You are on page 1of 11

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

net/publication/309447899

Analysis of Water Distribution Network Using Epanet and Vertex Method

Chapter · October 2016


DOI: 10.1007/978-3-319-40195-9_18

CITATION READS
1 964

2 authors:

Perumal Raja Sivakumar Ram Kailash Prasad


North Eastern Regional Institute of Science and Technology North Eastern Regional Institute of Science and Technology
23 PUBLICATIONS   71 CITATIONS    29 PUBLICATIONS   103 CITATIONS   

SEE PROFILE SEE PROFILE

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

Ph.D thesis View project

Out of Ph.D View project

All content following this page was uploaded by Perumal Raja Sivakumar on 05 July 2018.

The user has requested enhancement of the downloaded file.


ANALYSIS OF WATER DISTRIBUTION NETWORK USING
EPANET AND VERTEX METHOD

P Sivakumar1 and Ram Kailash Prasad2

1
Research Scholar, Department of Civil Engineering, NERIST, Nirjuli, Arunachal Pradesh,
2
Associate Professor, Department of Civil Engineering, NERIST, Nirjuli, Arunachal Pradesh,
Email: siva_nerist@yahoo.co.in, rkp@nerist.ac.in

Keywords: Water distribution network, EPANET 2, vertex method.

Abstract: The analysis of hydraulic behavior of the water distribution network is forefront part of the
planning and augmentation of any water supply projects. The analysis of water distribution
network determines the estimation of discharges, hydraulic gradient levels (HGL), nodal
concentrations etc. to fulfill the requirements of population. In the conventional approach of
analysis, unique value of pipe discharges and hydraulic heads are obtained. The results so
obtained may not give satisfactory performance in practice due to many uncertainties in nodal
demands, pipe roughness, lengths, diameters of pipes, water levels in reservoirs, head-discharge
characteristics of pumps etc. In this study, the uncertainty in discharges and hydraulic heads are
evaluated using two different pipe networks, the data for which is obtained from literature.
Further, the membership function of pipe roughness has been used to calculate the membership
function of discharges and hydraulic heads by incorporating EPANET with vortex method of
fuzzy approach. The uncertainties in discharges, hydraulic heads at different α-cuts are evaluated
considering the uncertainties in pipe roughness. The results of pipe discharges are found to vary
15% and 30 % whereas the hydraulic heads at nodes vary 0.3m and 3m when the uncertainty of
about 8 % in Hazen-Williams coefficient of pipe roughness is introduced in four and five pipe
networks respectively. Moreover, when this uncertainty in pipe roughness is combined with
other kind of uncertainty as discussed above, would further aggravate the uncertainty in
discharges and nodal heads of the pipe. As a result, the reliability of network would decrease in
terms of either meeting the required discharges or the nodal heads to the consumers. This study
would help to design the pipe network under the conditions of uncertainty in input parameters.

1 INTRODUCTION
The analysis of water distribution network roughness, lengths, diameters of pipes, water
(WDN) determines the estimation of discharges, levels in reservoirs, head-discharge
hydraulic gradient levels (HGL), nodal characteristics of pumps etc., Due to the complex
concentrations etc., to fulfill the requirements of behavior of WDN, the reliable measurements is
the population. In the conventional approach of not normally possible in each and every node and
analysis, unique value of the pipe discharges and links of the network. While augmenting the
hydraulic heads are obtained. The results so existing network, the length and diameters of
obtained may not give satisfactory performance pipe are assumed to be consistent even if the
due to many uncertainties in nodal demands, pipe network has been used for many years. The

1
diameter and friction coefficient of the pipe may the fuzzy variable is selected. Further, Smith et
vary due to scale formation on inside surface of al. (2002) observed that the disadvantage of the
pipes and aging process; length of pipe also vary vertex method is that the number of function
due to introduction of joints or removal of a evaluations needed increases by the number of α-
pipe line during the normal course of operation. cuts raised to the power of the number of random
However, the same have not been considered in variables.
the conventional method of analysis and hence do Revelli and Ridolfi (2002) analyzed a pipe
not give the expected result. Thus in the present network by fuzzy approach through optimization
study, pipe roughness is taken as an uncertainty considering uncertain parameters as pipe
parameter. All other parameters are taken as crisp roughness coefficient; nodal demands; pizometric
except the pipe roughness in vertex method of heads. Xu and Goulter (1999) and Bhave and
fuzzy set theory to avoid the excessive Gupta (2004) optimized the water distribution
computational requirements. The brief networks with fuzzy demands using linear
description of the EPANET, Fuzzy sets and programming. Gupta and Bhave (2007)
Vertex methods are given below. reanalyzed the network of Revelli and Ridolfi
EPANET (Rossman 2000) is a public domain (2002) by the usual method of analyses. They
software package that performs the extended concluded that their proposed methodology
period simulation of hydraulic and water quality requires less computational effort and time as
behavior within the pressurized pipe networks compared to that of Revelli and Ridolfi (2002).
which is based on the principle of gradient In the present study, two network problems
method (Todini and Pilati 1987). The output of are taken. Example 1 is a simple network
the analysis of WDN such as pipe discharges, problem taken from Bhave and Gupta (2006) to
nodal heads, nodal concentrations, etc., can be explain the proposed methodology. Example 2 is
readily calculated with available input parameters taken which has already been analyzed by Revelli
such as pipe roughness coefficient, length, and Ridolfi (2002) and Gupta and Bhave (2007).
diameter of pipe etc., Revelli and Ridolfi (2002) used Strickler
The fuzzy set theory was introduced by Zadeh coefficient of pipe roughness as fuzzy where as
(1965), it provides tool for dealing with Bhave and Gupta (2006) have considering
imprecision due to uncertainty and vagueness, Hazen-Williams coefficient of pipe roughness as
which is essential to many engineering problems. fuzzy. In this study, the Hazen-Williams
It resembles human decision making with its coefficient of pipe roughness is taken as fuzzy
ability to work from approximate data and for the simulation of WDN.
imprecise solutions. Since its inception, it has
been used to describe imprecision in ground 2 FUZZY SETS AND
water (Dou et al. 1997a; Guan and Aral 2004; VERTEX METHOD
Prasad and Mathur 2007); and in the pipe net
work problems (Revelli and Ridolfi 2002; Bhave Fuzzy set is a set of objects without clear
and Gupta 2004; Gupta and Bhave 2007). The boundaries or one without well defined
imprecisely known parameters can be taken as characteristics (Fresissinet et al. 1999). The
fuzzy and it can be defined by membership membership function establishes how much the
function (Ross 1995; Kaufmann and Gupta element “belongs” to the set and is included in
1991); however, it is computationally expensive the interval [0, 1], the convention being that the
for complex problems (Ross 1995; Guan and closer it is to 1, the more the element belongs to
Aral 2004), because of large number of the set and vice versa. Fuzzy sets, membership
simulation run of the model. functions, and α level cuts are defined
Dong and Shah (1987) suggested the vertex mathematically by Ross (1995). Thus, if X class
method which is based on the α level cut concept of objects denoted by x, then a fuzzy set in X is a
and interval analysis. In this method, the
set of ordered pairs A={x,µ A(x)|x∈X}, where
membership function is discretized, rather than
µ A(x) represents the membership function for the
discretizing the variable domain. The
fuzzy set A. The value representing the degree for
discretization of the membership domain is
an element x that belongs to the fuzzy set A is
accomplished by dividing the membership
defined as the degree of membership for x, which
domain into a series of equally spaced cuts,
is evaluated by the membership function. The α
called α-cuts; α represents the possibility. For
level cut of fuzzy set A is the set of those
each α- cut, the maximum and minimum value of
elements which have a membership value greater

2
than or equal to α (Ross 1995). When α = 0, the to real number y by the mapping y=f (x1, x2, . . . ,
corresponding interval is called the “support” of xN) the solution of fuzzy number B in y that
the fuzzy member with extreme boundaries of the would correspond to fuzzy numbers A1 in xN can
“minimum” and “maximum” values respectively. be obtained in the following procedure:
Similarly, with a triangular function when α =1, 1. The range of membership [0, 1] is discretized
the interval comes down to a crisp value, or the into a finite number of values, called α1, α2, . .
“most likely value” (Kaufmann and Gupta 1991). . , αM. The refinement in discretization
Two most common types of membership depends on the degree of accuracy desired.
function for fuzzy numbers are: (1) Triangular; 2. For each membership value αj, the
and (2) Trapezoidal. But, former one is preferred corresponding intervals for Ai in xi, i = 1,2, .
by some researchers (Dou et al. 1995, 1997a,b) . . , N are determined. These are the
because of its simple shape. The triangular supports of the αj cuts of A1, A2, A3, . . . , AN.
membership function is a special case of The end points of these intervals are
trapezoidal membership function, because in this represented by [a1, b1] , [a2, b2] , . . . , [aN ,
case, at α =1, there is a single point rather than a bN ]. Also ai may be equal to bi in which
flat line as in the trapezoidal function. However, case the interval would reduce to a point.
in this study triangular membership function has 3. Taking one end point from each of the
been used. intervals, the end points can be combined into
R. Moore (1979) has given a solution in 2N distinct permutations, giving 2N
interval computation, and found the basic combinations for the vector (x1, x2, . . . , xN).
problem as given a function f (xi, . . . ,,xn) and n Thus for 2 pipe network having the
intervals [xi-, xi+] find the interval range of the uncertain parameters as (a1, b1),
variable y=f(X) such that x∈ xi [xi-, xi+]. Yang et (a2, b2) can be combined in ordered pair as
al. (1993) defined the goal of the interval follows: [(a1, a2), (a1, b2), (b1, a2), (b1, b2)].
computation is to find the minimum and the Similarly, for 3 pipe network having the
maximum of the function when the different uncertain parameter as (a1, b1), (a2, b2), (a3,
possible values of the variables xi range in their b3) can be combined as
intervals [xi-, xi+]. Some methods are based on [(a1, a2), (a1, b2), (b1, a2), (b1, b2)] [a3, b3] can
finding a finite set of points (called be combined as
configurations or poles) on which this minimum [(a3, a1, a2), (a3, a1, b2), (a3, b1, a2), (a3, b1,
and maximum is attained. For each α-cut, the b2),
minimum and maximum value of the fuzzy (b3, a1, a2), (b3, a1, b2), (b3, b1, a2), (b3, b1,
variable is selected as shown in Fig. 1. Prasad b2)].
and Mathur (2007) compared the vertex method 4. The function f (x1, x2, . . . , xN). is evaluated for
with the ANN-GA approach in their study of each of the 2N combinations to obtain 2N for y
contaminant transport in groundwater flow. The , denoted as y1, y2, . . . , yN. The desired
formulation of vertex method (Dou et al. 1997b) interval for y is given by [∧k yk, ∨k yk], which
is as follows. In the vertex method, fuzzy define support of the αj - cut of B.
numbers A1, A2, A3, . . . , AN are defined on the 5. The process is repeated for other α - cuts to
real line L and the elements of A1 are denoted by obtain additional α - cuts of B and the solution to
xi, i = 1,2, . . . , N. If x1, x2, . . . , xN are related the fuzzy number B.

3
max
C HW − C HW
µ A (C HW ) = nor max
nor
, C HW max
≤ C HW ≤ C HW
C HW − C HW
(4)
µ A (C HW ) = 0, C HW ≥ C max
HW (5)

Here α - cut is represented by α*. If α*= 0, CHW


lies between 110 and 130; for α*= 0.4 , CHW lies
between 114 and 126; α*= 1, normal value of
CHW is 120.
Analysis of Hydraulic Network
Consider a looped pipe network having M source
nodes labeled j = 1, 2, 3, . . . , M ; N demand
nodes labeled j = M +1, . . . , M ; X pipes labeled
x = 1, 2, 3, . . . , X; C basic loops or circuits
labeled c = 1, 2, 3, . . . , C; and M – 1 pseudo
Fig . 1. Triangular membership function
loops labeled c = C+1, . . . C+M-1. Bhave and
for Hazen-Williams coefficient Gupta (2006) treated the Hazen-Williams
coefficient of pipe roughness as fuzzy. For
3 METHODOLOGY analyzing the networks, node-flow continuity
equations and head loss equations for loops have
Hazen-Williams formula for head loss been used. Node flow continuity equation states
that algebraic sum of flows at node is zero where
The head loss due to pipe friction, hf in metes as loop head loss equation states that algebraic
(Bhave and Gupta 2006) is given by sum of head losses in pipes of a loop is zero,
10.68 LQ 1.85 respectively. Hence
hf = (1)
∑ (Q x )α =α ∗ + q j = 0, j = 1,2,3,..., J
1.85
C HW D 4.87
x connected to j (6)
Where L=length of the pipe in metres; D=
diameter of the pipe in metres; Q= pipe discharge and
in cubic metres per second; and CHW =Hazen- 10.68 Lx (Q x1.85 )α =α ∗
Williams coefficient. ∑
x∈c C 1HW
.85
D x4.87
= 0, c = 1,2,3,..., C + M − 1

Fuzzy analysis through vertex method (7)


are the node flow continuity and loop head loss
Fig. 1. shows Hazen-Williams coefficient, CHW
equations respectively.
as a fuzzy parameter in a triangular membership
In the vertex method, for n piped
function, having extreme boundary values CHWmin
network, the order pair of the roughness
and CHWmax for each α- cuts (i.e., 0.0, 0.2, 0.4,
coefficient of each membership would be 2n.
0.6, 0.8). Where, CHWmin , CHWmax is the minimum
These roughness coefficients are formulated in
and maximum value of CHW. For α = 1, it
represents CHWnor , a single point value known as ordered pair and output parameters like pipe
crisp value, i.e., most likely or normal value of discharges and nodal heads are obtained using
EPANET 2. The output parameters can be
CHW. Hence for the triangular fuzzy number A
grouped for pipe discharges (Qxmin, Qxmax) and
(CHWmin, CHWnor , CHWmax) , the membership
function µ A is given by nodal heads (Hjmin, Hjmax )for each membership
function. Then the membership functions of
µ A (C HW ) = 0, C HW ≤ C HW
min
(2) discharges and nodal heads are plotted at each α
level cut.
min
C HW − C
µ A (C HW ) = nor
HW
min
min
, C HW nor
≤ C HW ≤ C HW
C HW −C HW
(3)

4
Fig . 2. A one loop network

3.1 Example. 1 [130,110,110,110];


[130,110,110,130]; [130,110,130,110];
Consider a one loop network of Bhave and Gupta [130,110,130,130];
(2006) as shown in Fig. 2. has source node 1 with [130,130,110,110];
HGL of 100 m and nodes 2, 3, and 4 are demand [130,130,110,130]; [130,130,130,110];
nodes with demands of 0.04, 0.045 and 0.025 [130,130,130,130].
m3/sec, respectively. The pipes are labeled as (1),
(2), (3) and (4). The length in metres and The maximum value of pipe discharge; Q1max
diameters in millimeters for different pipes are = 0.0442 m3/sec is obtained when CHW1 = CHW4 =
given in parentheses as pipe 1 (450 m, 300 mm); 130 and CHW2 = CHW3 = 110. The minimum value
pipe 2 (500 m, 350 mm); pipe 3 (450 m, 250 of pipe discharge Q1min = 0.0394 m3/sec is obtained
mm); and pipe 4 (500 m, 200 mm), respectively. as when CHW1 = CHW4 = 110 and CHW2 = CHW3 = 130
As explained, the Hazen-Williams coefficient in the membership function for α*= 0. 0.
of pipe roughness (CHW) is taken as uncertain and The maximum value of pipe discharge Q1max =
independent parameter to determine the 0.0437 m3/sec is obtained when CHW1 = CHW4 = 128
dependent parameters like pipe discharges and and CHW2 = CHW3 = 112. The minimum value of
nodal heads using EPANET 2. The results are pipe discharge Q1min = 0.0399 m3/sec is obtained as
shown in Table. 1. when CHW1 = CHW4 = 112 and CHW2 = CHW3 = 128
In the analysis of pipe network, taking all in the membership function for α*= 0.2.
independent fuzzy parameters at their normal All the minimum and maximum value (except
values (α = 1) of pipe roughness for all pipes i.e., α*= 1) of pipe discharges and nodal heads are
CHW1 = CHW2 = CHW3 = CHW4, the discharge value shown in the Table 1. The membership functions
of pipe (1) is Q1 = 0.0418 m3/sec (Table 1). of the corresponding parameters are shown in
For, α*= 0 the ordered pairs of CHW are Fig. 3.
obtained using vertex method (Dong and Shah
1987) .

[110,110,110,110];
[110,110,110,130]; [110,110,130,110];
[110,110,130,130];
[110,130,110,110];
[110,130,110,130]; [110,130,130,110];
[110,130,130,130];

5
6
Table 1. Values of dependent parameters for different α-cuts for network of Fig. 2
Ro Pipe Discharge (m3/sec) Nodal HGL (m)
w value Q1 Q2 Q3 Q4 H2 H3 H4
(1) (2) (3) (4) (5) (6) (7) (8) (9)
1 α ∗ = 0.0
Maximum 0.0442 0.0706 0.0306 0.0192 99.277 98.578 99.462
Minimum 0.0394 0.0658 0.0258 0.0144 98.953 98.062 99.168

2 α ∗ = 0.2
Maximum 0.0437 0.0701 0.0301 0.0187 99.251 98.536 99.439
Minimum 0.0399 0.0663 0.0263 0.0149 98.992 98.125 99.205

3 α ∗ = 0.4
Maximum 0.0432 0.0696 0.0296 0.0182 99.224 98.493 99.415
Minimum 0.0404 0.0668 0.0268 0.0154 99.030 98.186 99.240

4 α ∗ = 0.6
Maximum 0.0427 0.0692 0.0292 0.0177 99.195 98.448 99.389
Minimum 0.0408 0.0673 0.0273 0.0158 99.066 98.243 99.273

5 α ∗ = 0.8
Maximum 0.0423 0.0687 0.0287 0.0173 99.165 98.400 99.362
Minimum 0.0413 0.0677 0.0277 0.0163 99.101 98.298 99.304

6 α ∗ = 1. 0
Normal 0.0418 0.0682 0.0282 0.0168 99.134 98.350 99.334

Fig. 3. Membership functions of unknown


pipe discharges (m3/sec) and nodal HGL (m) Fig. 4. A two-loop network
for the netwok of Fig. 2

7
3.2 Example. 2 most likely value of CHW for all pipes is 120 with
minimum and maximum values of 110 and 130
Another two-loop pipe network problem is taken respectively (Fig. 1).
from Bhave and Gupta (2006). The network has In this network, there are 5 pipes with one
one source node, three demand nodes and five uncertain parameter (i.e Hazen-Williams
pipes as shown in Fig. 4. Node 1 is a source node coefficient, CHW) for which the ordered pair
with HGL of 100 m and nodes 2, 3, and 4 are becomes 25= 32 for each α–cut. Hence the
demand nodes with demands of 0.150, 0.300 and EPANET 2 model has been executed 32 times
0.230m3/sec, respectively. The pipes are labeled and the minimum and maximum values of
as (1), (2), (3), (4) and (5) and its length in metres discharges and heads are obtained.
and diameters in millimeters for different pipes The results obtained by proposed method and
given in parentheses as pipe 1 (1200 m, 500 mm); those of Bhave and Gupta (2006) are identical up
pipe 2 (1100 m, 500 mm); pipe 3 (1500 m, 500 to three decimal places. Consequently the
mm); pipe 4 (900 m, 350 mm); and pipe 5 (1000, membership functions for vertex method and
350) respectively. Bhave and Gupta (2006) are identical as shown in
In this analysis, the source HGL and nodal the Fig. 5.
demands are precise, but uncertainty exists in
Hazen-Williams coefficient, CHW of all pipes. The

8
Fig. 5. Membership functions of unknown
pipe discharges (m3/sec) and nodal
HGL (m) for the netwok of Fig. 4

4 CONCLUSIONS

Two water distribution network problems have


been analyzed to determine the membership
function of unknown pipe discharges and
hydraulic heads by assuming pipe roughness of
each pipe as uncertain. In this study, the
membership function of pipe roughness is
assumed to be triangular. Further, EPANET 2 is
used to find the pipe discharges and hydraulic
heads of each node by giving required input
parameters. The resulted membership functions of
output discharges and hydraulic heads are found
to be triangular in nature. Further, the results
obtained by proposed method and those of Bhave
and Gupta (2006) are identical up to three decimal
places. Consequently the membership functions
for vertex method and Bhave and Gupta (2006)
are identical. Thus, the proposed methodology
using vertex method and EPANET 2 has been
successful in quantifying uncertainty in pipe
discharges and nodal HGLs. The results of pipe
discharges are found to vary 15% and 30%
whereas the hydraulic heads at nodes vary 0.3m
and 3m when uncertainty of about 8% in Hazen-
Williams coefficient of pipe roughness is
introduced in four and five pipe networks
respectively. The maximum uncertainty in
discharges and hydraulic heads for both the
networks are found to be at α*= 0. This may be
due to maximum uncertainty in roughness
coefficient of the pipe.
Furthermore, EPANET 2 required to run 16
and 32 times for 4 and 5 pipe network problem

9
respectively at each α - cut level except α*= 1. using fuzzy set theory.” Water Resour. Res.,
Thus, the total number of EPANET 2 simulation 40(1), 1-20.
needed are 65 and 129 for 4 and 5 pipe network Guan, J., and Aral, M. M. (2005). “Remediation
problems respectively. The process would be time system design with multiple uncertain
consuming for large network even though there is parameters using fuzzy sets and genetic
a single uncertainty in input parameters. algorithm.” J. Hydrol. Eng., 10(5), 386-394.
Furthermore, when this uncertainty in pipe Gupta, R., and Bhave, P. R. (2007). “Fuzzy
roughness is combined with other kind of parameters in pipe network analysis.” Civil
uncertainty like nodal demands, lengths, diameter Eng. and Environ. Systems, 24(1), 33-54.
of pipes, water levels in reservoirs, head- Kaufmann, A., and Gupta, M. M. (1991).
discharge characteristics of pumps etc., would Introduction to fuzzy arithmetic: Theory and
aggravate the uncertainty in discharges and nodal applications, Van Nostrand Reinhold, New
heads of the pipe. As a result, the reliability of York.
network would decrease in terms of either Moore. R. (1979). Methods and applications of
meeting the required discharges or nodal heads to interval analysis. SIAM Studies in Applied
the consumers or both. This will lead the pipe Mathematics .
network problem to be more complex in real-life Prasad, R.K., and Mathur, S. (2007).
since the problem involves large number of pipes “Groundwater flow and contaminant transport
and different types of uncertainty in the pipes. As simulation with imprecise parametrers.” J.
a result, there is a need to reduce this Irrigation and Drainage Eng., 133(1), 61-70.
computational burden for the real life problem in Revelli, R., and Ridolfi, L. (2002). “Fuzzy
the proposed methodology. approach for analysis of pipe networks.” J.
Hydraul. Eng., 128(1), 93-101.
5 REFERENCES Revelli, R., and Ridolfi, L. (2003). Closure to
“Fuzzy approach for analysis of pipe
networks.” J. Hydraul. Eng., 129(7), 550-551.
Bhave, P. R., and Gupta, R. (2004). “Optimal
Ross, T. J. (1995). Fuzzy logic with engineering
design of water distribution networks for
applications, McGraw-Hill, New York.
fuzzy demands.” Civil Eng. and Environ.
Rossman, L. A. (2000), EPANET 2, users
Systems, 21(4), 229-245.
manual, National Risk Management Research
Bhave, P. R., and Gupta, R. (2006). Analysis of
Water Distribution Networks, Narosa Laboratory, U.S. Environmental Protection
Publishing Pvt. Ltd, New Delhi. Agency, Cincinnati, Ohio.
Smith, S. A., Krishnamurthy, T., and Mason, B.
Dong, W., and Shah, H. C. (1987). “Vertex
H. (2002). American Institute of Aeronautics
methods for computing functions of fuzzy
and Astronautics, 1-12.
variables.” Fuzzy sets and systems, 24, 65-78.
Dou, C., Woldt, W., Bogardi, I., and Dahab, M. Todini, E., and Plilati, S. (1987). “A gradient
method for the analysis of pipe networks.”
(1995). “Steady state groundwater flow
International Conference on Computer
simulation with imprecise parameters.” Water
Application for Water Supply and
Resour. Res., 31(11), 2709-2719.
Distribution. Leicester Polytechnic, U.K.
Dou, C., Woldt, W., Dahab, M., and Bogardi, I.
Xu, C. (2003). Discussion of “Fuzzy approach
(1997a). “Transient groundwater flow
for analysis of pipe networks.” J. Hydraul.
simulation using fuzzy set approach.” Ground
Water, 35(2), 205-215. Eng., 129(7), 549-550.
Xu, C., and Goulder, I. C. (1999). “Optimal
design of water distribution networks using
Dou, C., Woldt, W., Bogardi, I., and Dahab, M.
fuzzy optimization.”. Civil Eng. and Environ.
(1997b). “Numerical solute transport
simulation using fuzzy sets approach.” J. Systems, 16(4), 243-266.
Contam. Hydrol., 27,107-126. Yang, H. Q., Yao, R., and Jones, D (1993).
“Calculating functions of fuzzy numbers.”
Freissinet, C., Vauclin, M., and Erlich, M. (1999).
Fuzzy sets and systems, 55, 273-283.
“Comparison of first order analysis and fuzzy
Zadeh, L.A. (1965). “Fuzzy sets.” Inf. Control, 8,
set approach for the evaluation of imprecision
in a pesticide groundwater pollution screening 338-353.
model.” J. Contam. Hydrol., 37, 21-43.
Guan, J., and Aral, M. M. (2004). “Optimal
design of groundwater remediation systems

10

View publication stats

You might also like