You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/342310921

Risk-based classification of contamination events for intermittent water


distribution network security

Conference Paper · June 2020

CITATIONS READS

0 131

4 authors:

Shweta Nabira Maheshwari Rathi Swapnil Kamble


National Institute of Technology, Kurukshetra Nagpur Smart and Sustainable City Development Corporation Limited
35 PUBLICATIONS   288 CITATIONS    11 PUBLICATIONS   160 CITATIONS   

SEE PROFILE SEE PROFILE

Aabha Sargaonkar Rajesh Gupta


National Environmental Engineering Research Institute Visvesvaraya National Institute of Technology
26 PUBLICATIONS   887 CITATIONS    114 PUBLICATIONS   1,723 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Sensor placement for intermittent water distributiin network security View project

Risk based sensor placement approach View project

All content following this page was uploaded by Shweta Nabira Maheshwari Rathi on 19 June 2020.

The user has requested enhancement of the downloaded file.


HYDRO 2015 INTERNATIONAL IIT Roorkee, India, 17-19 December, 2015
20th International Conference on Hydraulics,
Water Resources and River Engineering

Risk-based classification of contamination events for intermittent water


distribution network security
Shweta Rathi1, Swapnil Kamble2, Abha Sargaonkar3 and Rajesh Gupta4
1Shweta Rathi, Research Scholar, Department of Civil Engineering, Visvesvaraya National Institute of
Technology, Nagpur, India, 440010.
2Project Assistant, ESDM division, CSIR-National Environmental Engineering Research Institute (NEERI),

Nagpur, Maharashtra, India,


3Principal Scientist, CSIR-National Environmental Engineering Research Institute (NEERI), Nagpur,

Maharashtra, India,
4Rajesh Gupta, Professor, Department of Civil Engineering, Visvesvaraya National Institute of Technology,

Nagpur, India, 440 010.


Email, rajeshguptavnit@hotmail.com
Telephone/Mobile No., +91 9823640157

Abstract
Water distribution networks (WDNs) are provided to supply safe quality of water in required quantities at
desired pressure throughout its design life. The water quality is monitored at salient locations throughout the
network to ascertain delivery of safe water. The water quality deteriorates in the distribution network because of
several reasons. The poor maintenance of the network causes leakage of water through pipeline that may form a
zone of contamination outside. In intermittent system, low pressure inside the pipeline during non-supply hours
causes ingress of contaminated water into the pipeline. The condition would be more crucial if sewerline/ open
drain/ or dirty water gathered in depressed area are very near to water pipeline. It is therefore necessary to
identify risk prone areas of contaminant ingress. The pipes/nodes of water distibution network can be classified
into different risk zones depending upon the risk values. Monitoring of water quality should be carried out so
that any accidental contamination through these risk-prone areas is detected at the earliest to safe gaurd the
consumers. In this study, an integrated risk assessment of water distrbution system (IRA-WDS) model is used
for identifying risk-prone areas. Based on involved risk, contamination events are classified as high, medium,
low and very-low risk events. The monitoring of water quality is proposed at such locations to identify any
accidental contamination through these risk prone areas at the earliest.

Keywords, Contamination events, intermittent water supply, Risk prone areas, Water quality monitoring, Water
distribution system.

1. Introduction

The pipelines in a water distribution network (WDN) are always under pressure in a continuous mode
of water supply. Hence, chance of accidental contamination due to ingress of polluted water is very
less. On the other hand, the intermittent mode of water supply having 1 to 2 hours of daily supply
force consumers to fetch water at the lowest possible level by either constructing underground tank or
digging a pit. The consumer withdraw as much water as possible during the supply period resulting in
very low pressure in the system. In such poorly maintained network, water leaks through joints and
cracks and may sometimes mix with polluted water (leakage through sewers, open drains or other
polluted water body in the vicinity of water pipeline). The pit used for collecting water itself becomes
a source of contamination. During non-supply hours, the pipeline sucks-back the mixed-water and get
contaminated. Such problems lead to increased health risk. Thus, the intermittent mode of water
supply is considered as highly vulnerable due to accidental contamination, posing a threat to public
health. Therefore, monitoring of water quality in distribution network is essential that can provide
early warning and reduces health risk to consumers.

A contaminant may enter in the network from any point and at any time. Further there could be
multiple contaminant intrusions at different locations at any time. The severity of damage due to
contaminant intrusion depends on several factors such as, location of node; direction of flow; type,
quantity of contaminant; time (day/night); etc. Thus, an important aspect in design of any monitoring
HYDRO 2015 INTERNATIONAL IIT Roorkee, India, 17-19 December, 2015
20th International Conference on Hydraulics,
Water Resources and River Engineering

system is to identify the locations at which contaminant entry be considered and the time and duration
of such entry. Large number of contamination events makes the design of monitoring system
computational intensive due to the requiremnt of large number of simulations. Usually, nodal points
are considered to reduce the number of events and ease in analysis as they are more vulnerable for
delibrate entry of contaminant through pumping. The other simple way to reduce number of
contamination events is to restrict the time say 4 to 6 events at a node of 5-min interval during the
period of higher demand. Even though the above measures would reduce the contamination events to
a large extent, the reduction may not be enough to reduce the computation work to desirable extent in
large water networks for designing a monitoring system. The recent advancements towards reduction
of number of contamination events have been reviewed in this paper. It can be observed that most of
these approaches have been developed considering deliberate contamination events wherein the
events of high consequenses are important to be considered. However, for accidental contamination
event selection, it is more important to identify the risk prone area that depends on several other
factors such as condition of pipe, its vicinity with sewer line, open drain or other foul water bodies
that may lead to a contamination event. Thus, a risk-based analysis is more proper to select accidental
contamination events as discussed later.

The aim of this paper is thus to suggest a methodology to identify risk prone nodes for accidental
contamination based on network’s structural and hydraulic conditions, soil conditions and
surrounding environmental conditions and classify them into different categories based on risk
analysis. A methodology for design of monitoring system can be developed to detect any
contamination event from risk prone area within desirable time depending upon the priority to events
from high risk prone areas.

2. Literature review pertaining to contamination event selection

As discussed earlier, only nodal points are considered to reduce the number of events (Chastain 2006;
Ostfeld et al. 2008; Xu et al. 2010; Dorini et al. 2010). Berry et al. (2006, 2009) even disregarded zero
demand nodes as possible locations of contamination at any fixed time. Krause et al. (2008, 2009)
considered only those contamination events that affect at least 10% of the water distribution network
nodes. Shen and Mcbean (2011) considered only non-terminal nodes. While Krause et al. (2008), Xu
et al. (2010), Dorini et al. (2010) considered possibility of contamination at each node in each 5-min
interval, Berry et al. (2006) considered only four times in a day for 5-min duration each, usually the
time when demand is more. Shen and Mcbean (2011) considered contamination occurring at peak
demand.

Eliades and Polycarpou (2006) and Wu and Walski (2006) generated random contamination events
using a Monte Carlo simulation technique. Preis and Ostfeld (2008) selected a few representative
contamination events whose geographic coordinates are similar to that of the entire set of possible
contamination events by developing an equation of a reduced size contamination matrix and provided
similar results when whole/full contamination events are simulated for optimization of sensor
locations. An importance-based sampling for contamination event selection is suggested by
Weickgenannt et al. (2010) and Perelman and Ostfeld (2010). Weickgenannt et al. (2010) used
heuristic methods to select those contamination events that generate more polluted water. Perelman
and Ostfeld (2010, 2012) used crossed entropy method to select critical contamination events that
have low probability of occurrence but have an extreme impact. Carr et al. (2006), Watson et al.
(2009), Xu et al. (2010) and Davis et al. (2014) suggested selection of events based on consequences.
Various citations as above are clubbed together based on the approach used for event selection are
presented in Table 1.

3. IRA-WDS Model to identify Risk Prone Nodes

Integrated risk assessment of water distrbution system (IRA-WDS) is a spatial decision support
system (SDSS) with Contaminant Ingress Model (CIM) model, Pipe condition assessment (PCA)
HYDRO 2015 INTERNATIONAL IIT Roorkee, India, 17-19 December, 2015
20th International Conference on Hydraulics,
Water Resources and River Engineering

model and Risk assessment model (RAM) (Vairavamoorthy et al. 2007). The CIM simulates the
movement of contaminated water from different pollution sources (open canals/drains and surface
water bodies, sewers, etc) through typical soils towards drinking water distribution pipes based on
contaminant zone (CZ) and contaminant transport (CT) model. The PCA model assesses the relative
condition of pipes based on physical, environmental and operational parameters, i.e. pipe age,
material, diameter, soil condition, groundwater, traffic along road; workmanship, bedding condition,
intermittency, number of breaks and bursts, leakage in the system etc. The RAM estimates the risk of
contaminant intrusion into water distribution pipes by combining the results of CIM and PCA.

Table 1. Review of methodologies based on selection of contamination events

Approaches based on Citation Contamination events

Chastain (2006) Considered contamination event at every node in the system


Berry et al. (2006) Considered 59 nodes having demands at four different times for
contamination intrusion (12 a.m, 6 a.m ,12 p.m, 6 p.m)
Berry et al. (2006, 2009) Considered only those nodes having demands as possible location of
Reducing contamination at one fixed time.
number of Ostfeld et al.(2008) All nodes as a possible injection nodes and simultaneous injections of
locations two nodes are also considered.
and few Krause et al.(2008,2009) Selects contamination events that affect at least 10% of the water
contamination in a distribution network nodes for simulation
day Xu et al. (2010) Possibility of a contamination at any node and in any 5-min interval.
Krause et al.(2008), Dorini et Considered possible contamination at each node and in each 5- min
al. (2010) for the first 24 hour of simulation.
Shen and Mcbean (2011) Considered 2912 non-terminals nodes for possible contamination
locations with contamination occurring at peak water demand
Wu and Walski (2006) Generating random events using Monte Carlo simulations
Eliades and polycarpou (2006)
Random Sampling
Ostfeld and Salomons (2004)
Ostfeld et al. (2008) Considers simultaneous contamination at 2 nodes
Development of Preis and Ostfeld (2008) Selects representative events that have similar geographic coordinate
equation of reduced with entire set of possible events
size contamination
matrix
Weickgenanth et al. (2010) Developed heuristics method to choose contamination events
Importance-based
sampling Perelman and Ostfeld (2010) Suggested crossed entropy approach to select critical contamination
events
Watson et al.(2009)
Carr et al.(2006)
High consequences
Xu et al.(2010) Considered possibility of contamination at any node and in any 5-min
contamination events
interval.
Davis et al.(2014)

4. Model Application

4.1 Study area

The WDN of Untkhana area (0.5 km2) lies between 21o7’35” to 21o8’15” latitude and 79o5’25” to
79o6’15” longitude. The network has 145 nodes and 172 pipe links in Figure 1. Among them 96 are
the demand nodes. Total demand of the network is 2149.9 m3/hour. The water supply in the study area
is intermittent, about 1 to 2 hours a day (from 6:00 a.m. to 7:30 a.m.). The existing distribution
network consists of pipes of Mild Steel (MS) laid in 1980, Cast Iron (CI) laid in 1980 and in 2003 to
2011, Ductile Iron (DI) in 2003 to 2011, and Galvanized Iron (GI) in 2003 with pipe diameter ranging
from 75 to 700 mm.

4.2 Network analysis


HYDRO 2015 INTERNATIONAL IIT Roorkee, India, 17-19 December, 2015
20th International Conference on Hydraulics,
Water Resources and River Engineering

EPANET 2.0 is used for hydraulic analysis of pressure-deficient (water starved) networks. The
outflow at a node is considered entirely dependent on available pressure. This is achieved by
considering an artificial reservoir at each node and setting the HGL at the reservoir to level of pipeline
from where withdrawal takes place (Ozger and Mays 2003, Ang and Jowitt 2006, Siew et al. 2012,
Chandapillai et al. 2012, Abdy Sayyed et al. 2015). The artificial reservoir in the model thus represent
storage vessel/infinite sink at the consumer end for which filling replicate the end user practices of
excess water withdrawn for storage till the end of water supply, or water loss after resuming the
supply due to open taps during non-supply hours. Pipe discharges and the velocity of flow are
obtained and used to determine travel time and other network parameters.

Figure 1. A WDN of Untkhana area in Nagpur city of Maharashtra State (India) simulated for
intermittent water supply (from Mohapatra et al. 2014)

4.3 Identification of risk prone nodes

In order to identify the pipes at risk for contaminant intrusion, the WDN from ESR to consumer end
was mapped as line shape file in GIS. Similarly, sewer network map was prepared in GIS. All the pipe
details for WDN and sewer network (i.e. pipe length, diameter, material, and age) were defined in the
corresponding attribute tables. Road network and open drains were digitized as line maps and geo-
referenced. Soil map, groundwater table in the study area and pressure in the system were prepared as
polygon shape files.

IRA-WDS model was set up considering all the shape files of WDN, sewer network, open drains, soil,
groundwater table, pressure in the system and their corresponding nodes with attribute data for the
infrastructure. Numerical data about the operation and maintenance (O&M) of the system e.g. annual
average pipe breaks and burst, water loss through leakage and unauthorized consumption were
derived from the records obtained from the waterworks. System specific model parameters such as
breakage and leakage rate as a function of pipe age, material, traffic conditions etc. were estimated
using the site-specific data and the methodology described elsewhere (Sargaonkar and Islam 2009;
Sargaonkar et al. 2013). During risk analysis, leakages were distributed in the entire network as a
function of age (installation year), material (corrosiveness) and traffic density.

The PCA model estimated the relative condition of the pipes in the WDN of the study area, the CIM
simulated movement of contaminated water from different pollution sources (open drains, sewers etc.)
HYDRO 2015 INTERNATIONAL IIT Roorkee, India, 17-19 December, 2015
20th International Conference on Hydraulics,
Water Resources and River Engineering

through typical soils towards drinking water distribution pipes and the RAM estimated the risk of
contaminant intrusion into water distribution pipes using the outputs from the CIM (hazard) and PCA
model (vulnerability) with appropriate weights. Risk score for each pipe was generated. Depending
upon the risk index, pipes can be classified into different risk zones. Higher risk index shows low risk
of contamination and vice versa (Sargaonkar et al. 2013). Four risk ranks are considered in this study.
Risk rank 1 having risk index between 0 - 0.5 is considered to be in very high risk of contamination.
Similarly, risk ranks 2, 3 and 4 having respective risk index between 0.5 - 0.6, 0.6 - 0.75 and 0.75 -1.0
are considered in high, medium and low risk of contamination respectively. Details of pipe data and
the results of IRA-WDS model indicating pipe numbers, i.e. pipes in Very High, High and Medium
risk are presented in Table 2.

Table 2. Details of pipe data and risk quantification using IRA-WDS model

Sr. Pipe Start End Installation Risk of


Material Diameter Remark
No. ID Node Node Year Contamination
1 1 1 2 CI 100 1980 High -
2 2 3 28 CI 100 1980 High Crossing
sewer 76
3 3 5 6 CI 100 1980 High Crossing
Sewer 75
4 4 7 1 CI 100 1980 High Parallel to
Sewer 31
within 2 feet
5 5 7 8 CI 100 1980 High -
6 6 9 7 CI 100 1980 High -
7 25 34 35 CI 100 1980 High -
8 104 127 9 CI 100 1980 High -
9 106 129 7 CI 100 1980 High -
10 103 127 126 CI 100 1980 High
11 102 126 125 CI 100 1980 High
12 111 135 113 CI 150 1980 Medium -
13 34 42 37 CI 100 1980 Medium -
14 19 17 26 CI 100 1980 Medium Crossing
Sewer 105
15 114 136 137 CI 225 1980 Medium -
16 118 140 142 CI 400 1980 Medium -
17 98 119 120 CI 400 1980 Medium -
18 144 163 2000 CI 250 1980 Medium -
19 107 182 3000 CI 225 1980 Medium -
20 164 163 182 CI 225 1980 Medium -
21 108 132 133 CI 225 1980 Medium -
22 150 163 132 CI 225 1980 Medium -
23 105 129 128 CI 150 1980 High -
24 66 81 80 CI 75 1980 High -
25 86 96 99 CI 150 1980 Medium Crossing
Sewer 461
26 54 65 104 CI 100 1980 Medium Crossing
Sewer 525
27 145 109 111 CI 150 1980 Very High -

The overall PCA statistics indicated that out of 178 pipes in the network, 1 pipe in very bad condition,
12 pipes in bad condition, 12 in good condition and 153 in very good condition (Table 3). Further
based on the risk index pipes are ranked 1 to 4 with very high, high, medium and low risk of
HYDRO 2015 INTERNATIONAL IIT Roorkee, India, 17-19 December, 2015
20th International Conference on Hydraulics,
Water Resources and River Engineering

contamination as shown in Table 3 and depicted in Figure 2. The numbers of pipes in risk ranks 1 to 4
were 1, 13, 13 and 151. The pipes in low risk areas are in very good conditions and therefore the
chance of contamination would be very less and can be ignored to reduce consideration of large
number of contamination events. Thus, the number of pipes in risk prone areas is 27 only.

Table 3. Overall statistics of pipe condition and risk assessment

PCA rank PCA Index Pipe condition No. of pipes %


1 0.000 – 0.045 Very Bad 1 0.56
2 0.046 – 0.181 Bad 12 6.74
3 0.182 – 0.545 Good 12 6.74
4 0.546 – 1.000 Very Good 153 85.96
Risk rank Risk index Risk of contamination No. of pipes %
1 0.000 – 0.522 Very High 1 0.56
2 0.523 – 0.613 High 13 7.30
3 0.614 – 0.773 Medium 13 7.30
4 0.774 – 1.000 Low 151 84.83

Figure 2 Pipes and nodes in very high, high, medium and low risks in WDN of study area
Untkhana, India

Risk assessment methodology provided classification of pipes in risk zones. However, for selection of
contamination events nodes on either end of pipes in risk zones are considered.

Total 27 number of pipes are in risk in the study area. It is observed that one pipe is on very high risk,
thirteen are on high risk and thirteen are on medium risk area. Node details (i.e pipe id and node ids)
presented in Table 2 indicates that total 42 nodes are located in risk areas, also 2 nodes in very high
risk, 18 nodes in high risk and 22 nodes are in medium risk areas.

It is observed that the zones of contaminant ingress were mainly observed near the areas where water
supply lines cross with sewer network shown in remark column (Table 2).
HYDRO 2015 INTERNATIONAL IIT Roorkee, India, 17-19 December, 2015
20th International Conference on Hydraulics,
Water Resources and River Engineering

4.4 Quantifying the probability of contamination at Risk-Prone Nodes

A simple way to quantify the probability of occurrence of events is to consider that each node in risk
prone zones (very high, high and medium risk) has same probability of occurance. However, the
probability of occurrence of event in high risk zone is more than that in medium risk zone, hence,
proper weights should be assigned based on risk score (index).

Herein, we have 42 risk prone nodes with 2 nodes in very high, 18 in high and 22 in medium risk
zone. Equal probability of occurrence of event at risk prone node gives probability of contamination
at any node among the risk prone nodes as 1/42. However, if weights are assigned based on risk index
as 5, 3.5 and 1.25, the combined probability of event occurrence at a node in high risk zone is 5% [i.e.
= (2×5)/(2×5+3.5×18+1.25×22)], and at medium and low risk zone is 2/5. Further, the probability of
occurrence of contamination event at individual node in high risk zone is 3.33 %; in medium risk zone
is 1.36 %).

These probability values of contamination event at different nodes are useful in combining the
objective function values for different events while optimizing monitoring system design.

5. Summary and Conclusions

An integrated risk assessment of water distrbution system (IRA-WDS) developed by (Vairavamoorthy


et al. 2007) is suggested for identifying the risk prone area and selecting nodes at which
contamination events need to be useful when designing the monitoring system to detect accidental
contamination in water distribution network. The pipes/nodes of water distibution network are
classified as very high, high, medium and low risk pipes/nodes. A typical water starved network from
the field is considered to illustrate the methodology. It is observed that number of contamination
events need to be considered is reduced to 42 as against 145 if each node is considered as possible
location of contaminant entry. The proposed methodology can be applied to any type of network. The
probability of occurrences of accidental contamination is more in high risk prone zones and therefore
monitoring system should be designed to provide more emphasis to detect these events in desirable
time. The work on design of sensor-based optimal monitoring system is in progress and comparative
study would be presented in future.

References

Abdy Sayyed, M.A.H., Gupta, R., and Tanyimboh, T. (2015). Non-iterative application of EPANET for pressure
dependent modelling of water distribution systems. Water Resources Management, DOI 10.1007/s11269-015-
0992-0, Published online 16/4/2015.
Ang, W. K., and Jowitt, P. W. (2006). Solution for water distribution systems under pressure-deficient
conditions. J. Water Resources Planning and Management, 132(3), 175–182.
Berry J, Hart, W. E., Phillips, C. A., Uber, J. G., and Watson, J-P. (2006) Sensor placement in municipal water
networks with temporal integer programming models. J. Water Resources Planning and Management, 132(4),
218-224.
Berry, J., Carr, R. D., Hart, W. E., Leung, V. J., Phillips, C. A., and Watson, J-P. (2009). Designing
contamination warning systems for municipal water networks using imperfect sensors. J. Water Resources
Planning and Management, 135(4), 253–263.
Carr, R. D., Greenberg, H.J., Hart, W.E., Konjevod, G., Lauer, E., Lin, H., Morrison, T., and Phillips, C. A.
(2006). Robust optimization of contaminant sensor placement for community water systems. Math Program
Ser. B, 107, 337–356.
Chandapillai, J., Sudheer, K. P., and Saseedran, S. (2012). Design of water distribution network for equitable
supply. Water Resources Management, 26, 391-406.
Chastain, J. R. (2006). Methodology for locating monitoring stations to detect contamination in potable water
distribution systems. J. Infrastructure System, 12(4), 252–259.
Davis, M., Janke, R., and Phillips, C. A. (2014). Robustness of designs for drinking-water contamination
warning systems under uncertain conditions. J. Water Resources Planning and Management, 140(10), DOI,
04014028.
View publication stats

HYDRO 2015 INTERNATIONAL IIT Roorkee, India, 17-19 December, 2015


20th International Conference on Hydraulics,
Water Resources and River Engineering

Dorini, G., Jonkergouw, P., Kapelan, Z., and Savic, D. (2010). SLOTS, Effective algorithm for sensor
placement in water distribution systems. J. Water Resources Planning and Management,136(6), 620-628.
Eliades, D., Polycarpou, M. (2006). Iterative deepening of Pareto solutions in water sensor networks. Proc. 8th
Annual Water Distribution Systems Analysis Symp. ASCE, Reston, VA.
Krause, A., Leskovec, J., Guestrin, C., VanBriesen, J., and Faloutsos, C. (2008) Efficient sensor placement
optimization for securing large water distribution networks. J. Water Resources Planning and Management,
134(6), 516–526.
Krause, A., and Guestrin, C. (2009). Robust sensor placement for detecting adversarial contaminations in water
distribution systems. Proc. 8th Annual Water Distribution Systems Analysis Symp., ASCE, Reston, VA.
Ostfeld, A., and Salomons, E. (2004). Optimal layout of Early Warning Detection Stations for Water
Distribution Systems Security. J. Water Resources Planning and Management, 130(5), 377-385.
Ostfeld, A., et al. (2008). The battle of the water sensor networks (BWSN): A design challenge for engineers
and algorithms. J. Water Resources Planning and Management, 134(6), 555-568.
Ozger, S. S., and Mays, L. W. (2003). “A semi-pressure-driven approach to reliability assessment of water
distribution networks.” Proc. 30th IAHR Congress, Thessaloniki, Greece, 345–352.
Perelman, L., and Ostfeld, A. (2012). Extreme impact contamination events sampling for real-sized water
distribution systems. J. Water Resources Planning and Management, 138(5), 581-585.
Perelman, L., and Ostfeld, A. (2010). Extreme impact contamination events sampling for water distribution
systems security. J. Water Resources Planning and Management, 136(1), 80-86.
Preis, A., and Ostfeld, A. (2008). Multiobjective contaminant sensor network design for water distribution
systems. J. Water Resources Planning and Management, 134(4), 366-377.
Sargaonkar, A., Islam, R. (2009). Application of GIS in water distribution system assessment. J. Environmental
Science and Engineering, 51(4), 321-324.
Sargaonkar, A., Kamble, S., and Rao R. (2013). Model Study for Rehabilitation Planning of Water Supply
Network. Computers, Environment and Urban Systems, 39, 172-181.
Shen, H., and McBean, E. (2011). Pareto optimality for sensor placements in a water distribution system. J.
Water Resources Planning and Management, 137(3), 243-248.
Siew, C., Tanyimboh, T.T., and Seyoum, A. G. (2012). Pressure-dependent EPANET extension. Water
Resources Management, 26(6), 1477-1498.
Vairavamoorthy, K., Yan, J., Galgale, H. M., and Gorantiwar, S. D. (2007). IRA-WDS: A GIS-based risk
analysis tool for water distribution systems. Environmental Modelling & Software, 22, 951- 965.
Watson, J-P., Murray, R., Hart, W. E. (2009). Formulation and optimization of robust sensor placement
problems for drinking water contamination warning systems. J. Infrastructure System, 15(4), 330-339.
Weickgenannt, M., Kapelan, Z., Blokker, M., and Savic, D. A. (2010). Risk based sensor placement for
contaminant detection in water distribution systems. J. Water Resources Planning and Management, 136(6),
629-636.
Wu, Z. Y., and Walski, T. (2006). Multi objective optimization of sensor placement in water distribution
systems. Proc. 8th Annual Water Distribution Systems Analysis Symp., ASCE, Reston, VA.
Xu, J., Johnson, M. P., Fischbeck, P. S.,, Small, M. J., and VanBriesen, J. M. (2010). Robust placement of
sensors in dynamic water distribution systems. European Journal of operational research, 202, 707-716.

You might also like