Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Hydraulic Uniformity Index for Water Distribution Networks

Gimoon Jeong 1 and Doosun Kang 2

Abstract: Various reliability measures have been developed to evaluate design and operational aspects of water distribution system (WDS)
Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

performance. Conventional reliability indicators mainly focus on nodal supply conditions, such as nodal pressure and available discharge.
Here, a new link-based reliability index is proposed that considers pipe head loss–distribution throughout a network. The proposed index, the
hydraulic uniformity index (HUI), is approximated using the equivalent hydraulic gradient and intends to distribute the hydraulic gradient as
uniformly as possible throughout the network by enlarging steep-gradient pipes and reducing the size of low-gradient pipes, while satisfying
the design constraints. The HUI is intended to overcome the shortcomings of conventional node-based indices by synchronizing the evalu-
ation components with the design elements for intuitive system evaluation. An applied study using the Hanoi benchmark network demon-
strates that the proposed HUI can be used as an indicator to evaluate system design and operational reliability. DOI: 10.1061/(ASCE)
WR.1943-5452.0001158. © 2019 American Society of Civil Engineers.
Author keywords: Equivalent hydraulic gradient; Hydraulic uniformity index (HUI); System reliability; Water distribution system.

Introduction of pipe failure and repair time. Moreover, Yamijala et al. (2009) and
Fadaee and Tabatabaei (2010) proposed regression models of pipe
Water distribution systems (WDS) are critical infrastructure com- reliability based on pipe breakage and failure probability analysis.
ponents that supply drinking water over a wide area. WDS should Recently, Shuang et al. (2017) reviewed historical research on
have sufficient capacity to ensure stable water supply in terms of mechanical reliability of WDS and evaluated system reliability
system pressure and service flow. The system capacity often is in- during cascading pipe failures using a simulation-based method.
terpreted using the concept of “system reliability,” which quantifies In the context of connectivity/topological reliability, Yazdani
marginal serviceability over the minimum requirement. Wagner and Jeffrey (2010) and Candelieri et al. (2015) summarized struc-
et al. (1988) were the first to introduce and apply the concepts of tural metrics of WDNs and demonstrated the relation between con-
mechanical and hydraulic reliability to the design of WDS. The nectivity and reliability under network failures. Creaco et al. (2016)
mechanical reliability focuses on the network topology, to evaluate found that network performance is related to the uniformity of the
the system’s connectivity given the failure conditions, and often pipe diameters in the generic loop, and quantified its uniformity
ignores the system’s ability to effectively supply water to end users. as the loop–diameter uniformity coefficient. In addition, Di Nardo
Meanwhile, the hydraulic reliability refers to the ability of a system et al. (2018) suggested the reliability assessment based on fractal
to meet the requirements of water flow and pressure given normal and topological metrics. They investigated relations among geo-
and abnormal conditions (Mays 1989). Halhal et al. (1997) pro- metrical metrics, topological metrics, and energy resilience, and
posed three types of benefits, such as hydraulic, physical integrity, introduced the advantages of recognizing individual pipe having
and flexibility as the objectives for reliable WDS design. Later, greater impact-to-network resilience.
Ostfeld (2004) proposed to categorize WDS reliability into three Regarding entropic reliability, Awumah et al. (1990) formulated
properties, such as connectivity/topological, hydraulic, and entropic an entropy reliability index, which represents water supply diver-
contexts. sity in a network, and Tanyimboh and Templeman (1993) advanced
A number of studies have attempted to quantify WDS reliabil- the entropy index and applied the concept to several study net-
ity; several representative studies are reviewed in the following. In works. Afshar et al. (2005) suggested a reliability-based WDS
the context of mechanical (physical) reliability, Su et al. (1987) sug- design that simultaneously considers the network layout and pipe
gested a mechanical reliability indicator using a calculation of pipe size as the decision variables. Later, Di Nardo et al. (2010), Raad
availability based on pipe failure probability. Fujiwara and Tung et al. (2010), and Greco et al. (2012) explored the use of energetic
(1991) advanced the concept of pipe availability by considering and entropic indices to evaluate the effects of pipe closures on a
pipe size variation. Khomsi et al. (1996) further expanded the network’s hydraulic performance.
application of pipe availability by including the satisfied service Finally, in the context of hydraulic reliability, Xu and Goulter
time and demonstrated the method using a Monte-Carlo simulation (1999) assessed system reliability based on available nodal pres-
sure, and later Shi (2006) suggested a reliability index using actual
1
Ph.D. Candidate, Dept. of Civil Engineering, Kyung Hee Univ., 1732, supplied flows at demand nodes. The available pressure and sup-
Deogyeong-daero, Giheung-gu, Yongin-si, Gyeonggi-do 17104, Republic plied flows can be considered simultaneously as an energy term.
of Korea. Email: gimoon1118@gmail.com The resilience index (RI) developed by Todini (2000) was the first
2
Associate Professor, Dept. of Civil Engineering, Kyung Hee Univ., energy-based hydraulic reliability index. Prasad and Park (2003)
1732, Deogyeong-daero, Giheung-gu, Yongin-si, Gyeonggi-do 17104,
and Jayaram and Srinivasan (2008) further developed various
Republic of Korea (corresponding author). Email: doosunkang@khu.ac.kr
Note. This manuscript was submitted on December 5, 2018; approved
energy-based reliability indices, such as network resilience index
on June 27, 2019; published online on December 12, 2019. Discussion per- (NRI) and modified resilience index (MRI), respectively, and dem-
iod open until May 12, 2020; separate discussions must be submitted for onstrated them using benchmark networks. Cabrera et al. (2010)
individual papers. This paper is part of the Journal of Water Resources proposed five energy-efficiency indicators for the evaluation of
Planning and Management, © ASCE, ISSN 0733-9496. WDS performance, and later Cabrera et al. (2014) suggested

© ASCE 04019078-1 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


a new approach for assessing energy components in WDS.
Similarly, Dziedzic and Karney (2013) analyzed water distribution
sustainability using four energy components (supplied, delivered,
lost, and dissipated) in water networks. Moreover, Hashemi et al.
(2015) proposed five energy metrics based on six types of energy
components (supplied, delivered, required, downstream flow, leak-
age, and friction) in WDS. Recently, Jeong et al. (2017) suggested a
new approach to estimate the minimum required head at a node,
which considers the flow direction in pipes and the hydraulic gra-
dient within a network. Applying their approach, they found that Fig. 1. Energy distribution in a WDS. (Reprinted from Jeong et al.
conventional resilience indices, such as RI, NRI, and MRI, are
Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

2017, © ASCE.)
likely to overestimate the system reliability.
The aforementioned studies developed hydraulic reliability in-
dices based on nodal information, such as available pressure and
supplied demand. However, the effects of the head loss and flow Einput ¼ Eloss þ Eoutput ð1Þ
rate at links on the nodal pressure and flow depend on pipe char-
acteristics, such as diameter, length, material, and layout. Thus, to In this study, a link-based reliability index is developed based on
evaluate the nodal index, an investigation of pipe design should be the head loss distribution of a network. In detail, the energy loss in a
performed beforehand. For more intuitive and effective evaluation WDS can be categorized into three components: friction loss, leak-
of system reliability, it would be beneficial to develop an index age loss, and minor loss. The friction loss (Efriction ) is caused by
using the pipe characteristics. Conventionally, node-based indices friction occurring at the pipe wall; the leakage loss (Eleak ) occurs
are based on available nodal pressure and actual water supply. through leaks in the pipes. The minor loss (Eminor ) takes place at
Therefore, the water supply satisfaction is determined based on valves, joints, and other mechanical components in a system. Since
whether the available nodal pressure is greater than the minimum Eminor is generally negligible, the energy loss (Eloss ) can be ap-
required pressure head at the nodes. Meanwhile, there exists no proximated as the sum of Efriction and Eleak , as expressed in
specific standard for the pipes to determine system reliability Eq. (2a). Efriction and Eleak are further expressed in Eqs. (2b)
except for a flow velocity limit. Colombo and Karney (2009) sug- and (2c), respectively
gested equating water and energy leakage at pipes to determine the
appropriate pipe size, but did not go so far as to suggest a specific Eloss ¼ Efriction þEleak ð2aÞ
boundary for design. Kang and Lansey (2011) used the pipe flow
velocity as a criterion for determining the initial pipe diameter for X X
T npipe

the optimal design of a WDS. In addition, Suribabu et al. (2016) Efriction ¼ γ Qti Htloss;i Δt ð2bÞ
t¼1 i¼1
proposed an approach to improve the system reliability by enlarg-
ing pipes with excessively high flow velocity. However, no gener-
X X
T npipe
ally accepted standard or index was suggested for link-based Eleak ¼ γ Qtleak;i H ti Δt ð2cÞ
system design and evaluation. t¼1 i¼1
In this study, a link-based reliability index, the so-called hy-
draulic uniformity index (HUI), was proposed. First, we introduce where T = operation time; Δt = time step; npipe = number of
the concepts of equivalent head loss (hlequi ) and equivalent hy- pipes in a system; Qti = flow rate in the ith pipe at the tth time
draulic gradient (HGequi ), which are suggested as standard values step; Qtleak;i = leakage rate in the ith pipe at the tth time step;
for pipe evaluation. The HUI is then calculated at each pipe by Htloss;i = friction head loss in the ith pipe at the tth time step;
comparing the equivalent hydraulic gradient with the actual and H ti = total head in the ith pipe at the tth time step, which is
hydraulic gradient. Finally, a systemwide HUI value (HUI sys ) is generally calculated as an average of total heads at the upstream
defined to evaluate overall system design. A well-known bench- and downstream nodes of the relevant pipe.
mark network (Hanoi network) was used to demonstrate the pro- The friction loss (Efriction ) decreases the supply pressure at
posed index. Applied studies revealed that a hydraulically uniform demand nodes, while the leakage loss (Eleak ) is associated with
network is more reliable under abnormal conditions, and the pro- the loss of delivered flow and pressure. Since our proposed
posed HUI can successfully serve as a reliability indicator to quan- concepts (i.e., the equivalent head loss and equivalent hydraulic
tify a system’s hydraulic uniformity. gradient) are associated only with the head loss in pipelines,
Eleak is not considered in the current study; therefore, Eloss only
refers to Efriction hereafter.
Methodology
Concept of Acceptable Head Loss
Energy Loss in a Water Distribution Network
The conventional reliability indices were defined based on nodal
Energy distribution in a WDS is classified into three categories: supply conditions, such as nodal pressure and serviceable outflow,
input, loss, and output energy, as shown in Fig. 1. Input energy while the decisions for system design are the link-based informa-
(Einput ) is the energy supplied to the system from the source tion, such as diameter, material, and layout. Evaluating the nodal
via pumps; loss energy (Eloss ) is the energy lost during the trans- supply conditions does not clearly identify the exact locations
mission process; and output energy (Eoutput ) is the energy delivered where inappropriate system components are used. Thus, the con-
to and utilized by the users at demand nodes. The input energy ventional reliability indices (i.e., the node-based indices) only focus
must be sufficient to supply the output energy after accounting on system outcomes, but not on the link-based design aspects. To
for energy loss; thus, energy conservation is satisfied, as expressed overcome this shortcoming, system reliability should be evaluated
in Eq. (1). based on the link information as well, synchronizing the design

© ASCE 04019078-2 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


Fig. 2. Concept of hydraulic gradient.
Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

components and evaluation components, while satisfying the nodal four designs. As seen in Fig. 3, the service pressure is highest in
conditions. The system reliability is interpreted as the marginal Design 1, with the most uniform hydraulic gradient, whereas the
serviceability over the minimum requirement. In water pipe net- nodal pressures are relatively low in Designs 3 and 4, which have
works, a system performance evaluation is conducted to deter- inconsistent hydraulic gradients. The results reveal that the system
mine whether the energy loss (Eloss ) is acceptable to secure reliable reliability (i.e., service pressure) is proportional to the uniformity of
water supply. In this study, a concept of acceptable head loss is the hydraulic gradient, given identical design constraints.
suggested based on equivalent head loss (hlequi ) and equivalent Further investigation of the correlation between system reliabil-
hydraulic gradient (HGequi ), which are derived based on total en- ity and gradient uniformity was pursued. That is, given the four
ergy loss and pipe characteristics (i.e., flow rate, diameter, and alternative designs in Fig. 3, the nodal demands were increased
length). Here, the HGequi can be understood as hlequi per unit pipe by 10%, simulating abnormal conditions. Table 1 summarizes the
length. pressure drop of each network for the abnormal demand increase
condition. As seen, the overall pressure reduction is inversely pro-
Definition of Hydraulic Gradient portional to the hydraulic uniformity, such that Design 1, with the
The hydraulic gradient (HG) of a pipe is defined as the head loss most uniform hydraulic gradient, shows the lowest pressure drop
per unit length of a pipe. The head loss through a pipe is the head for the increased demand scenario. Meanwhile, Designs 3 and 4
difference between the upstream and downstream nodes in the di- suffer greater pressure drops than the other designs.
rection of flow, and it varies depending on pipe size and conveying For simple branch networks, the hydraulic gradients can be
flow rate (Fig. 2). Particularly steep or mild slope implies that the estimated easily, and it is possible to evaluate individual pipes. For
relevant pipe is either underdesigned or overdesigned, respectively, large and complex, looped networks, however, a quantification
for a given flow rate. A network with consistent hydraulic gradient metric based on a formulated approach would be beneficial to
indicates that the individual pipes are designed to have consistent quantitatively evaluate the system performance.
slope throughout the network. Common design practices are likely
to distribute the hydraulic gradient as uniformly as possible over the
network by enlarging the steep-gradient pipes and reducing the size Calculation of Hydraulic Uniformity Index
of the mild-gradient pipes.
Equivalent Head Loss
Hydraulic Gradient and System Reliability Using the loss energy equation, the equivalent head loss of a system
Each of the four alternative designs in Fig. 3 is comprised of three can be estimated by assuming that each pipe in the system has
pipes with diameters of 200, 250, and 300 mm for different layouts. identical head loss, as seen in Eq. (3a), and is finally expressed
Note that the pipe lengths and nodal demands are identical for all in Eq. (3b).

Fig. 3. Hydraulic gradient and system pressure of alternative designs: (a) Design 1; (b) Design 2; (c) Design 3; and (d) Design 4.

© ASCE 04019078-3 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


Table 1. Nodal pressure drop (m) caused by 10% increase in nodal water hlequi × npipe
demand HGequi ¼ Pnpipe ð4Þ
i¼1 Li
Node Design 1 Design 2 Design 3 Design 4
B 0.40 0.40 2.87 2.87 where HGequi = the equivalent hydraulic gradient of a system; and
C 0.86 1.75 3.06 3.33 Li = the length of the ith pipe.
D 1.23 1.88 3.19 3.38 Finally, by substituting the equivalent head loss (hlequi ) of
Average 0.83 1.34 3.04 3.19 Eq. (4) with Eq. (3b), HGequi can be derived as Eq. (5)
P
npipe × npipe
i¼1 Qi hli
HGequi ¼ Pnpipe P npipe ð5Þ
i¼1 Li i¼1 Qi
X
npipe X
npipe
Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

γ Qi hli ¼ γhlequi Qi ð3aÞ


i¼1 i¼1
Hydraulic Uniformity Index
The hydraulic uniformity index (HUI) of an individual pipe is
calculated as the ratio of the hydraulic gradient of a pipe to the
Pnpipe
i¼1 Qi hli
equivalent hydraulic gradient and is formulated in Eq. (6)
hlequi ¼ P npipe ð3bÞ
i¼1 Qi HGi hli
HUI i ¼ ¼ ð6Þ
HGequi HGequi × Li
where hlequi = the equivalent head loss of a system; and hli = the
head loss in the ith pipe. where HUI i = the hydraulic uniformity index of the ith pipe; and
The proposed equivalent head loss (hlequi ) does not contain pipe HGi = hydraulic gradient of the ith pipe.
length information; thus, it implies identical head loss for individ- An HUI value greater than 1.0 indicates that the pipe may be
ual pipes. Although the head loss is identical, the hydraulic gradient underdesigned, degrading the system performance by inducing ex-
can change, depending on the pipe length, as seen in Fig. 4(b). cessive head loss. On the contrary, a pipe with an HUI value lower
Therefore, in order to estimate an equivalent value over the entire than 1.0 can be considered to be overdesigned. Therefore, the ideal
network, it is necessary to calculate the equivalent hydraulic HUI value of an individual pipe is close to 1.0.
gradient (HGequi ), considering the pipe length. Furthermore, a systemwide HUI (HUI sys ) can be quantified as
shown in Eq. (7), which is the standard deviation of individual
Equivalent Hydraulic Gradient pipes’ HUI from the ideal value of 1.0. A HUI sys value of zero
The equivalent hydraulic gradient (HGequi ) is calculated as the total indicates that the hydraulic gradients of individual pipes are
head loss over a network divided by the total pipe length and is perfectly uniform. Therefore, the system design with the lowest
expressed as Eq. (4): HUI sys value (closest to zero) would be the most reliable solution

Fig. 4. Concept of equivalent head loss: (a) equivalent head loss with equivalent hydraulic gradient; and (b) equivalent head loss with varying
hydraulic gradient.

© ASCE 04019078-4 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Comparisons of alternative designs with different hydraulic gradients.

when satisfying the design conditions, such as minimum pressure


requirements and minimum design cost
sPffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
npipe 2
i¼1 ðHUI i 1Þ
HUI sys ¼ ð7Þ
npipe

where HUI sys = the systemwide hydraulic uniformity index.


It should be noted that HUI sys only constrains the uniformity of
the hydraulic gradient of an entire network. If the hydraulic gra-
dients of individual pipes are consistent throughout the network,
the HUI sys is equal to zero even for different designs [e.g., Designs
(A), (B), and (C)], as shown in Fig. 5. Here, Design (A) is oversized
with redundant capacity representing a reliable but unfavorable de-
sign, while Design (C) is infeasible and unfavorable since it cannot
meet the minimum pressure requirement. Meanwhile, Designs (B),
(B’), and (B’’) are all feasible by satisfying the minimum pres-
Fig. 6. Hanoi network with node and pipe IDs.
sure requirement but have different HUI sys values. Among them,
Design (B) is the most hydraulically uniform and reliable design
with the lowest HUI sys . As seen, the HUI sys can be used to evalu-
ate the consistency of the hydraulic gradient throughout a network.
Table 2. Available pipe diameters with unit cost
To further evaluate individual pipes in a network, the pipe HUI
value should be reviewed to enhance the system design. Pipe diameter
(inch) (mm) Unit cost (USD/m)
a
8 203 22.7
Applications 12 305 45.7
16 406 70.4
Applied Designs 20 508 98.4
24 610 129.3
The proposed indices were applied to the Hanoi network for dem- 30 762 180.7
onstration (Fig. 6). The Hanoi network has been used widely as a 40 1,016 278.3
benchmark network for optimal design practices. It is a looped net- 48a 1,219 368.9
work supplying 19,940 m3 =h of demand to 31 nodes through 34 a
Marked diameters are newly added for network modification.
pipes (with no pumps or valves). The network has a single source
with a total head of 100 m, and the elevation of all demand nodes is
equal to 0 m. The Hanoi network has been optimized in many pre-
vious studies, and Cunha and Sousa (1999) suggested the lowest Base Design (Cost: USD 6,056,101)
cost design. Here, the least-cost optimal design suggested by Cunha Fig. 7(a) shows pipe diameters of the least expensive design of
and Sousa (1999) was used as a base design for comparison. Table 2 Cunha and Sousa (1999). The HUI value of each pipe was calcu-
summarizes the available pipe sizes and unit costs applied for lated for the base design and displayed in Fig. 7(b). The thick
Hanoi network optimization. In Cunha and Sousa (1999), the avail- lines in Fig. 7(b) indicate underdesigned pipes (P1, P2, P18,
able pipe diameters ranged from 305 to 1,016 mm (12 to 40 in.). and P19) with high HUI values (≥1.5), while the dotted lines
In this study, 203- and 1,219-mm (8- and 48-in.) pipes were also (P12, P16, P28, and P31) indicate overdesigned pipes with low
included for system modification. HUI values (≤0.1). Note the design cost of the base design is

© ASCE 04019078-5 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Base design (Cost = USD 6,056,101, HUIsys ¼ 0.8391): (a) pipe diameter (millimeters); and (b) pipe HUI.

Fig. 8. Modified design 1 (Cost = USD 6,136,906, HUIsys ¼ 0.8359): (a) pipe diameter (millimeters); and (b) pipe HUI.

USD 6,056,101 with a system HUI (HUI sys ) value of 0.8391. The since they already were the minimum diameter of 305 mm (12 in.).
lowest and system-averaged pressures are 29.7 m (N27) and Other pipes with an unacceptable HUI value (either much higher or
41.0 m, respectively. lower than 1.0) were modified manually until no further improve-
ment could be obtained. Note that the minimum pressure constraint
Modified Design 1 (Cost: USD 6,136,906) was applied when modifying the pipe diameter; that is, the mini-
The base design was manually modified to improve the HUI values mum nodal pressure should be greater than 29.7 m. In Modified
of individual pipes by enlarging the underdesigned pipes and design 1, the lowest and system-average pressures are 29.8 m
reducing the sizes of overdesigned pipes. Note that optimization and 42.6 m, respectively. It is still observed that P1 and P2 are
approaches may apply to improve system design, especially for underdesigned (thick lines), requiring enlargement, while P12,
large and complex, looped networks. In the modifications, the P31, and P33 are overdesigned (dotted lines), and need to be
available pipe sizes were set to be the same as those used in the decreased further. Overall, marginal improvement of system perfor-
base design (i.e., 305–1,016 mm, Table 2). Fig. 8 shows the modi- mance was obtained with a slight increase of design cost in
fied pipe diameters and HUI values of Modified design 1. The final Modified design 1 compared to the base design.
design cost was USD 6,136,906 (1.3% increase compared to the
base design), and the HUI sys value was marginally improved to Modified Design 2 (Cost: USD 5,981,392)
0.8359. Note the lower HUI sys indicates the more consistent To further improve Modified design 1, the available pipe sizes were
system layout. extended by including 203- and 1,219-mm (8- and 48-in.) pipes, as
The specific modifications are as follows. The underdesigned listed in Table 2. Fig. 9 shows the pipe sizes and HUI values for
pipes P1 and P2 could not be enlarged further since they already Modified design 2. The design cost was reduced to USD 5,981,392
were the maximum available size of 1,016 mm (40 in.). Similarly, (1.2% decrease compared to the base design), while the HUI sys
the overdesigned pipes P16, P28, and P31 were not further reduced value was improved to 0.4746, showing remarkable improvement

© ASCE 04019078-6 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Modified design 2 (Cost = USD 5,981,392, HUIsys ¼ 0.4746): (a) pipe diameter (millimeters); and (b) pipe HUI.

Fig. 10. Pipe HUI comparison of three designs: (a) HUI values for pipes in three designs; and (b) pipe HUI distribution of three designs.

in hydraulic uniformity. That is, the system performance was im- of uniformity of pipe diameters in WDN loops. These indicators
proved, while decreasing the design cost. are well known for network reliability quantification and reflecting
It should be noted that the design cost could be decreased below uniformity in concept of network composition, respectively, and
that of the least-cost design (the base design) because the available thus were chosen for comparison with HUI.
pipe sizes were extended in Modified design 2. Given the design Table 3 summarizes the comparison results of three indices for
condition, the system HUI was significantly improved (from the applied designs. Here, the lower system HUI value, the higher
0.8391 to 0.4746) and the lowest and system-average pressures RI value, and the higher LDU value indicate the more reliable
were calculated as 38.3 and 63.5 m, respectively, showing substan- system design. As seen, both the HUI and RI indicate a slight en-
tial improvement in system reliability. In addition, as seen in hancement in Modified design 1 and a remarkable improvement in
Fig. 9(b), pipe HUI values are close to 1.0 (except P1, P2, and Modified design 2 compared to the base design. Note that the RI
P31), indicating hydraulically uniform network design. value of Modified design 2 is remarkably increased from the value
Fig. 10 compares the pipe HUI values of the three designs. As of the base design (i.e., from 0.2111 to 0.5084), which reveals that
shown, Modified design 2 is the most uniform design in terms of the network modification based on the pipe HUI clearly enhances
hydraulic gradient, with HUI values of most pipes close to 1.0, overall network resilience/reliability. On the other hand, LDU val-
whereas the base design and Modified design 1 contain several ues of the three designs are almost consistent, which indicates that
pipes with unfavorable HUI values. the system modifications based on the pipe HUI does not alter the
diameter uniformity within a loop. This proves that the proposed
Comparison with Conventional Reliability Indices modification approach enhances the system reliability, while main-
taining the loop-based diameter uniformity in a topological aspect.
The proposed HUI was compared with conventional reliability in-
dices using the three designs obtained in the previous section. The
Abnormal Operation Scenarios
resilience index (RI) (Todini 2000) is the representative reliability
indicator based on energy analysis of WDNs, and loop diameter In the previous section, it was found that Modified design 2 is
uniformity (LDU) (Creaco et al. 2016) is the topological index the most hydraulically uniform network, followed by Modified

© ASCE 04019078-7 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


Table 3. Comparison of system reliability indices for three designs
Index Base design Modified design 1 Modified design 2
HUI sys 0.8391 0.8359 0.4746
Resilience index (RI) (Todini 2000) 0.1847 0.2111 0.5084
Loop diameter uniformity (LDU) (Creaco et al. 2016) 0.5108 0.5237 0.5204

Table 4. Comparison of three designs under base and abnormal scenarios


Measure Base design Modified design 1 Modified design 2
Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

HUI sys 0.8391 0.8359 0.4746


Design cost 6,056,101 6,136,906 5,981,392
Base condition
H min (m) 29.7 29.8 38.3
H average (m) 41.0 42.6 63.5
Scenario 1
H min (m) 18.9 19.0 28.9
H average (m) 32.0 33.8 57.9
No. of unsatisfactory nodes 18 14 1
Scenario 2
H min (m) 22.2 26.1 31.7
H average (m) 40.4 42.1 62.4
No. of unsatisfactory nodes 5 4 1

design 1 and the base design. To investigate the relationship pipes (P14, P15, P16, P27, P28, and P31) were selected to simulate
between HUI and system reliability, the three designs were tested the single-pipe failure conditions. Note that multiple-pipe failures
with two abnormal operation scenarios, including an increase in were not considered in this study. Here, the six pipes were selected
water demand and unexpected pipe failure. since they do not induce a negative pressure when failed. The six-
pipe failure conditions are simulated one by one and minimum no-
Scenario 1: Water Demand Increase dal pressure, average pressure, and number of unsatisfactory nodes
The nodal demands were increased by 8%, assuming peak water are calculated. The average values of six simulation results are sum-
usage during the summer, and the system pressure of the three de- marized in Table 4. Comparing the base design, Modified design 1,
signs were compared. Note that the scenario of an 8% increase in and Modified design 2, the minimum nodal pressures are 22.2,
water consumption during the summer season due to climate 26.1, and 31.7 m, respectively, and the system-average pressures
change was adopted from an Intergovernmental Panel on Climate are 40.4, 42.1, and 62.4 m, respectively. In addition, the numbers
Change (IPCC) technical paper (Bates et al. 2008). Table 4 sum- of nodes with unsatisfactory pressure head are 5, 4, and 1, respec-
marizes the system performance of the three designs for normal and tively. Overall, the scenario analysis reveals that the Modified de-
abnormal demand conditions. sign 2 is the most reliable and robust design as affected by the pipe
Comparing the base design, Modified design 1, and Modified failure events, followed by Modified design 1 and the base design.
design 2, due to the demand increase, the maximum nodal pressure The lower system HUI represents the more hydraulically uni-
drops are 10.8, 10.8, and 9.4 m, respectively; and the system- form network that avoids over- and underdesigned components,
average pressure reductions are 9.0, 8.8, and 5.6 m, respectively. leading to a more reliable system design with a similar design cost.
In addition, the number of nodes with unsatisfactory pressure head Comparing the applied designs, it was clearly observed that a net-
can be used to compare the system reliability of each design. As work with lower system HUI (with similar design cost) provides
shown, the system serviceability is improved in the modified de- sufficient pressure more reliably under abnormal operating condi-
signs compared to the base design since the number of unsatisfac- tions. The applied study also demonstrates that the suggested HUIs
tory nodes is reduced from 18 locations for the base design to 14 can be used as indicators for evaluating individual pipe and overall
and 1 locations for Modified designs 1 and 2, respectively. system reliability.
Overall, the base design and Modified design 1 show similar
results in terms of design cost and system HUI. Comparing the
system performance (reliability), the two designs provide very sim- Conclusions
ilar system pressures, though Modified design 1 is marginally im-
proved, with a slightly higher design cost. Comparing the base A link-based reliability index was developed based on the head
design and Modified design 2, Modified design 2 outperforms loss distribution of a network. The proposed index, called a hy-
the base design in every aspect, including cost, HUIs, and system draulic uniformity index (HUI), is intended to overcome the short-
pressure for normal and abnormal demand conditions. Note that the comings of conventional node-based reliability indices that are
system HUI of Modified design 2 is clearly decreased compared dependent mainly on nodal supply conditions, such as nodal pres-
with that of the base design and Modified design 1. sure and serviceable discharge. The newly proposed, link-based
index synchronizes the evaluation components with the design
Scenario 2: Single Pipe Failures elements. Thus, it can suggest the optimal system design using a
The second abnormal operation scenario demonstrates the effects more intuitive approach. The proposed HUI intends to distribute
of single-pipe failures to the system reliability. Six representative the hydraulic gradient as uniform as possible throughout the

© ASCE 04019078-8 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


network by enlarging the steep-gradient pipes and reducing the size Eoutput = total output energy;
of the mild-gradient pipes, while satisfying the nodal pressure Efriction = lost energy due to pipe wall friction;
requirements. The HUI of an individual pipe is calculated as the Eleak = lost energy by leakage;
ratio of the hydraulic gradient of a pipe to the equivalent hydraulic Hti = total head in ith pipe at tth time step;
gradient; therefore, the ideal value is 1.0. Furthermore, a system-
HGequi = equivalent hydraulic gradient of system;
wide HUI can be quantified as the standard deviation of individual
HGi = actual head loss in ith pipe;
pipes’ HUI from the ideal value of 1.0; thus, a system HUI of zero
indicates that the hydraulic gradients of individual pipes are HUI i = hydraulic uniformity index in ith pipe;
perfectly uniform, providing the most reliable solution for given HUI sys = hydraulic uniformity index of system;
design constraints, such as minimum pressure requirements and hlequi = equivalent head loss of system;
limited design cost. hli = head loss in ith pipe;
Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

The proposed indices were applied to the Hanoi benchmark Li = length of ith pipe;
network for demonstration. The least expensive optimal design T= the operation time;
suggested by Cunha and Sousa (1999) was used as a base design npipe = total number of pipes;
for comparison. The base design was then modified manually to Qti = discharge in ith pipe at tth time step;
improve HUI values (both pipes’ HUI and systemwide HUI) and Qtleak;i = leakage rate in ith pipe at tth time step;
two modified designs were suggested. Through the system modi- γ= specific weight of water; and
fications (enlarging the underdesigned pipes and reducing the sizes
Δt = simulation time step.
of overdesigned pipes), the pipe HUI values were improved to be
close to the ideal value of 1.0, and the system HUI value was
reduced compared to the base design. The lower system HUI is
indicative of the more hydraulically uniform network, which avoids References
over- and underdesigned components, leading to a more reliable Afshar, M. H., M. Akbari, and M. A. Mariño. 2005. “Simultaneous layout
design with similar design cost. Comparing the system perfor- and size optimization of water distribution networks: Engineering ap-
mance of the applied designs under normal and abnormal demand proach.” J. Infrastruct. Syst. 11 (4): 221–230. https://doi.org/10.1061
conditions, it was demonstrated that a network with lower system /(ASCE)1076-0342(2005)11:4(221).
HUI (assuming similar design cost) can provide sufficient pressure Awumah, K., I. Goulter, and S. K. Bhatt. 1990. “Assessment of reliability in
at demand nodes more reliably. The applied study also shows that water distribution networks using entropy based measures.” Stochastic
the suggested HUI can be used as an indicator for evaluating indi- Hydrol. Hydraul. 4 (4): 309–320. https://doi.org/10.1007/BF01544084.
vidual pipe and overall system reliability. Bates, B. C., M. Kundzewicz, S. Wu, and J. P. Palutikof. 2008. Climate
The proposed system HUI can be used to evaluate the change and water, IPCC technical paper IV. Geneva: Intergovernmen-
consistency of the hydraulic gradient throughout the network tal Panel on Climate Change.
and possibly can be implemented as an objective or constraint Cabrera, E., E. Gómez, E. Cabrera, Jr., J. Soriano, and V. Espert. 2014.
quantifying system reliability for system design optimization. To “Energy assessment of pressurized water systems.” J. Water Resour.
Plann. Manage. 141 (8): 04014095. https://doi.org/10.1061/(ASCE)
further evaluate individual pipes in a network, the pipe HUI value
WR.1943-5452.0000494.
can be used to enhance the system design. The proposed HUI can
Cabrera, E., M. A. Pardo, R. Cobacho, and E. Cabrera Jr. 2010. “Energy
be improved by reflecting the individual pipe importance (based on audit of water networks.” J. Water Resour. Plann. Manage. 136 (6):
pipe diameter, length, and flow rate) in calculation to avoid bias and 669–677. https://doi.org/10.1061/(ASCE)WR.1943-5452.0000077.
precisely evaluate the system reliability. Implementing the pro- Candelieri, A., D. Soldi, and F. Archetti. 2015. “Network analysis for
posed indices in network design is a fruitful area to pursue in future resilience evaluation in water distribution networks.” Environ. Eng.
studies. Manage. J. 14 (6): 1261–1270. https://doi.org/10.30638/eemj.2015.136.
Colombo, A. F., and B. W. Karney. 2009. “Leaks and water use represen-
tation in water distribution system models: Finding a working equiva-
Data Availability Statement lence.” J. Hydraul. Eng. 135 (3): 234–239. https://doi.org/10.1061
/(ASCE)0733-9429(2009)135:3(234).
Some or all data, models, or code generated or used during the Creaco, E., M. Franchini, and E. Todini. 2016. “The combined use of resil-
study are available from the corresponding author by request. ience and loop diameter uniformity as a good indirect measure of net-
work reliability.” Urban Water J. 13 (2): 167–181. https://doi.org/10
.1080/1573062X.2014.949799.
Acknowledgments Cunha, M. D. C., and J. Sousa. 1999. “Water distribution network design
optimization: Simulated annealing approach.” J. Water Resour. Plann.
This work is supported by the Basic Science Research Program Manage. 125 (4): 215–221. https://doi.org/10.1061/(ASCE)0733-9496
through the National Research Foundation of Korea (NRF) funded (1999)125:4(215).
by the Ministry of Education, Science and Technology (NRF- Di Nardo, A., M. Di Natale, C. Giudicianni, R. Greco, and G. F.
Santonastaso. 2018. “Complex network and fractal theory for the as-
2016R1A2B4014273); and the EDISON (Education-research
sessment of water distribution network resilience to pipe failures.”
Integration through Simulation On the Net) Program through the Water Sci. Technol. Water Supply 18 (3): 767–777. https://doi.org/10
National Research Foundation of Korea (NRF) funded by the Min- .2166/ws.2017.124.
istry of Science, ICT, and Future Planning (2017M3C 1A6075016). Di Nardo, A., R. Greco, G. F. Santonastaso, and M. Di Natale. 2010.
“Resilience and entropy indices for water supply network sectorization
in district meter areas.” In Proc., Int. Conf. on Hydroinformatics,
Notation 2365–2373. Tianjin, China: Tianjin Univ.
Dziedzic, R. M., and B. W. Karney. 2013. “Energy metrics for water dis-
The following symbols are used in this paper: tribution assessment.” In Proc., 46th Annual Stormwater and Urban
Einput = total input energy; Water Systems Modeling Conf. on Journal of Water Management Mod-
Eloss = total loss energy; eling. Guelph, ON, Canada: Computational Hydraulics International.

© ASCE 04019078-9 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078


Fadaee, M., and R. Tabatabaei. 2010. “Estimation of failure probability in Raad, D. N., A. N. Sinske, and J. H. Van Vuuren. 2010. “Comparison of
water pipes network using statistical model.” World Appl. Sci. J. 11 (9): four reliability surrogate measures for water distribution systems
1157–1163. design.” Water Resour. Res. 46 (5): W05524. https://doi.org/10.1029
Fujiwara, O., and H. D. Tung. 1991. “Reliability improvement for water /2009WR007785.
distribution networks through increasing pipe size.” Water Resour. Res. Shi, P. 2006. “Seismic response modeling of water supply systems.” Ph.D.
27 (7): 1395–1402. https://doi.org/10.1029/91WR00882. dissertation, School of Civil and Environmental Engineering, Cornell
Greco, R., A. Di Nardo, and G. Santonastaso. 2012. “Resilience and en- Univ.
tropy as indices of robustness of water distribution networks.” J. Hydro- Shuang, Q., Y. Liu, Y. Tang, J. Liu, and K. Shuang. 2017. “System reli-
inf. 14 (3): 761–771. https://doi.org/10.2166/hydro.2012.037. ability evaluation in water distribution networks with the impact of
Halhal, D., G. A. Walters, D. Ouazar, and D. A. Savic. 1997. “Water net- valves experiencing cascading failures.” Water 9 (6): 413. https://doi
work rehabilitation with structured messy genetic algorithm.” J. Water .org/10.3390/w9060413.
Resour. Plann. Manage. 123 (3): 137–146. https://doi.org/10.1061 Su, Y.-C., L. W. Mays, N. Duan, and K. E. Lansey. 1987. “Reliability-
Downloaded from ascelibrary.org by "University of California, Santa Barbara" on 12/14/19. Copyright ASCE. For personal use only; all rights reserved.

/(ASCE)0733-9496(1997)123:3(137). based optimization model for water distribution systems.” J. Hydraul.


Hashemi, S., Y. R. Filion, and V. L. Speight. 2015. “Pipe-level energy Eng. 114 (12): 1539–1556. https://doi.org/10.1061/(ASCE)0733-9429
metrics for energy assessment in water distribution networks.” Procedia (1987)113:12(1539).
Eng. 119 (17): 139–147. https://doi.org/10.1016/j.proeng.2015.08.864. Suribabu, C. R., K. Prashanth, S. V. Kumar, and N. S. Ganesh. 2016.
Jayaram, N., and K. Srinivasan. 2008. “Performance-based optimal “Resilience enhancement methods for water distribution networks.”
design and rehabilitation of water distribution networks using life Jordan J. Civ. Eng. 10 (2): 219–231. https://doi.org/10.14525/JJCE
cycle costing.” Water Resour. Res. 44 (1): W01417. https://doi.org/10 .10.1.3538.
.1029/2006WR005316. Tanyimboh, T. T., and A. B. Templeman. 1993. “Calculating maximum
Jeong, G., A. Wicaksono, and D. Kang. 2017. “Revisiting the resilience entropy flows in networks.” J. Oper. Res. Soc. 44 (4): 383–396.
index for water distribution networks.” J. Water Resour. Plann. Man- https://doi.org/10.1057/jors.1993.68.
age. 143 (8): 04017035. https://doi.org/10.1061/(ASCE)WR.1943 Todini, E. 2000. “Looped water distribution networks design using a resil-
-5452.0000792. ience index based heuristic approach.” Urban Water 2 (2): 115–122.
Kang, D., and K. Lansey. 2011. “Revisiting optimal water-distribution sys- https://doi.org/10.1016/S1462-0758(00)00049-2.
tem design: Issues and a heuristic hierarchical approach.” J. Water Wagner, J. M., U. Shamir, and D. H. Marks. 1988. “Water distribution reli-
Resour. Plann. Manage. 138 (3): 208–217. https://doi.org/10.1061 ability: Simulation methods.” J. Water Resour. Plann. Manage. 114 (3):
/(ASCE)WR.1943-5452.0000165. 276–294. https://doi.org/10.1061/(ASCE)0733-9496(1988)114:3(276).
Khomsi, D., G. A. Walters, A. R. D. Thorley, and D. Ouazar. 1996. “Reliabil- Xu, C., and C. Goulter. 1999. “Reliability-based optimal design of water
ity tester for water-distribution networks.” J. Comput. Civ. Eng. 10 (1): distribution networks.” J. Water Resour. Plann. Manage. 125 (6):
10–19. https://doi.org/10.1061/(ASCE)0887-3801(1996)10:1(10). 352–362. https://doi.org/10.1061/(ASCE)0733-9496(1999)125:6(352).
Mays, L. W. 1989. Reliability analysis of water distribution systems. Yamijala, S., S. D. Guikema, and K. Brumbelow. 2009. “Statistical models
Reston, VA: ASCE. for the analysis of water distribution system pipe break data.” Reliab.
Ostfeld, A. 2004. “Reliability analysis of water distribution systems.” Eng. Syst. Saf. 94 (2): 282–293. https://doi.org/10.1016/j.ress.2008
J. Hydroinf. 6 (4): 281–294. https://doi.org/10.2166/hydro.2004.0021. .03.011.
Prasad, T. D., and N. S. Park. 2003. “Multiobjective genetic algorithms Yazdani, A., and P. Jeffrey. 2010. “Robustness and vulnerability analysis of
for design of water distribution networks.” J. Water Resour. Plann. water distribution networks using graph theoretic and complex network
Manage. 1 (73): 73–82. https://doi.org/10.1061/(ASCE)0733-9496 principles.” In Proc., 12th Annaul Conf. on Water Distribution Systems
(2004)130. Analysis (WDSA), 933–945. Reston, VA: ASCE.

© ASCE 04019078-10 J. Water Resour. Plann. Manage.

J. Water Resour. Plann. Manage., 2020, 146(2): 04019078

You might also like