was presented for tracking progresses in leakage management. The results demonstrated that using the ILI to assess the leakage reduction achievements is not consistent with the expected hydraulic WDN behavior. Consequently, the use of ILI for regulation purposes in the WDN sector would be misleading without the support of appropriate hydraulic modelling. In more detail, the analysis reported herein shows that, depending on the current leakage rate and pressure control scheme, the ILI might be invariant or even increase in the face of a large reduction of leakage volume from the controlled network.
As a result of this decrease in NRW value, the ILI has decreased from 26.88 (D) to 12.80 (C) (Fig. 3). As can be seen, the type of pipe material affects the NRW ratio and the ILI in the region where the PM was applied. Especially in areas with exible materials, the crack narrows due to the decrease in pressure and the leakage rate and total leakage volume per unit time decrease. It was determined that these decreases in leakage due to PM in regions with exible material density are re ected in the ILI.
Depending on the PM, the possible leakage volume is calculated according to the FAVAD equation in case the pressure is reduced to the desired level. The evaluation was made by comparing the eld data and the results of the FAVAD approach (Table 1). The N1 was selected according to the pipe material in DMA. Considering the main line and service connections in the pilot areas, mixed pipe material is available in DMA1 (mostly Cast type), DMA5 (predominantly Steel) and DMA6 (predominantly Cast). Therefore, N1 was chosen as 1 in these regions ( Table 1). The N1 was chosen as 1.5 for DMA2 (PVC) and for DMA3 (HDPE) the coe cient was determined as 2. It is seen that the loss levels calculated according to the FAVAD equation depending on the PM in DMAs are very close to the loss levels measured in the eld (Table 1). Considering the effect of the N1 on the FAVAD equation, it is very important to choose the most appropriate N1 value for the pipe type in the region. For this reason, it is thought that the gain to be obtained by using the relevant equation and N1 for the regions where PM is planned can be calculated in a way that is very close to the reality by knowing the weighted pipe type of the network exactly.

Application Of Favad Approach To Other Dmas
In the previous section, it was determined that the leakage rates calculated with the FAVAD equation in DMAs are compatible with the eld data. For this reason, in the second stage, 6 DMAs without PM were selected. The results obtained in case of applying PM in these regions were evaluated and possible bene ts (leakage reduction, ILI change) were estimated. For this purpose, the characteristic data of the regions were collected and, the GGS volume and ratio, UARL, CARL and ILI were calculated.
The NRW rates in DMAs under current conditions vary between 11.05% (DMA6) and 71.75% (DMA8). In the current situation, DMA6 is at a very good level, and the rates in DMA7 and DMA9 are at an acceptable level. Based on ILI classi cation, DMA 6 and DMA9 are in class A, DMA7 is in class C (DMA7) and DMA 4, DMA 5 and DMA8 are in class D. When the NRW rates and ILI in DMAs are compared, the NRW rates in the A-class regions are generally lower than 25%. However, in DMA7 and DMA9, although leakage and NRW volumes are very close to each other, the ILI is 8.20 (C) in DMA7 and 3.97 (A) in DMA9. As can be seen, although the leakage volumes are close to each other, the differences in the inlet volumes proportionally ensure that the performance of the system is good or bad.
In the second stage, the changes in leakage and ILI in case of PM were analyzed (Fig. 3). The leakage can be reduced between 0.6 l/s (DMA9) and 8.21 l/s (DMA8) in 6 regions. The gain ow rate to be obtained is directly related to the weighted pipe type of the network as well as the pressure change. In DMA4 and DMA5 where pressure reduction rates are very close to each other, signi cant decreases in leaks (7.12-17.31%) are expected depending on the pipe material.
On the other hand, NRW rates in DMA6, DMA7 and DMA9 are at relatively acceptable levels (11.05%, 26.23% and 24.47%), while ILI indicators are at good and intermediate levels (A, C and A). Although these areas are not considered as the priority area in theoretical intervention, they are very suitable for PM due to high pressures. It was observed that serious gains will be achieved by pulling the pressure to ideal levels. Useful ow rates will be added to the system by applying PM. The expected changes in ILI and NRW by adding these bene cial ow rates to the system are shown in Fig. 3.
When the possible changes in the NRW rates are examined, a decrease is expected in all DMAs. It is seen that there will be a serious decrease in the GGS ratio in regions where the current loss rate and pressure change is high (DMA8). Especially in DMAs where N1 is greater than 1, the change reaches serious levels compared to other regions (DMA7 and DMA8). The main reason for this is that elastic pipes are highly sensitive to pressure. In exible pipes, it is thought that the crack diameter will expand more than rigid pipes with high pressures. In regions with more rigid pipe types (DMA4 and DMA9), the changes in losses due to pressure remain at a lower level.
Moreover, the reduction of the NRW rate depending on the pressure was also examined with the ILI and quite important results were obtained. The decrease in NRW rates with the reduction of pressure does not always cause a positive change in the ILI. As a result of PM, leakage rates prevented in DMA4 and DMA9 are 0.92 l/s and 0.60 l/s, respectively. Moreover, the decrease in NRW rates is about 4% (38.98-34.30% and 24.47-20.04%). However, although the NRW rates decreased, an increase was observed in the ILI indicator. In DMA4, the initial ILI was 19.20 (D), while it was calculated as 23.54 (D) at the end of the PM. Similarly, in DMA6, the ILI increased from 3.97 (A) to 5.12 (B). The main reason for this is that the pressure variable affects differently depending on the N1 in the calculation of FAVAD and UARL.
When the UARL used in ILI analysis is examined, the pressure is a direct factor and the change in pressure will change the ILI linearly. However, the N1 is used as the exponent of the pressure change in the FAVAD equation. The new leakage amount will change in a nonlinear way according to the value of the N1. In DMA4 and DMA9 where there are rigid pipes in the network, the UARL and ILI decreases and increases at the same rate with the rate of change in pressure. However, in DMAs, the leakage will be affected by the N1 (0.5) of the change in pressure. As a result, pressure variation in these regions can have a negative effect on ILI, although it causes a decrease in leakage volume. Likewise, since the N1 is selected as 1 in DMA5 and DMA6, the change in pressure affects ILI and NRW at the same rate, even though the ow rate has bene ted in these regions, ILI will not change or will be affected very little (Fig. 3). As a result, ILI class was improved in 3 regions (DMA3: D > C, DMA7: C > B, DMA8: D > B), in one region (DMA9) ILI class dropped from A to B. Although the ILI class did not change in the other 5 regions, it was observed that the loss rates decreased. For this reason, using the ILI alone in the regions where PM is used can create misleading results in performance analysis. It is thought that it would be more accurate to evaluate losses with performance indicators in PM areas.
In this study, the bene ts and ILI changes to be obtained in case of different pipe types in the isolated zone were also calculated ( Table 2). For this purpose, an application has been carried out for the cases of different pipe material in DMA 8. Although DMA 8 consists entirely of HDPE pipes under current conditions (N1 = 2), the change in losses was calculated in the case of asbestos, steel and PVC pipes respectively (N1 = 0.5, 1.0, 1.5). In this context, if the current pressure level is reduced from 51 meters to 20 meters, the rate of NRW decreases from 71.75-28.06% under current conditions. Moreover, according to the initial value, it has been calculated that a signi cant reduction will be achieved by decreasing the ILI parameter from 16.56 to 6.49. If the same region consists of asbestos pipes, it is expected that there will be a 10% reduction in NRW ratio, while it is seen that the ILI parameter will increase from 16.56 to 26.43. Similarly, if the line consists of steel pipes, although a 22% decrease is predicted in NRW, it is seen that the ILI parameter does not change (16.56 -Class D). As can be understood from the examples, the network pipe type has a quite important place in pressure management. It is seen that more bene ts can be obtained in exible pipes compared to rigid pipes. In addition, it has been observed that if the N1 is chosen 1 or less than 1, the ILI parameter may be insu cient to evaluate the performance of the pressure management application. As a result, although losses in the network have been reduced by pressure management, the ILI indicator may remain stable or increase. As it is known, the ILI performance indicator offers signi cant bene ts in terms of evaluating the initial performance of WDSs and comparing them with other networks. The current performances of the networks are calculated with basic data such as main length, number of service connection, and length of the service connection on private property and pressure, which provides serious insight into the loss status of the network. For this reason, the ILI indicator provides a signi cant advantage in determining the priority region where water losses should be intervened primarily. In addition, the performance of leakage reduction methods that are acoustic listening method, fault repair management and network rehabilitation can be monitored with the ILI indicator. In pressure management, it can be used to evaluate the performance of the ILI method in cases where the N1 is selected greater than 1 (if the line is mainly composed of exible pipes).

Conclusions
In this study, the relationship between PM and the ILI in WDS was analyzed and the effectiveness of the ILI in monitoring system performance was investigated. is thought that the system should be evaluated with various performance indicators together with ILI in PM studies. Figure 1 General view of the study area and isolated areas (Source: https://www.harita.gov.tr/urun/map-of-turkeys-administrativeboundries/266) Figure 2 Flow-pressure graphs in isolated areas