Computations of Flows For On Demand Irrigation Systems

Performance analysis of on-demand pressurized irrigation systems 15
Chapter 3
Computation of flows for on-demand
irrigation systems
One of the most important problems for an on-demand irrigation system designer is the
calculation of the discharges flowing into the network. Such discharges strongly vary over time
depending on the cropping pattern, meteorological conditions, on-farm irrigation efficiency and
farmers' behaviour.
In this paper, each group of hydrants operating at a given instant is called hydrants
configuration. Each hydrants configuration produces a discharge configuration (or flow
regime) into the network. The term node includes both hydrants and junctions of two pipes,
whereas the term section is used to describe the pipe connecting any two nodes.
The design capacity is usually determined considering short-term peak demand and
considering an average cropping pattern for the whole system. But, the individual cropping
pattern may differ from the designed one, and the irrigation system may be either undersized or
oversized.
In view of the difficulty of this problem, empirical methods have been used. For example, the
US Bureau of Reclamation (1967) recommends solving each case on individual basis and gives
only general indications like: the maximum demand may generally be estimated at 125-150% of
the average demand. Systems operating for a 12- month season may require a capacity large
enough to carry from 10 to 15% of the total annual demand in the peak month. Those operating
for a 7-month season may require a capacity large enough to carry from 20 to 25% of the total
annual demand during the peak month. However, the actual maximum demand should be
determined by detailed analysis of individual projects.
The advent of on-demand, large-scale irrigation systems in the early 1960s in France fostered
the development of statistical models to compute the design flows. Examples of such models are
the first and the second Clment formula (1966). But only the first demand formula has been
widely used because of its simplicity.
Although these models are theoretically sound, the assumptions governing the determination
of their parameters do not take into account the actual functioning of an irrigation system. In
view of these limitations a number of researchers tackled the problem by simulating irrigation
strategies. As an example, Maidment and Hutchinson (1983) modeled the demand pattern over a
large irrigation area taking into account the size of irrigated area, the soil type, the cropping
pattern, the irrigation strategy and the weather variation. However, they had to average out the
demand hydrograph over time to avoid unrealistic very high water demand one day and very low
the next.
Recently other approaches have been developed combining simulation of irrigation strategies,
based on the soil water balance, and statistical models (Abdellaoui, 1986; Walker et al., 1995;
Teixeira et al., 1995). The result of these methods is a single distribution of one design flow for
16 Computation of flows for on-demand irrigation systems
each pipe section of the network. They will be referred to in the following as One Flow Regime
Models (OFRM).
OFRM do not actually take into account the hydraulic functioning in an on-demand collective
irrigation network. Indeed, in such systems there is occurrence of several flow regimes
according to the spatial distribution of the hydrants that are simultaneously in operation.
Therefore, improving the design and the performance of an irrigation system operating on-
demand requires consideration of these flow regimes in the design process. The new approach,
called Several Flow Regimes Models (SFRM), is based on this concept. In this chapter OFRM
will be reviewed before presenting the process of generating flow regimes for the SFRM.
ONE FLOW REGIME M ODELS (OFRM)

Statistical models: the Clment models
One study that deals explicitly with calculation of pressurized irrigation systems capacity for on-
demand operation is the work of Clment (1966). Two different models were proposed. One
(called the first Clment model) is based on a probabilistic approach where, within a population
of R hydrants, the number of hydrants being open simultaneously is considered to follow a
binomial distribution. The other (called the second Clment model) is based on simulating the
irrigation process as a birth and death process in which, at a given state j (j hydrants open), the
average rate of birth is proportional to (R-j) and the average rate of death is proportional to j.
The Clment models, although based on a theory, were extensively used for designing sprinkler
irrigation systems in France, Italy, Morocco and Tunisia.
The first Clment model

Background equations
In on-demand irrigation systems, the nominal discharge of the hydrants (d) is selected much
higher than the duty, D (the duty is the flow based on peak period water requirement on a 24-
hour basis: D=qs Ap, where qs is the continuos specific discharge and Ap is the area of the plot
irrigated by the hydrant). It allows farmers to irrigate for a duration lower than 24 hours. This
condition implies that the event to find all the hydrants simultaneously operating has very low
probability. Thus, it is not reasonable to calculate the irrigation network by adding the discharges
delivered at all the hydrants simultaneously. Consequently, probabilistic approaches for
computing the discharges into the sections of an on-demand collective network have been widely
used in the past and are still used actually.
The most utilized is the probabilistic approach proposed by Clment (1966) and it is
summarized hereafter.
Let q s be the specific continuous discharge, 24 hours per day (l s 1 ha 1 ), A is the irrigated
area (ha), R the total number of hydrants, d the nominal discharge of each hydrant (l s 1 ), T the
duration of the peak period (h), T' the operating time of the network (h) during the period T, r the
coefficient of utilization of the network (defined as the ratio T'/T). The average operation time t'
of each hydrant during the peak period (h) is then
qs AT
t' = ( )/ d (1)
R
The elementary probability, p, of operation of each hydrant is defined as
t' t' q s AT 1
p= = = (2)
T' r T R d rT
thus,
qsA
p= (3)
rRd
Therefore, for a population of R homogeneous hydrants, the probability of finding one hydrant
open is p, while (1 - p) is the probability to find it closed.
The number of operating hydrants is considered a random variable having a binomial
distribution with mean
=Rp (4)
and variance
2 = R p (1-p) (5)
Therefore, the cumulative probability, P q, that among the R hydrants there will be a maximum
of N hydrants simultaneously operating is:
N
Pq = C KR p K (1 p ) (R K) (6)
K= 0
where:
R!
C KR = (7)
K!( R K)!
is the number of combinations of R hydrants taken K at a time. When R is sufficiently large (R >
10) and p > 0.2-0.3, the binomial distribution approximates the Laplace-Gauss normal distribution
whose cumulative probability (P q) for having a maximum of x hydrants simultaneously operating
(with - < x < N) is:
2
U(Pq ) u
1

2
Pq = e du (8)
2p
where U(P q) is the standard normal variable corresponding to the probability P q, and u is the
standard normal deviate given by:
xRp
u= (9)
R p (1 - p)
2
u

2
The integral (8) is solved developing in series the exponential function e . The solutions
of this integral have been tabulated (see Table 1) and so, according to a prefixed value P q , it is
possible to determine the corresponding value U(P q).
Knowing U(P q), it is possible to calculate the number of hydrants simultaneously operating, N,
through the relationship (9). In fact, for u = U(P q) we have:
N = R p + U( Pq ) R p (1 - p) (10)
that is the first formula of Clment.
Considering hydrants with the same discharge, the total TABLE 1

discharge downstream a generic section k is given by: Standard normal cumulative
distribution function
Q k = R p d + U(Pq ) R p (1 p ) d 2 (11) Pq
0.90
U(Pq)
1.285
0.91 1.345
and, for different discharges of hydrants (di,), where I is the
0.92 1.405
hydrant class number 0.93 1.475
0.94 1.555
Q k = i R i p i d i + U( Pq ) i R i p i ( 1 p i ) d i
2
(12) 0.95 1.645
0.96 1.755
The first Clment model is based on three major hypotheses 0.97 1.885
that limit its applicability (CTGREF, 1974; CTGREF, 1977; 0.98 2.055
Lamaddalena and Ciollaro, 1993). 0.99 2.324
The first hypothesis concerns the parameter r. It is defined by
Clment as coefficient of utilization of the system in the sense
that, during the design phase, the duration of the day for irrigation, within the peak period, is
considered shorter than 24 hours. This parameter, defined at the network level in an irrigation
system operating on-demand, should have a value equal to one because these systems may
have to work 24 hours per day. In practice, the parameter r should correspond to the
operating time of each hydrant and, therefore, it is not correct to use it for the global design of
the system. Nevertheless, from a conceptual point of view, it may be considered as a
parameter which helps adjusting the theoretical formulation to a homogeneous population of
discharges, withdrawn in the field, appropriately chosen through a statistical approach. It must
be pointed out that the Clment model, like all other models, only offers a schematic
representation of an actual network. Therefore, it must be adjusted or calibrated by
introducing field data relative to existing networks. In particular, values of the parameter r
should be, whenever possible, selected for homogeneous regions and for particular crops. An
example of the field Clment model calibration, for an Italian irrigation network, is reported in
Annex 1.
The second hypothesis concerns the elementary probability of opening each hydrant. It refers
to an estimation of the average operating time of each hydrant. But, the probability to find a
hydrant working at a given time t depends on its state at the previous time t-1. In order to
justify the binomial law, this probability should characterize a series of events such that when
the farmer opens his hydrant at a time t, he would close it after a laps of time dt, and he
decides to re-open or to leave it closed at a successive time t+dt, and so on. This is not real
because a farmer opens his hydrant and leaves it in the same state for a large number of laps
of time dt. Moreover, the elementary probability varies during the day according to the
farmer's behaviour.
The third hypothesis considers the independence of the hydrants and their random operation
during the peak period. This hypothesis might seem justified because the farmers should
behave autonomously and not according to the operation of the neighbour farmers.
Nevertheless, the rhythm of nights and days and the similitude of the crops within an irrigation
district condition the farmer's behaviour, so that this hypothesis is not fully reliable.
The importance of the r coefficient is stressed also in Figure 10. In this figure two parameters
are defined: the elasticity of the network, eR (Clment and Galand, 1979):
QCl
eR= (13)
qs A
and the average elasticity of the hydrants (called also farmers degree of freedom), e h,
Rd
eh= (14)
(q s A)
The ratio e R is a measure of the over-capacity of the network and is a characteristic of on-
demand operation. The ratio e h defines the freedom afforded to farmers to organize their
irrigation.
The values of eR refer to a network designed to supply equal flows at all hydrants. When the
hydrant design flows are unequal, the values of the ratio are slightly greater. Nevertheless,
whether the hydrants are homogeneous or not, taking into account the probability of the demand
being spread results in a network peak design flow which is very much smaller than that which
would be obtained by summating the flows at all hydrants.
The degree of freedom that is to be afforded to farmers should be selected according to
criteria such as size and dispersion of plots, availability of labor, type of on-farm equipment,
frequency of irrigation. Hydrants with capacities of one and a half to twice the value of the duty
correspond to the lowest feasible degree of freedom. With smaller values, the probability of an
hydrant being open becomes too great for the demand model to apply. Conversely, hydrant
capacities should not exceed six to eight times the value of the duty. This corresponds to a very
high degree of freedom.
Figure 10 illustrates the variation of the elasticity of the network, eR, versus the total number
of hydrants, R, for different values of the elasticity of the hydrants, e h. The curves have been
drawn for U(P q)=1.645. Considering a value of elasticity at the hydrant eh = 4.5, for a network
having R=100 hydrants, the elasticity of the network eR varies from about 1.43 (corresponding to
r = 0.9) to about 2.03 (corresponding to r = 0.6). It means that the upstream discharge in an on-
demand network (from Eq. 13):
QCl = eR qs A (15)
may increase about 45% if a coefficient r=0.6 is chosen instead of a coefficient r=0.9.
Furthermore, from Figure 10 it can be seen that, for on-demand systems, the ratio of the peak
flow in the network to the assumed continuous flow (elasticity of the network) increases as the
number of hydrants decreases. With hydrant capacities two to four times greater than the duty,
by selecting r=0.9 the peak flow in a network having 100 hydrants is only 27 to 40 percent
greater than the continuous flow; while by selecting r=0.6 the peak flow in a network having 100
hydrants is 79 to 98 percent greater than the continuous flow. It means that the coefficient r has
much more influence on the design capacity of the network respect to the elasticity of the
hydrants. Therefore, in order to give more freedom to farmers, it is more appropriate to select
higher hydrant elasticity.
The values selected for the parameter r normally lie between 16/24 (r=0.67) and 22/24
(r=0.93). The performance analysis of existing networks is the most reliable approach for
selecting the coefficient r best suited to a given irrigation context.
The parameter U(P q) defines the "quality of operation" of the network; it normally has a
values ranging from 0.99 to 0.95. It is hardly possible to go below a value of 0.95. A significant
reduction of this parameter beyond these values can lead to the occurrence of unacceptable
failures to satisfy the demand in certain parts of the network (Galand et al., 1975).
In view of the hypotheses made when formulating the on-demand model it is recommended
that a deterministic approach be adopted at the extremities of the network by cumulating the
flows at the hydrants when their number falls below a certain value which, in practice, lies
between four and ten.
FIGURE 10
Variation of the elasticity of the network, eR, versus the total number of hydrants for different
values of r and eh, for U(Pq) = 1.645.
In certain cases it may happen that the calculated discharge of a section serving five or six
hydrants is less than that of the downstream section serving four hydrants whose flows have
been summed. In this case the discharge in the upstream section will be equal the discharge in
the downstream section.
Determination of the specific continuous discharge

In order to apply the above methodology it is necessary to know the value of the specific
continuous discharge, qs (l s-1 ha-1) in the network downstream of the section under consideration.
Its value can readily be determined when:
The cropping pattern is identical throughout the area. If this is so the specific continuous
discharge, qs (l s-1 ha-1), estimated by giving due weight to each of the crops, holds good for
every farm and all branches of the network under consideration.
The cropping intensity is identical throughout the area. When this is so, the ratio between the
net irrigated area and the gross area also holds good for every holding and all parts of the
network under study.
A number of computer packages are available for such a computation (CROPWAT,
ISAREG, etc.), as well as an extensive literature. Therefore, the reader is referred to them for
its calculation.
Discharge at the hydrants

Although a farmer supplied by an on-demand system is free to use his hydrant at any time, a
physical constraint is nevertheless imposed as regards the maximum flow he can draw. This is
achieved by fitting the hydrant with a flow regulator (flow limiter). The discharge attributed to
each hydrant is defined according to the size and crop water requirements of the plot. It is
always greater than the duty so as to give the farmer a certain degree of freedom in the
management of the irrigation.
The ratio between the discharge attributed to each hydrant and the duty is a measure of the
"degree of freedom" which a farmer has to manage irrigation. The wide variety of agronomic
situations is reflected by the wide range of the value of the degree of freedom found in practice
(FAO-44, 1990):
High degree of freedom: family holdings with limited labour, low crop water requirements,
small or scattered plots, low investment level in on-farm equipment;
Low degree of freedom: large size plots, large scale farming, abundant labour, high
investment level in on-farm equipment.
Since the maximum flow at hydrants is fixed by flow regulators it is usual to opt for a
standard range of flows. Such ranges vary from country to country.
In southeastern France, for instance, a range of six hydrants has been standardized,
corresponding to the following discharges:
Class of hydrant 0 1 2 3 4 5
-1
Discharge (l s ) 2.1 4.2 8.3 13.9 20.8 27.8
In Italy the range of hydrants corresponds to the following discharges:

Class of hydrant 0 1 2 3 4 5
-1
Discharge (l s ) 2.5 5 10 15 20 25
In Box 1, an example of the Clment formula is worked out and the results are shown in
Table 2 (generated by COPAM).
TABLE 2
Discharges flowing into each section of the network under study (output of the COPAM package:
computation with the first Clment model)
Section Initial Final Number of Area 1st Clment
Number Node Node Hydrants (ha) discharge (l/s)
********************************************************
1 0 1 19 57.00 60.00
2 1 2 18 54.00 60.00
3 2 3 17 51.00 50.00
4 3 4 16 48.00 50.00
5 4 5 15 45.00 50.00
6 5 6 14 42.00 50.00
7 6 7 11 33.00 40.00
8 7 8 8 24.00 40.00
9 8 9 7 21.00 40.00
10 9 10 6 18.00 40.00
11 10 11 5 15.00 40.00
12 11 12 5 15.00 40.00
13 12 13 4 12.00 40.00
14 13 14 3 9.00 30.00
15 14 15 2 6.00 20.00
16 15 16 1 3.00 10.00
17 7 17 3 9.00 30.00
18 17 18 3 9.00 30.00
19 18 19 2 6.00 20.00
20 19 20 1 3.00 10.00
21 6 21 3 9.00 30.00
22 21 22 2 6.00 20.00
23 22 23 1 3.00 10.00
24 8 24 1 3.00 10.00
BOX 1: APPLICATION OF THE FIRST CLMENT MODEL FOR THE COMPUTATION OF FLOWS IN AN
ON-DEMAND NETWORK
L=41
3m
L=17
L=4 3 m
3m
Data:
-1 -1
qs = specific continuous discharge (24/24 hours) = 0.327 l s ha
Ap = area of the plot to be irrigated = 3 ha
A = total irrigable area = 3 19 = 57 ha
-1
d = nominal discharge of the hydrants = 10 l s
U(Pq ) = operation quality = 1.645
r = use coefficient = 0.667
R = total number of hydrants = 19
Nba= minimum number of hydrants in simultaneous operation = 4
****************
BOX 1 Contd
qs A 0.327 57
p = elementary probability = = = 0.147
rRd 0.667 19 10
1) Number of hydrants simultaneously operating downstream the section 0-1:
N0-1 = Rp +U(Pq ) Rp (1- p) = 19 0.147 + 1.645 19 0.147 (1- 0.147) = 5.33

(1)
We assume N 0-1 = 6
The design discharge downstream the section 0-1 is:
-1
Q0-1 = N 0-1 d = 6 10 = 60 l s
2) Number of hydrants in simultaneous operation downstream the section 1-2:
N1-2 = R1- 2 p +U(Pq ) R1-2 p (1 - p) = 18 0.147 + 1.645 18 0.147 (1- 0.147) = 6.11
We assume N 1-2 = 6
-1
Q1-2 = N 1-2 d = 6 10 = 60 l s
N2-3 = R 2-3 p+ U(Pq ) R 2-3 p (1- p) = 17 0.147 + 1.645 17 0.147 (1- 0.147) = 4.90
We assume N 2-3 = 5
-1
Q2-3 = N 2-3 d = 5 10 = 50 l s
and so on for the calculation of the discharges in the other sections.
In order to facilitate the calculation of large networks, a computer software program called
COPAM (Combined Optimization and Performance Analysis Model) has been developed by
Lamaddalena (1997). This software (enclosed) has several options that will be explained in this
paper.
All the computer programs for computation of irrigation systems require detailed information
on the pipe network transporting water from the source to the demand points (hydrants). In the
following section some general information on the use of COPAM, as well as the installation of
the program and the preparation of the input data files, are presented.
Installation of COPAM
Basic Windows knowledge is required for installing the COPAM Package.
Create an appropriate directory in the hard disk (it can be called Copam),
Insert the install disk in the appropriate drive,
Copy all files from the install disk to the directory previously created (e.g. Copam). Verify that
the *.dll files have been copied,
From Windows Explorer, open the directory Copam and create a FIGURE 11
shortcut on the desktop for the file (icon) copam.exe, COPAM icon
Click the icon label once to change the name. It is suggested to
call it COPAM (Figure 11).
Starting COPAM
Double click on the COPAM icon. Figure 12 will occur for few
seconds: COPAM
FIGURE 12
First screen of the COPAM package
Then Figure 13 will appear.

Three different sets of programs are available in the COPAM package:
Discharges computation
Pipe size computation
Analysis
Two programs are available under the set Discharges computation: Clment and Random.
One program is available under the set Pipe size computation: optimization. Two programs are
available under the set Analyses: Configurations and Hydrants. For all the above software, the
basic input file is the same and it is explained below.
FIGURE 13
Layout of the COPAM package
FIGURE 14
File menu bar and its sub-menu options
Preparation of the input file

The line of words beginning File is a menu bar. When you click on any of the words with your
mouse, a sub-menu will drop down. For example, if you click on the File menu, you will see the
sub-menu in Figure 14.
Different options are available in the File menu: New, Open, Save, Save as, Print input file,
Setup printer, View file, Exit. All these options are familiar for windows users.
When you want to create a new file, the option New is clicked. Then the menu bar Edit
is selected and the sub-menu in Figure 15 will drop down.
FIGURE 15
Edit menu bar and its sub menu items
By clicking on each option, the input data may be inserted using the Edit menu. It is strongly
recommended to prepare the input data before entering in the program.
First select the sub-menu Edit/Hydrants Discharge by clicking on it. Figure 16 will appear
and the list of the nominal discharges of the hydrants is inserted in the appropriate edit box. The
list of hydrants discharge is introduced in an increasing order and the user should enter any
standard values of discharge selected for the project, as used in the respective country.
The command OK will close the option and store the information.
By clicking on the option Network layout, input data are inserted in the sub-menu
Edit/network layout.
FIGURE 16
Sub menu Edit/Hydrants discharge
It is assumed the network is of the branching type. Each node (hydrants and/or linking of
sections) is positioned by a number. The node numbering is extremely important for the correct
execution of the program. It has to be allocated as follows:
The upstream node (source) must have number 0
The other nodes are numbered consecutively, from upstream to downstream. Any node may
be jumped.
The number of the section is equal to the number of its downstream node.
All terminal nodes of the branches must have a hydrant.
No more than two sections may be derived by an upstream node. If so, an imaginary section
with minimum length (i.e.: lmin = 1 m) must be created and an additional node must be
considered. This node must have a sequential number.
No hydrants may be located in a node with three sections joined. If so, an additional node
with a sequential number must be added.
If hydrants with two or more outlets exist in the network, one number for each outlet needs to
be allocated by creating an imaginary section with minimum length.
When the numbering has been completed, the following information has to be entered in the
Edit/Network layout:
FIGURE 17
Examples of node numbering
area irrigated by each hydrant (in hectares); if no hydrant occurs in the node, Area=0 has to
be typed,
hydrant discharge (in l s-1). It may be selected by clicking in the combo box,
section length (in m),
land elevation of the downstream node (in m a.s.l.),
nominal diameter of the section pipe (in mm). This information is needed when the program is
used for the analysis of the network. In the design stage, Diameter=0 must be considered.
In Figure 18, an example of the sub-menu Edit/network layout is reported. Additional
options are available on the bottom of the screen. They may be activated by clicking on the
button: Add Node, Canc Node, Ins Node and Find Section, respectively for adding a new node,
delete a node, insert a node and for finding a section. The exit button closes the sub menu.
FIGURE 18
Edit/network layout sub-menu
When the network layout is completed, the Edit/list of pipes sub menu is selected and
Figure 19 will appear.
The list of commercial diameters (in mm) is inserted in the Edit/list of pipes sub-menu. The
list has to be completed by the thickness (in mm) of the pipes, the roughness (, Bazin
coefficient) and the unitary cost of the pipe. An internal procedure of the COPAM package will
link, automatically, the currency to the regional setting properties of your computer. The pipes
unitary costs are typed in increasing order. The nominal diameters are typed in the grid. When
the nominal diameter corresponds to the internal diameter, the pipe thickness is considered equal
to zero. The types of pipes are identified by the Bazin roughness coefficient.
In the sub-menu Edit/Description (Figure 20), the description of the file may be typed. This
information is important when a large number of data files are managed to aid with recognition.
FIGURE 19
Edit list of pipes sub-menu
FIGURE 20
Edit description sub-menu
FIGURE 21
Toolbar buttonCheck input file
FIGURE 22 st
Clment parameters: 1 Clment formula
An additional option is available in the COPAM package: the toolbar button Check input
file (see Figure 21). It checks for the most common errors in the input file.
Computation of discharges
When the set Discharge computation is selected, two different programs are available:
Clment and Random. The program Clment allow the computation of the discharges flowing
into the network through the first and the second Clment models. When the first one is selected,
additional parameters have to be typed in the Clment parameters/sub-menu (Figure 22). They
are the:
specific continuous discharge (in l s-1 ha-1)
minimum number of terminal open hydrants
percentage of uncultivated land (in %)
Clment use coefficient (r)
Clment operation quality, U(P q).
An example of the output file of the program Clment is the one reported in Table 2. The
name of the output file is typed in the appropriate edit box and the extension .cle is
automatically assigned to the file.
The Second Clment's model

Basic theory
Considering the limitations in the first formulation, Clment developed a second model for
calculating discharges in irrigation systems operating on-demand (Clment, 1966). This second
Clment model is based on the Markovian stochastic theory of birth and death processes. It is
summarized hereafter. The complete formulation is reported in Clment (1966) and
Lamaddalena (1997). The derivation of the second Clment model is based on some
fundamental concepts on the theory of the stationary Markovian processes.
Consider a set of customers arriving at a service station (hydrants in the case of an irrigation
system). The pattern of arrivals is described by a distribution function of times of arrival. The
customers require different times to be served and, thus, the times of service are described by
another statistical distribution function. For an irrigation system we can consider:
customers in service, which is the average number of arrivals during the average time period
of operation of a hydrant;
customers served, which is the average number of hydrants operating at a given instant t;
customers in the queue, which is the average number of arrivals when the system is saturated
(during the average waiting time).
Consider a generic system characterized by a random function, X(t), assuming values 1,
2, .........., N, which represents any possible state of the system.
In the case of an irrigation system, the state of the system is defined by the number of
hydrants in operation, while birth and death correspond to opening or closing one hydrant,
respectively.
Consider an irrigation network having R hydrants. Let us assume that the operating time for
all hydrants follows the same distribution function with average duration of the hydrants
operation equal to the average irrigation time, IT. Furthermore, let us assume that the network
has been designed with the hypothesis of having N hydrants simultaneously open (N<R).
Define PSAT as the probability of saturation of a network equipped with (R-1) hydrants and u'
the standard normal variable. It may be demonstrated (Clment, 1966; Lamaddalena, 1997) that:
N R p
u' = (16)
R p (1 - p)
and
1 ( u' )
PSAT = (17)
R p (1 - p) ( u' )
where (u') and (u') are, respectively, the Gaussian probability distribution function and the
Gaussian cumulative distribution function. Making:
( u' )
F(u' ) = (18)
( u' )
the Equation 17 becomes
1
P SAT = F(u' ) (19)
R p (1 - p)
At this stage, it is easy to fix the standard normal variable, u', according to the target
probability, p, and to determine the corresponding values of (u') and (u') and also the value of
F(u')1. In fact, p is given by the Equation 2, P SAT can be selected (usually, P SAT = 0.01 is
suggested) and F(u') can be calculated using:
F(u' ) = PSAT R p (1 - p) (20)

from a diagram representing F(u') as a function of u', or directly from the equation representing
such a function (Figure 23). It is then possible to determine the corresponding value of u' and,
using the Equation 16, we can finally calculate the number of hydrants simultaneously operating
in the network:
N = R p + u' R p (1 - p) (21)
where Equation 21 represents the second formula of Clment.
1
It is possible to use the statistical tables giving, for each value of u', the corresponding values of (u')
and (u') and the corresponding functions. For the present work the functions (u') and (u') have
been calculated and introduced in a computer program for solving the discharge calculation by using
the 2nd Clement's formula.
FIGURE 23
Diagram representing u' as a function of F(u')
The structure of this second formula of Clment is similar to the first Clment formula but, in
this case, u' is not a constant depending on the selected cumulative probability (corresponding to
the quality of operation of the network), but it is a function of PSAT, p and R.
The second Clment model is based on the theory of birth and death processes. This
hypothesis limits its applicability. In fact, this theory is well applied for designing telephone lines,
where if the busy line is engaged (saturation) the customer has to call later. But for irrigation
systems it is not so easy to establish saturation conditions. Furthermore, also when the system is
saturated farmers may decide to irrigate with a lower pressure and/or discharge at the hydrant.
Finally, the complexity in mathematical approach and the negligible differences in results pushed
all designers to apply anytime the first model instead of the second one. In Table 3, the
discharges flowing into each section of the network in the example are reported.
nd
Applicability of the 2 Clment model
The computer package COPAM may be used for computing the discharges into each section of
an irrigation network by using the 2nd Clment model. The Clment parameters/sub-menu of
the Clment program (see Figure 24) is filled with the parameters described above.
The application of the 1st and the 2nd Clment models for computing the discharges into an
large Italian irrigation network (the one illustrated in the Annex 1) is shown hereafter. The
following design data were used for these calculations. The network is equipped with 660
hydrants of 10 l s-1 and the design values of the irrigated area Ai, the specific continuous
discharge, qs, and the coefficient of utilization of the network, r, are respectively: Ai = 2030 ha,
qs = 0.327 l s-1 ha-1 and r = 0.66. The 1st Clment's model was applied using the cumulative
probability Pq=95% (corresponding to U(P q) = 1.645), while the 2nd Clment's model was applied
using the probability of saturation PSAT =1%.
BOX 2: APPLICATION OF THE SECONDCLMENT MODEL FOR THE COMPUTATION OF FLOWS IN AN

ON-DEMAND NETWORK
Data:
-1 -1
qs = specific continuous discharge (24/24 hours) = 0.327 l s ha
Ap = area of the plot to be irrigated = 3 ha
A = total irrigable area = 3 19 = 57 ha
-1
d = nominal discharge of the hydrants = 10 l s
PSAT = probability of saturation = 0.01
r = use coefficient = 0.667
R = total number of hydrants = 19
Nba= minimum number of hydrants in simultaneous operation = 4
****************
BOX 2 Contd
qs A 0.327 57
p = elementary probability = = = 0.147
rRd 0.667 19 10
0.2623
u = standard normal variable = 3.9715 4.1693 ( PSAT Rp (1 - p) )
1) Number of hydrants simultaneously operating downstream the section 0-1:
N0-1 = Rp+u' Rp (1- p) =
{ [
19 0.147 + 3.9715 - 4.1693 ( 0.01 19 0.147 (1- 0.147) ]0.2623 } 19 0.147 (1 - 0.147)
= 6.77
1
We assume N 0-1 = 7
-1
Q0-1 = N 0-1 d = 7 10 = 70 l s
N1-2 = Rp+u' Rp (1- p) =
{ [
18 0.147 + 3.9715 - 4.1693 ( 0.01 18 0.147 (1- 0.147) ]0.2623 } 18 0.147 (1 - 0.147) =
= 6.53
We assume N 1-2 = 7
-1
Q1-2 = N 1-2 d = 7 10 = 70 l s
N2-3 = Rp+u' Rp (1- p) =
{ [
17 0.147 + 3.9715 - 4.1693 ( 0.01 17 0.147 (1- 0.147) ]0.2623 } 17 0.147 (1 - 0.147) =
6.29
We assume N 2-3 = 6
-1
Q2-3 = N 2-3 d = 6 10 = 60 l s ,
and so on for the calculation of the discharges in the other sections.
_________________________________
1 In this example the mathematical approximation was used for computing the discharges in each section of the
network.
TABLE 3
Discharges flowing into each section of the network under study (output of the COPAM package:
computation with the second Clment model)
Section Initial Final Number of Area 2nd Clment
Number Node Node Hydrants (ha) discharge (l/s)
********************************************************
1 0 1 19 57.00 70.00
2 1 2 18 54.00 70.00
3 2 3 17 51.00 60.00
4 3 4 16 48.00 60.00
5 4 5 15 45.00 60.00
6 5 6 14 42.00 60.00
7 6 7 11 33.00 50.00
8 7 8 8 24.00 40.00
9 8 9 7 21.00 40.00
10 9 10 6 18.00 40.00
11 10 11 5 15.00 40.00
12 11 12 5 15.00 40.00
13 12 13 4 12.00 40.00
14 13 14 3 9.00 30.00
15 14 15 2 6.00 20.00
16 15 16 1 3.00 10.00
17 7 17 3 9.00 30.00
18 17 18 3 9.00 30.00
19 18 19 2 6.00 20.00
20 19 20 1 3.00 10.00
21 6 21 3 9.00 30.00
22 21 22 2 6.00 20.00
23 22 23 1 3.00 10.00
24 8 24 1 3.00 10.00
FIGURE 24
nd
Clment parameters: 2 Clment formula
FIGURE 25
st nd
Comparison between discharges computed with the 1 and the 2 Clment equations applied to
the Italian irrigation network illustrated in Annex 1
DISCHARGE [l/s] (1st Clement' formula)
FIGURE 26
st
Probability of saturation versus the 1 Clment model discharges
5
PROBABILTY OF SATURATION (PSAT)
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
DISCHARGE (1st Clement's formula) [l/s]
The results of the application are summarized in Figure 25 which shows that, for the case
under study, using the second model slightly higher discharges are obtained. Further analyses
have shown that when the probability of saturation PSAT is computed using the discharges given
by the 1st Clment model (using the above design parameters) it becomes evident (Figure 26)
that PSAT would be near to 5% for low discharges but would approach PSAT= 1% for higher
discharges. When U(P q) is computed from the discharges corresponding to the 2nd Clment
model, similar results are obtained with a limit value of U(P q) close to U(P q) = 1.645 (Figure 27).
FIGURE 27
nd
Operation quality, U(Pq) versus the 2 Clment model discharges
3
OPERATION QUALITY (U(Pq))
2.75
2.5
2.25
1.75
1.5
1.25
1
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
DISCHARGE (2nd Clement's formula) [l/s]
This example calculation showed that results from both Clment equations are compatible
and differences are negligible in practice. Because of the limitation of the theory of birth and
death processes applied to the irrigation networks and the mathematical complexity almost all
designers use the first model instead of the second.
SEVERAL FLOW REGIMES M ODEL (SFR M ODEL)

The configuration of discharges (flow regimes) in an irrigation network greatly affects design
and operation conditions. In the next sections a model for simulating the water demand in a
pressurized irrigation system operating on-demand is described.
Model for generating random discharges

The following approach is proposed in this section for simulating the possible operation conditions
of an on-demand irrigation system. It operates using the random generation of K hydrants
simultaneously opened among the total number R (with K<R). This generation is performed to
consider the temporal variability of the discharges flowing into the irrigation network. Each
generation, according to the definition reported in the technical literature (CTGREF, 1979;
Bethery et al., 1981) produces a hydrants configuration. Each hydrants configuration
corresponds to a discharge configuration. Therefore, the discharge in the sections of the network
is calculated as the sum of the discharges withdrawn from the downstream opened hydrants.
Before illustrating the procedure used for the generation of discharge configurations, it is
important to underline that the use of a computer for generating random numbers is considered
conceptually impossible. In fact, any program will produce an output that is predictable, hence
not truly random because the numbers are not unrelated. Nevertheless, practical computer
random number generators are in common use (Knuth, 1981). To distinguish purely random
physical process from computer-generated sequences, the term pseudo-random is used.
There exists a body of random number generators which mutually do satisfy the definition
over a very broad class of application programs. But what is random enough for one application
may not be random enough for another. The chi-square test may be used to verify the goodness
of the random generation for a particular application (Press et al., 1989; Knuth, 1981).
For generating random discharge configurations, uniform deviates are applied. The uniform
deviates are just random numbers which lie within a specified range, with any one random
number in the range just as likely as any other (Press et al., 1989).
The procedure utilized is as follows (see Figure 28):
1- Every node of the system receives an identification number, j.
2- Since some nodes do not correspond to hydrants, a code identification of the hydrants is
assigned (ch=0 when there is no hydrant at the node, ch=1 when there is a hydrant with
nominal discharge of 5 l s-1, ch=2 when there is a hydrant with nominal discharge of 10 l s-1,
ch=3 when there is a hydrant with nominal discharge of 20 l s-1).
3- Definition of a list of discharges to be tested.
4- Definition of the number of hydrants configuration to be generated, corresponding to the
discharges defined in the previous list.
5- Aggregation of the hydrants discharges in each configuration, in order to calculate the
discharges flowing into the sections of the network (discharge configuration).
FIGURE 28
Typical layout of a network
-1 -1 Nodes without hydrant

O Hydrants for d=5 l s Hydrants for d = 10 l s
A computer program written in Turbo Pascal, version 6.0 generates random sequences using
the internal procedure called Random (Schildt, 1986). The random number generator is
initialized by the procedure Randomize. The reliability of the Turbo Pascal random number
generator has been verified by Schildt (1986). Additional tests have been also performed by El
Yacoubi (1994) and Lamaddalena (1997). The program for the random generation of discharge
configurations (RGM: Random Generation Model) is integrated in the COPAM package. It is
selected by clicking on the appropriate button (Figure 29). Then Figure 30 will appear on the
screen. The input data for the irrigation network were illustrated in the previous section.
FIGURE 29
Random generation program
The RG Model may be used for two different purposes:

analysis of existing irrigation systems;
design of new irrigation systems.
In the first case, this model is based on the knowledge of the demand hydrograph at the
upstream end of the network. In fact, it allows the selection of the upstream discharge
corresponding to various hydrant configurations. The value of the upstream discharge is inserted
in the appropriate edit box (see Figure 30). Corresponding to the selected discharge, a number of
hydrants simultaneously operating (hydrant configuration) is automatically withdrawn. This
procedure is repeated for several configurations and is used for analysing the system, as
illustrated in the chapter 5. The number of configurations (or flow regimes) to generate is typed
in the appropriate edit box (see Figure 30). It must be multiple of 10.
All the generated flow regimes are stored in an output file with its name typed in the
appropriate edit box. The extension .ran is automatically assigned to this output file.
When a new irrigation system is designed, the upstream discharge is not known a priori.
Therefore, to allow the generation of different hydrants configurations, such discharge may be
computed, for example, with the Clment models 1. This is achieved through the COPAM
1
Other models for generating the upstream discharges are under study. Nevertheless, despite the good
improvements that have been made in the generation of the upstream withdrawn volumes, the problem
of transformation from volumes to discharges has still not been solved because of a number of
FIGURE 30
Random generation parameters
package as explained in the previous section. The discharge value is allocated in the edit box
Upstream discharge of the Random generation parameters. After including the number of
configurations to be generated in the appropriate edit box (it must be multiple of 10), operation of
the network is simulated by generating the hydrants configurations. These hydrants
configurations, stored in a pre-selected output file, are taken into account for computing the
optimal pipe size of the network by using the approach illustrated in the chapter 4. Applications
of the above methodologies for design and analysis of Italian irrigation networks are reported in
Lamaddalena (1997).
uncertainties involved in that process. Some authors (Abdellaoui, 1986; El Yacoubi, 1994; Pereira et al.,
1995; Lamaddalena, 1997) presented interesting approaches for the solution of such a problem but only
a few applications are available and additional tests are needed for validation and for generalizing the
applicability of such models. For this reason, the first Clement model is preferred for the computation of
the discharge at the upstream end of the network and, in correspondence of such discharge, a large
number of hydrant configurations may be generated. In this way, the actual operation conditions are
simulated and are taken into account for the design (and/or the analysis) of the irrigation system.
Chapter 4
Pipe-size calculation
The problem of calculating the optimal pipe size diameters of an irrigation network has attracted
the attention of many researchers and designers. Many optimization models based on linear
programming (LP), non-linear programming (NLP) and dynamic programming (DP) techniques
are available in literature. These models give important improvements to practical problems
through less costly solutions and less computation time respect to the classical approaches.
One limitation of the optimization models is that they consider only One Flow Regime
(OFR) during the process of the pipe size computation. With this kind of approach, there is no
assurance that the system selected is the least costly and compatible with the required
performance. Indeed, in on-demand irrigation systems, the distribution of flows in each section
may strongly vary in time and space. Thus, several flow regimes (SFR) should be taken into
account during the computation process.
In this chapter an optimization model using the Labye's Iterative discontinuous algorithm is
presented and applied for the case of One Flow Regime and extended to the case of Several
Flow Regimes.
OPTIMIZATION OF PIPE DIAMETERS WITH OFRM

Review on optimization procedures
Research on optimization procedures to design water distribution networks have been reported
since the 1960s. Karmeli et al. (1968), Schaake and Lai (1969), and Lieng (1971) developed
optimization models for solving branched networks where a given demand pattern was used to
define the flow uniquely in the pipes. Afterwards, important efforts have been made by
Alperovits and Shamir (1977), Quindri et al. (1981), Morgan and Goulter (1985) for solving
looped systems, where infinite number of distributions of flow can meet a specific demand
pattern. Alperovits and Shamir (1977) used fundamental linear programming formulation but
their approach is severely limited to small size systems. The approach by Quindri et al. (1981)
may solve larger systems but only pipes are included in the design and no additional
components are considered (e.g. pumping station, reservoirs, valves). In the approach by
Morgan and Goulter (1985). the cost of components is not included in the objective function and
multiple runs are required to determine the best solution.
Lansey and Mays (1989) proposed a methodology for determining the optimal design of
water distribution systems considering the pipe size, the pumping station and reservoirs. In
addition, the optimal setting for control and pressure-reducing valves can be determined. This
methodology considers the design problem as a non-linear one, where all the typical
components of a network are designed while analyzing multiple alternative discharges. This
procedure still is computationally intensive and requires a large number of iterations to obtain
the solution, especially when applied to large networks.
44 Pipe-size calculation
In view of the concerns sometimes expressed about the use of linear programming when
large networks are designed, other approaches have been developed (Labye, 1966 and 1981)
using the dynamic programming formulation. Later, another approach was developed (Ait Kadi,
1986) to optimize the network layout and the pipe sizes simultaneously for a network composed
by a mainline and secondary parallel branches. This approach (Ait Kadi, 1986) involves two
stages. In the first one, an initial solution is constructed by obtaining the optimum layout of the
mainline, with each pipe of the network having the smallest allowable diameter. In the second
stage, the initial solution is improved by reducing, iteratively, the upstream head and,
consequently, varying the layout of the mainline and pipe sizes simultaneously. This iterative
process is continued until the optimum head giving the minimum cost of the network is reached.
In recent years, several comprehensive reviews on the state-of-the-art in this field were
undertaken (Walsky, 1985b; Goulter, 1987; Walters, 1988). Furthermore, a number of
optimization models were assessed (Walsky et al., 1987). An interesting result of the analysis
by Walsky et al. (1987) was that models produce similar design, both in terms of costs and
hardware components selected. The cost associated with the solutions determined by different
models only varied by 12%. In addition, the most expensive systems were those with increased
level of reliability, imparted by additional storage in the system. Linear programming and
dynamic programming for calculating the optimal pipe diameters in irrigation networks have
also been compared (Di Santo and Petrillo, 1980a,b; Ait Kadi, 1986).
Di Santo and Petrillo (1980a) demonstrated that a network solved by using both LP and the
DP with Bellman's formulation has the same optimal solution. Results obtained by Ait Kadi
(1986) for networks solved by using both LP and the Labye's Iterative Discontinuos Method
(LIDM) were similar.
The above analyses (Di Santo and Petrillo, 1980a,b; Ait Kadi, 1986; Walsky et al., 1987)
indicate that the optimization models utilized were relatively robust and that the optimization is
not sensitive to the technique itself. However, it is useful to improve the ability to design and
analyze water distribution systems. In fact, system design must consider various critical demand
hydrographs to ensure system reliability (Templeman, 1982; Hashimoto et al., 1982). This is
valid for both branched and looped networks where the same daily pattern demand may
correspond to several configurations of flows in the pipes. In addition, it is useful to examine the
extent to which the model techniques and approaches are incorporated into engineering practice.
Better designs have been obtained through optimization models but there has been no
guarantee of global optimality, i.e. on what should be the objective of an optimal design.
Assuming that an optimal design is the one which meets the applied demands at the least cost
(Goulter, 1992), it should incorporate multiple demand conditions, failures in the system
components and reliability.
For considering the reliability of the network, a definition of "failure" is necessary. In
general, a network failure is an event in which a network is not able to provide sufficient flow or
sufficient pressure to meet the demand. Under this definition, failure can occur either if a
component (e.g., a pipe) is undersized, or if the actual demand exceeds the design demand.
These two cases are independent for practical purposes. A considerable effort has been directed
to the reliability question over the last few years.
Relatively little success has been achieved in obtaining comprehensive measures of network
reliability that are computationally feasible and physically realistic (Goulter, 1987; Lansey and
Basnet, 1990).
The measures that give good representation of reliability are computationally impractical,
like the model by Su et al. (1987) where 200 min. of computer time is needed for solving a three
loop example. On the other hand, those approaches that are computationally suitable provide
very poor description of the network performance. Also, empirical solutions may be considered
for improving reliability performance, like the one by Bouchard and Goulter (1991) which
proposed to add valves to the links in order to isolate branches during failure for repair.
Despite the efforts in developing optimization models, they are not widely used in practice.
The main reasons for this are: they are often too complex to be used and designers are not
comfortable with the optimization approaches.
Indeed, Walsky et al. (1987) have shown that, by using optimization models, important
answers to practical problems are possible which can be verified using simulation techniques.
Furthermore, optimization models require less computation time than the classical approaches.
Usually, optimization models are difficult to use. The main reason is that they are developed in
academic environments where the algorithm is much more important than the input-output
interface. In addition, many older engineers have not had opportunity to study formal
optimization techniques.
Considering the effort in developing optimization models, it seems reasonable to assert that
research should be oriented to integrate simulation and optimization models rather than to
develop new optimization algorithms. Techniques for designing optimal pipe diameters for
irrigation networks under several discharge configurations need to be improved and tested.
Furthermore, reliability of systems need to be well defined in order to be included in the
optimization process. Finally, the need for improving the user interface in optimization models
software is important (Walsky et al., 1987).
An interesting commercial program for the calculation of the optimal pipe diameters for
irrigation networks has been developed by CEMAGREF (1990). This program (XERXES-
RENFORS, vers. 5.0) has a user-friendly interface. It computes the discharges with the first
Clment formula, or by adding hydrants' discharges, while Labye's discontinuous method is
used to compute the optimal diameters. Also, optimal pumping station and optimal
reinforcement of the networks may be calculated. This program has interfaces in French and
head losses are computed using the Calmon and Lechapt formula.
In the present publication, an effort has been made to develop and distribute the computer
program (COPAM). It has an English user-friendly interface and all the calculations are easy. It
integrates models for the optimal pipe size computation with models for the analysis of
irrigation systems and allow to present the outputs under form of files and graphics. The
program facilitates full understanding by the user with the capability to verify the results.
When the design of the pumping stations is required the economic aspects are an important
component but the performance analysis of the network is also required. This latter aspect is
covered in Chapter 5.
As far as the regulation of the pumping stations is concerned, the reader is referred to
Irrigation and Drainage Paper No. 44 (page 128) where the subject is treated in detail. However,
the regulation by variable speed pumps is a promising approach concerning energy saving and
has been detailed in Chapter 5 (page 74).
Labye's Iterative Discontinuous Method (LIDM) for OFR

The approach proposed by Labye (1981), called Labye's Iterative Discontinuous Method
(LIDM), for optimizing pipe sizes in an irrigation network is described in this section. This
method is developed in two stages.
In the first stage, an initial solution is constructed giving, for each section k of the network,
the minimum commercial diameter (Dmin) according to the maximum allowable flow velocity
(vmax) in a pipe, when the pipe conveys the calculated discharge (Qk ). The diameter for the
section k is calculated by the relationship:
4Q k
(Dmin)k = (22)
p v max
After knowing the initial diameters, it is possible to calculate the piezometric elevation (Z0 )in
at the upstream end of the network, which satisfies the minimum head (Hj,min) required at the
most unfavorable hydrant (j):
(Z0 )in = Hj,min + ZTj + Yk (23)
0 M j
where Yk are the head losses along the pathway (Mj) connecting the upstream end of the
0 M j
network to the most unfavourable hydrant.
The initial piezometric elevation (Z0 )in relative to the initial diameters solution is therefore
calculated through the relationship (23).
In the second stage, the optimal solution is obtained by iteratively decreasing the upstream
piezometric elevation (Z0 )in until reaching the effectively available upstream piezometric
elevation, Z0 , by selecting, for each iteration, the sections for which an increase in diameter
produces the minimum increase of the network cost. The selection process at each iteration is
carried out as described below.
At any iteration i, the commercial pipe diameters (at most two diameters per section (Labye,
1966)) Ds+1 and Ds (with Ds+1 > Ds) are known. The coefficient:
Ps+1 Ps
s = (24)
J s J s+1
is defined (Fig. 31), where P s [ITL] and Js [m m-1] are, respectively, the cost and the friction
loss per unit length of pipe diameter Ds [m], and Ps+1 [ITL] and Js+1 [m m-1] are, respectively, the
cost and the friction loss per unit length of pipe diameter Ds+1 [m].
FIGURE 31
Characteristic curve of a section
The minimum cost variation, dP, of the elementary scheme FIGURE 32

(SN)* (Figure 32) of any sub-network, (SN), and a section k in Characteristic curve of
series with (SN), for any given variation, dH', of the head H' [m], a section
at the upstream end of (SN)* , is obtained by solving the following
"local" linear programming (Ait Kadi et al., 1990):
min. dP = - s, SN dH - s,k dYk (25)
subject to:
dH + dYk = dH (26)
where dH [m] and dY k [m] are, respectively, the variation of the
head at the upstream end of (SN) and the variation of the friction
loss in section k.
The optimal solution of the equations (25) and (26) is:
dH = dH' and dY k = 0 if s,SN < s,k (27)
dH = 0 and dY k = dH' s,SN > s,k
if (28)
Therefore, the minimum cost variation, dP, of (SN)* can be written as:
dP = - *
dH' with (29)
* = min ( s,SN , s,k ) (30)
Hence, proceeding from any terminal section of the pipe network, the equation (30) can be
used to determine the section that will vary at each iteration. Note that in this process, s,SN of
the assembly of two sections in derivation is equal to (Fig. 32):
s,SN = s,1 + s,2 (31)
whereas, for two sections in series it would be equal to:
s,SN = min ( s,1 , s,2 ) (32)
In the case of a terminal section with a head in excess at its downstream end (Hj > Hj,min), the
value of s,SN to be used in the process is equal to zero as long as the excess head prevails.
The magnitude of dH i , for each iteration i, is determined as:
dHi = min (EHi , Yi , Zi ) (33)
where:
EHi is the minimum value of the excess head prevailing in all the nodes where the head will
change;
Yi is the minimum value of (Yk,i - Y* ) for those sections which will change in diameters, with
Yk,i being the value of the head loss in the section k at iteration i, and Y* is, for this section,
the value of the head loss corresponding to the largest diameter over its entire length if the
section has two diameters, or the next greater diameter if the section has only one diameter.
Note that for those terminal sections with head in excess (H j > Hj,min ), Yi is equal to the
value of this excess (Hj - H j,min ).
Zi is the difference between the upstream piezometric elevation, (Z0 )i , at iteration i, and the
piezometric elevation, Z0 , effectively available at the upstream end of the network.
The iterative process is continued until Z0 is reached, obtaining the optimal solution. An
example is described in Box 3.
BOX 3
Using the LIDM, compute the optimal pipe sizes for the network in figure. The following range of flow
velocity is considered: vmax = 2.5 ms-1, vmin = 0.2 ms -1. The available upstream piezometric elevation
is: Z0 = 165 m a.s.l.
TABLE 3.1
Characteristic of the network
Section Initial Final Discharge Length Land Elevation Hmin
Number Node Node (l s -1) (m) (m a.s.l.) (m)
1 0 1 35 1000 110 20
2 1 2 15 1000 120 30
3 2 3 20 1000 122 30
TABLE 3.2
List of commercial pipes
Thickness Bazin Cost
( mm ) (mm) (m 0.5 ) (ITL m -1)
110 5.3 0.06 14 000
160 7.7 0.06 29 300
200 9.6 0.06 55 000
225 10.8 0.06 65 000
250 11.9 0.06 80 000
315 15.0 0.06 105 000
Darcy's equation was used for calculating the friction slope J [m m-1] in the pipes:
J = u Q2
where
u = 0.000857 (1 + 2 D-0.5 )2 D-5
is the roughness parameter of Bazin, expressed in m 0.5. Q (m 3 s -1) is the discharge flowing into the
section and D (m) is the diameter of the section.
BOX 3 Contd
By applying the Darcy's equation, the following values of u are obtained:
( mm ) u
110 168.343079
100 23.462156
200 7.293085
225 3.945896
250 2.269560
315 0.683712
The following flow velocities are obtained for each section:
Section 1 Section 2 Section 3

Q = 35 ls -1 Q = 15 ls -1 Q = 20 ls -1
( mm ) v ( ms-1) ( mm ) v ( ms-1) ( mm ) v ( ms-1)
110 4.51 110 1.93 110 2.58
160 2.13 160 0.91 160 1.22
200 1.36 200 0.58 200 0.78
225 1.08 225 0.46 225 0.62
250 0.85 250 0.37 250 0.50
315 0.55 315 0.24 315 0.31
The list of the commercial pipe diameters available for each section, according the allowable range
of flow velocities is:
Section
Number (mm ) (mm ) (mm ) (mm ) (mm ) (mm )
1 - 160 200 225 250 315
2 110 160 200 225 250 315
3 - 160 200 225 250 315
Section Number Discharge max min

(l s -1) (mm ) (mm )
1 35 315 160
2 15 315 110
3 20 315 160
The friction losses are:

( mm ) Jx1000 ( mm ) Jx1000 ( mm ) Jx1000
160 28.74 110 37.88 160 9.39
200 8.93 160 5.28 200 2.92
225 4.83 200 1.64 225 1.58
250 2.78 225 0.84 250 0.91
315 0.84 250 0.51 315 0.27
315 0.15
BOX 3 Contd
The values of the coefficient are:
Ps+1 Ps
s =
J s J s+1
(mm) /1000 (mm) /1000 (mm) /1000
160 110 160

1297 469 3974
200 160 200
2439 7064 7469
200
225 225
7304 13278 22370
225
250 250
1287 39769 39411
250
315 315
315 70064
- Initial solution
For each section k of the network, the initial solution is constructed giving the minimum commercial
diameter (Dmin ) according to the maximum allowable flow velocity (vmax) in the pipe.
After knowing the initial diameters, it is possible to calculate the piezometric elevation (Z0)in at the
upstream end of the network, which satisfies the minimum head (Hj,min) required at the most unfavourable
hydrant (j):
Z0-2 = 150 + J110,2 * L 2 + J 160,1 * L 1 = 150 + 37.88 + 28.74 = 216.63 m a.s.l.
Z0-3 = 152 + J160,3 * L 3 + J 160,1 * L 1 = 152 + 9.39 + 28.74 = 190.13 m a.s.l.
The most unfavourable hydrant in this example is number 2 and, therefore, the initial solution is
represented by the upstream piezometric elevation Z0,in = 216.62 m a.s.l. and the initial minimum
diameters:
Section min
Number (mm )
1 160
2 110
3 160
By starting from the node 0 along the path 0 3, it is possible to compute the pressure head on the
node 2 and 3. On the node 2 there is no excess head, while on the node 3 there is excess equal to:
H3 = 216.62 28.74 9.39 152 = 26.49 m
This implies that 3=0. According to the theory of LIDM, we have to compare 1=1297 and 2=469 in
series. The minimum value correspond to 2, so we have to change the diameter of the section 2 from =
110 mm to = 160 mm.
At this stage, we have to decrease Z0,in, so we select the minimum value among:
- Excess head on node where the head change: excess on node 3 = 26.49 m
BOX 3 Contd
- Z0,in Z 0 = 216.62 - 165 = 51.62 m
Y by changing the diameter of the section 2 from = 110 mm to = 160 mm:
Y = 37.88 5.28 = 32.60 m

The minimum value is 26.49 m
- The solution at the first iteration is:
Z0,1 = 216.62 26.49 = 190.13 m a.s.l.

This gives the diameters shown below:
Section min
Number (mm )
1 160
2 160 *
3 160
* By changing the whole diameter on the section 2 we would recover 32.60 m and, in fact, we decrease
the upstream piezometric elevation only 26.49 m. This implies that we have a mixage on section 2
between the diameters = 110 mm and = 160 mm. The lengths of such mixage are:
Note that 1000 / 32.60 = X / 26.49 and
1000 26.49
X = = 813 m
32.60
This means that 813 m of the section 2 needs a diameter = 160 mm and (1000-813) = 187 m needs a
diameter = 110 mm.
At this stage, there is no excess head on the nodes 2 and 3. Therefore, the sections 2 and 3 are in
parallel, and the values of the coefficients for identifying the sections to be changed are
- for the sections 2 and 3: = 2 + 3 = 469 + 3974 = 4453
- for the sections 1: 1=1297
The minimum value is 1, therefore we have to increase the diameter on the section 1 from = 160 mm to
= 200 mm.
BOX 3 Contd
We select a the minimum value for Z0,1 among:
- Excess head on the nodes where the head change: no head variations on the nodes 2 and 3
occur;
- Z0,1 Z 0 = 190.13 -165 = 25.13 m
Y = 28.74 8.93 = 19.81 m
The minimum value is 19.81 m, therefore we have to change diameter on the whole section 1 from =
160 mm to = 200 mm.
Therefore, the solution at the second iteration is:
Z0,2 = 190.13 19.81 = 170.32 m a.s.l.

Section min
Number (mm )
1 200
2 160-110
3 160
At this stage, there is no excess head on the nodes 2 and 3. Therefore, the sections 2 and 3 are
considered in parallel, and the values of the coefficients for identifying the sections to be changed are
The minimum value is 1, so we increase the diameter on the section 1 from = 200 mm to = 225
mm.
We select a minimum value Z0.1 among:
- Excess head on the nodes where the head change: no head variations on the nodes 2 and 3
occur;
- Z0,1 Z 0 = 170.32 -165 = 5.32 m
Y by changing the diameter of the section 1 from 200 mm to 225 mm:
Y = 8.93 4.83 = 4.10 m
The minimum value is 4.10 m, so we change the diameter on the whole section 1 from = 200 mm to
= 225 mm.
Therefore, the solution at the second iteration is:
Z0,2 = 170.32 4.10 = 166.22 m a.s.l.
Section min
Number (mm )
1 225
2 160-110
3 160
BOX 3 Contd
At this stage, there is still no any excess head on the nodes 2 and 3, so sections 2 and 3 are
considered in parallel, and the values of the coefficients for identifying the sections to be changed are
The minimum value is = 2 + 3, so we increase the diameter on the sections 2 and 3.
We select the minimum Z0,2 value among:
- Excess head on the nodes where the head change: no head variation on the nodes 2 and 3 occur;
- Z0,2 Z 0 = 66.22 -165 = 1.22 m

Y3 = 9.38 2.92 = 6.46 m
Y by changing the whole diameter of the section 2 from = 110 mm to = 160 mm:
Y2 = 32.60 26.49 = 6.11 m
= 110 mm in the section 2 to =

The minimum value is 1.22 m, so we change part of the diameter
160 mm and we change part of the diameter = 160 mm in the section 3 into = 200 mm. We will
have mixage on the sections 2 and 3.
The solution at the third iteration is:
Z0,3 = 166.22 1.22 = 165 m a.s.l.
Section min
Number (mm )
1 225
2 160-110
3 200-160
Mixage on the section 3:
Note that 1000 / 6.47 = X / 1.22, so
1000 1.22
X = = 189 m
6.47
This means that 189 m of the section 3 needs a diameter = 200 mm and (1000-189) = 811 m
needs a diameter = 160 mm.
BOX 3 Contd
Mixage on the section 2:
Note that 187 / 6.11 = X / 1.22, so
187 1.22
X = = 37 m
6.11
This means that additional 37 m of the section 2 needs a diameter = 160 mm, so that (813+37)= 850
m needs a diameter = 160 mm, and (1000-850)=150 m needs a diameter = 110 mm.
The final solution is:
The cost of the network is: 126 162 300 ITL
OPTIMIZATION OF PIPE DIAMETERS WITH SFRM

The above described optimization method considers fixed discharges flowing into the pipes of
the network. This is not true for irrigation systems, especially when on-demand delivery
schedules are adopted. In these systems, the discharges are strongly variable in time because of
the extreme variability of the configurations of hydrants operating simultaneously, which
depend on cropping patterns, irrigated areas, meteorological conditions, and farmer's behaviour.
For this reason the LIDM has been extended (Ait Kadi et al., 1990; Lamaddalena, 1997) to take
into account the hydraulic happenings in the network which consists of Several Flow Regimes
(SFRM). Using a computer program, flow regimes were generated using the RGM approach
presented in chapter 3.
Labye's Iterative Discontinuous Method extended for SFR (ELIDM)

The algorithm of the Labye's iterative discontinuous method extended for SFR (Labye, 1981;
Ait Kadi et al., 1990) is divided in two stages. In the first stage, an initial solution is constructed
giving, for each section k of the network, the minimum commercial diameter (Dmin)k according
to the maximum allowable flow velocity (vmax) when the pipe conveys the maximum discharge
(Qmax,k ) for all the configurations. The diameter for the section k is calculated by the
relationship:
4 Q max ,k
(Dmin ) k = (34)
v max
After knowing the initial diameters, it is possible to calculate, for each configuration r, the
piezometric elevation (Z0 )in,r [m a.s.l.] at the upstream end of the network, satisfying the
minimum head, Hj,min [m], required at the most unfavorable hydrant, j:
(Z0 )in,r = Hj,min + ZTj + Yk ,r (35)
0 M j
where Yk ,r are the head losses along the pathway (Mj) connecting the upstream end of the
0 M j
network to the most unfavorable hydrant when the flow regime occurs.
Within all the configurations of discharges, only those requiring an upstream piezometric
elevation, (Z0 )in,r, greater than the effectively available one, Z0 , are considered.
With the first of C possible configurations (r1 ), the initial piezometric elevation (Z0 )in,r1 ,
relative to the initial diameters solution, (Dmin)k , is calculated through the relationship (35). The
optimal solution for r1 is obtained by iteratively decreasing the upstream piezometric elevation.
For each iteration, select sections for which a change in diameter produces the minimum
increase in the network cost. The selection process for each iteration is described below.
At any iteration (I), the commercial pipe diameters are known. There are no more than two
diameters per section (Labye, 1966) with Ds+1 > Ds. The coefficient:
Ps+1 Ps
s = (24)
J s J s+1 FIGURE 33
-1 Elementary scheme
is defined, where Ps [ITL] and Js [m m ] are, respectively, the
cost and the friction loss per unit length of pipe diameter Ds [m],
and Ps+1 [ITL] and Js+1 [m m-1] are, respectively, the cost and the
friction loss per unit length of pipe diameter Ds+1 [m].
The minimum cost variation, dP, of the elementary scheme
(SN)* (Figure 33) of any sub-network, (SN), and a section k in
series with (SN), for any given variation, dH', of the head H' [m],
at the upstream end of (SN)* , is obtained by solving the
following "local" linear programming (Ait Kadi et al.,1990):
min. dP = - s,SN dH - s,k dYk (36)
subject to:
dH + dYk = dH (37)
where dH [m] and dY k [m] are, respectively, the variation of the head at the upstream end of
(SN) and the variation of the friction loss in section k. The optimal solution of the equations
(36) and (37) is:
dH = dH' and dY k = 0 if s,SN < s,k (38)
dH = 0 and dY k = dH' if s,SN > s,k (39)
*
Therefore, the minimum cost variation, dP, of (SN) can be written as:
dP = - *
dH' with (40)
* = min ( s,SN , s,k ) (41)

Hence, proceeding from any terminal section of the pipe network, the equation (40) can be
used to determine the section that will vary at each iteration. Note that in this process (Figure
33), s,SN of the assembly of two sections in derivation is equal to:
s,SN = s,1 + s,2 (42)
whereas, for two sections in series it is equal to:
s,SN = min ( s,1 , s,2 ) (43)
It should be mentioned that in the case of a terminal section with a head in excess at its
downstream end (Hj > Hj,min ), the value of s,SN which is used in the process is equal to zero as
long as the excess head prevails.
The magnitude of dH i,r, for each iteration i and configuration r, is determined as:
dHi,r = min (EHi , Yi , Zi ) (44)
where:
EHi is the minimum value of the excess head prevailing in all the nodes where the head will
change;
Yi is the minimum value of (Yk,i - Y* ) for those sections that change in diameter, with Yk,i
being the value of the head loss in the section k at iteration i, and Y* is, for this section, the
value of the head loss corresponding to the largest diameter over its entire length if the
section has two diameters, or the next greater diameter if the section has only one diameter.
Note that for those terminal sections with head in excess (H j > Hj,min ), Yi is equal to the
value of this excess (Hj - H j,min ).
Zi is the difference between the upstream piezometric elevation, (Z0 )i,r1 , at iteration i when the
flow regime r1 occurs, and the piezometric elevation, Z0 , effectively available at the upstream
end of the network.
The iterative process is continued until Z0 is reached, obtaining the optimal solution for the
examined configuration, r1 . This solution is considered as the initial solution for the next
configuration (r2 ) and the iterative process is repeated. The process is continued until all the
discharge configurations, C, have been considered. Diameters of the sections can only increase
or remain constant for the previous analysed configurations to be satisfied. The final optimal
solution should satisfy all the examined discharge configurations.
A computer program for calculating the optimal pipe size of an irrigation network under both
several discharge configurations and in case of single flow regime has been developed
(Lamaddalena, 1997) and integrated into the COPAM package. The ELIDM has been validated
using the linear programming formulation extended for several flow regimes (Ben Abdellah,
1995; Lamaddalena, 1997). The commercial computer package (LINDO) was adapted for
calculating the optimal solution for the irrigation network with linear programming.
Darcy's equation was used in ELIDM for calculating the friction slope J [m m-1] in the pipes:
J = 0.000857 (1 + 2 D-0.5 ) 2 Q2 D-5 = u Q2 (45)
where is the roughness parameter of Bazin, expressed in m0.5 . The other variables have been
already defined in the text. In Table 4 the values of the roughness Bazin parameters for different
types of pipes are reported. They are compared with the values of the equivalent homogeneous
roughness (mm). Information on the different types of pipes is reported in Annex 2.
TABLE 4
Bazin roughness parameter () for different types of pipes
Equivalent
TYPE OF PIPE homogenous
roughness ( m 0.5 )
( mm )
1- Technically smooth tubes ( glass, brass, drawn copper, resin ) 0 0.02 --
2- Steel pipes
A ) Time degradable coverings
- New pipes, varnished by centrifugation 0.05 --
- Bitumened by immersion 0.1 0.015 0.06
- In current duty with light rust 0.2 0.4 0.10
- With asphalt or tar applied by hands 0.5 0.6 0.16
- With diffused tubercolisation 13 0.23
B ) Non degradable coverings
- Cement applied by centrifugation 0.05 0.15 0.06
3- Welded sheet-pipes
- In good conditions 0.2 0.3 0.10
- In current duty with crusting 0.4 1.0 0.16
4- Nailed sheet-pipes
- 1 line of longitudinal nails 0.3 0.4 0.10
- 2 lines of longitudinal nails 0.6 0.7 0.16
- Idem with crusting Till 3.0 0.30
- 4-6 lines of longitudinal nails 2.0 0.23
- 6 lines of longitudinal nails + 4 transversal 3.0 0.30
- Idem with crusting Till 5.0 0.36
5- Cast iron pipes
- With centrifuged-cemented covering 0.1 0.06
- New, covered internally with bitumen 0.15 0.06
- New, not covered 0.2 0.4 0.10
- With light crusting 0.4 1.0 0.16
- In current duty, partially rusted 1.0 2.0 0.23
- strongly encrusted 3.0 5.0 0.36
6- Cement-pipes
- Asbestos cement 0.1 0.06
- New reinforced concrete, plaster perfectly smooth 0.1 0.15 0.06
- Reinforced concrete with smooth plaster, in work for many years 2.0 0.23
- Tunnels with cement plaster, depending on the degree of finish 2.0 5.0 0.23 0.36
The applicability of the ELIDM in the case of a network with SFR is presented in the next
section.
Applicability FIGURE 34
In this section, the applicability of Costs (106 ITL) of the network in the example, calculated
by using the ELP and the ELIDM for different sets of flow
the Labye iterative discontinuous
regimes
method has been verified through
the linear programming formula-
tion. Both the methods have been
applied to a network under several
flow regimes.
The RGM presented in chapter
3 was previously used for
generating different sets of
random discharge configurations.
These sets of configurations have
been utilized for the calculation of
the optimal diameters using the
ELP and ELIDM programs.
The difference in cost of the
network under study is shown in
Figure 34. It is always less than
2% because in the ELP
formulation all the discharge
configurations are considered as a unique block and, thus, the exact solution is obtained. In the
ELIDM program the configurations are considered one by one, with the final solution for the
configuration ri being the initial solution for the ri+1 configuration. Therefore, diameters of the
sections can only increase or remain constant when the configurations change. For this reason,
the order of the discharge configurations during the calculation has little effect on the optimal
solution.
The differences between the ELP and the ELIDM are small. When several discharge
configurations are considered, the latter method is preferred because it is more flexible and the
computation time is very short (a few seconds for each discharge configuration). Furthermore,
the computer software integrated in the COPAM package is user-friendly and allows the
calculation of networks up to 1000 nodes under more than 1000 flow regimes.
THE INFLUENCE OF THE PUMPING STATION ON THE PIPE SIZE COMPUTATION

The basic relationship between the pumping station at the upstream end of an irrigation system
and the pipe network is that the higher upstream pressure the lower the diameters of the
network, and vice versa. The other important relationship is that the higher the pressure the
higher the costs of the pumping station and the energy.
Through the optimization programs we can compute the variation of the network costs for
different upstream piezometric elevations z. On the other hand, the piezometric elevation can
be translated into energy and pumping station costs. The total cost for the irrigation system
(network, pumping station and energy) is given by the sum of both costs. The total cost is
represented by a curve having a minimum value corresponding to an optimal solution (Figure
35). It is understood that all costs must be referred to the same life span and therefore must be
actualized.
FIGURE 35
Optimization of the pumping station costs
COST
Total Cost 2
C min,2
Total Cost 1
C min,1
Pumping Station + Energy 2
Pipe Network
Pumping Station + Energy 1
Z opt,2 Z opt,1 Z
Notwithstanding the correctness of this mathematical approach, a practical limitation should

be highlighted concerning the uncertainty in the energy cost evolution. In fact, there are
examples of pumping stations that were designed when energy costs were relatively low (before
the petrol crisis) and at that time an optimal upstream piezometric elevation was found that
nowadays it is not economical and causes very high water fees (Figure 35).
According to these considerations, the designer should use the optimization procedures with
a critical view. For this reason the automatic optimization of the upstream piezometric elevation
has not been included in COPAM. Nevertheless COPAM gives the possibility to compute the
optimal cost of the network for several piezometric elevations but each one has to be computed
one by one. The designer can use these results to evaluate different pumping station alternatives
taking into account that related costs can only be estimated, in a reliable way, under present
conditions.
THE USE OF COPAM FOR PIPE SIZE COMPUTATION

The use of the optimization program integrated in the COPAM package is easy. After
completing the input file, as described in the chapter 3, by clicking on the Optimization button
(Figure 36) the page control in Figure 37 will appear on the screen.
Three alternatives are available within the Options Tab control, concerning the flow
regimes (Figure 37): single flow regimes, several flow regimes directly generated by the
program, and several flow regimes previously generated and stored in a file to be read. The last
option allows network optimization also in the case of rotation delivery schedule. In fact, the
flow regimes are computed according to the planned irrigation schedule and stored in a file that
will be read by the program. Concerning the option single flow regimes, the program allows
the computation of discharges by using the 1st or 2nd Clment model.
FIGURE 36
Pipe size computation program
FIGURE 37
Optimization program: Options Tab control
FIGURE 38
Optimization program: MixageTab control
The name of the output file is typed in the appropriate edit box. The extension .opt is
automatically assigned to this output file. All information concerning the computation is stored
in that file (see the example of the output file in Annex 3). Finally, two additional design
options are available in the program: new design and rehabilitation. In the first case, the
program computes the optimal pipe size diameters starting from an initial solution obtained by
using the smallest diameters respecting the maximum flow velocities constraints (as explained
in the chapter 4). In the case of rehabilitation, the initial solution is given by the actual diameters
of each section of the network.
The program, within the Mixage tab control (Figure 38), gives the possibility to select one
diameter for each section or to consider the mixage with two diameters for each section where
required. From a practical point of view, one diameter should be selected in order to avoid
possible mistakes during the construction phase.
In the Data Tab control (Figure 39), all the parameters related to the 1st or 2nd Clment
formula are introduced. These parameters were illustrated in chapter 3. The upstream
piezometric elevation and the minimum pressure head required at the hydrants are typed in the
appropriate edit boxes.
The software allows also the computation of the optimal pipe size diameters when the
minimum pressure head at the hydrants is not a constant. In this case, the radio button variable
is selected and, in the last column of the input file, the appropriate minimum pressure heads
(Hmin) are typed in correspondence of each hydrant (Figure 40).
FIGURE 39
Optimization program: Data Tab control
FIGURE 40
Input file: Edit network layout
FIGURE 41
When Several flow regime is selected within the Options Tab control (Figure 41), the
final solution or the analytical solution for each flow regime may be printed on the output file
by clicking on the appropriate radio button. The name of the output file is typed in the edit box.
In this case the final solution is recommended, especially when many flow regimes are
generated and the output file might be too long.
When the option Several flow regimes is selected, only the upstream piezometric elevation
and the minimum pressure head are typed in the appropriate edit boxes in the Data Tab
control (Figure 42).
In the case of flow regimes generated directly by the program (Option: Several-random
generation in Figure 41), the name of the file where the flow regimes are stored needs to be
specified in the appropriate box (Figure 41). Also, the maximum upstream discharge and the
number of flow regimes to be generated (remember that this number must be multiple of 10) are
specified in the appropriate boxes (Figure 41).
These last two inputs are not required when the flow regimes are read from a file (Option:
Several-read from file, in Figure 43). Only the name of the file for reading the flow regimes is
specified in the appropriate edit box.
Sensitivity analysis of the ELIDM
The actual number of flow regimes in an irrigation network is large and the optimal solution
varies according to the number of flow regimes considered (see Figure 34). Consequently, a
sensitivity analysis is required to examine the variation of the optimal solution for several sets
of flow regimes.
FIGURE 42
Optimization program: Data Tab control
FIGURE 43
FIGURE 44
Layout of the network
The ELIDM was used for the irrigation network in Figure 44. Fifty different sets of flow
regimes, corresponding to 50 l s-1, were randomly generated by the RGM (described in the
chapter 3), and used for investigating the sensitivity of the optimal solution versus the number
of flow regimes. The input file of the network is reported in the Annex 3, as well as an example
of the output file for one random flow regime.
Initially, ten random discharge configurations were generated for calculating the cost P1,10.
Subsequently, another ten random discharge configurations were generated for calculating the
cost P 1,10 , and so on for every P 1,10 up to a total of 50 sets of flow regimes. This procedure was
repeated for an increased number of configurations in each set, from 20 up to 100. Therefore, if
C is the number of configurations in each set, varying from 10, 20, ...., 100, the average cost for
each C is:
50
Pi, C
PC = i =1 (46)
50
In Figure 45, the average cost of the network and the relative confidence interval at 95% of
probability is reported for each set of discharge configurations.
Figure 45 shows that when a larger number of flow regimes is considered a more stable
solution is obtained. When the number of flow regimes increases, the designed network is able
to satisfy a larger population of possible configurations. Figure 45 also shows that when a larger
number of flow regimes is considered, the optimal solution obtained for different repetitions of
the same number of flow regimes falls within a narrower range of confidence interval. The
results indicate that a minimum of 50 discharge configurations (C = 50) is needed for
calculating the network in the example.
FIGURE 45
Cost of the network as a function of the number of configurations. The average costs and the
confidence interval concern 50 sets of discharge configurations.
220
215 10 20 30 40 50 60 70 80 90
COST (ITL10E6)
100
210
205
200
AVERAGE COST OF THE
195 NETWORK
CONFIDENCE INTERVAL AT 95%
OF PROBABILITY
190
185
10 20 30 40 50 60 70 80 90 100
NUMBER OF CONFIGURATIONS
The sensitivity analysis (Figure 45) shows how the cost of the network depends upon the
number of flow regimes. Thus, the need to establish criteria for selecting the most suitable
solution arises. We can define it as a Satisfactory Solution instead of an Optimal Solution.
This satisfactory solution is related to performance criteria. In the next chapter, models for the
analysis of on-demand irrigation systems and performance indicators are presented and applied.

Computations of Flows For On Demand Irrigation Systems

Uploaded by

Copyright:

Available Formats

Computations of Flows For On Demand Irrigation Systems

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Computations of Flows For On Demand Irrigation Systems

Uploaded by

Copyright:

Available Formats

Performance analysis of on-demand pressurized irrigation systems 15

ONE FLOW REGIME M ODELS (OFRM)

The first Clment model

Considering hydrants with the same discharge, the total TABLE 1

Determination of the specific continuous discharge

Discharge at the hydrants

In Italy the range of hydrants corresponds to the following discharges:

1) Number of hydrants simultaneously operating downstream the section 0-1:

N0-1 = Rp +U(Pq ) Rp (1- p) = 19 0.147 + 1.645 19 0.147 (1- 0.147) = 5.33

Then Figure 13 will appear.

Preparation of the input file

The Second Clment's model

F(u' ) = PSAT R p (1 - p) (20)

BOX 2: APPLICATION OF THE SECONDCLMENT MODEL FOR THE COMPUTATION OF FLOWS IN AN

1) Number of hydrants simultaneously operating downstream the section 0-1:

N0-1 = Rp+u' Rp (1- p) =

N1-2 = Rp+u' Rp (1- p) =

N2-3 = Rp+u' Rp (1- p) =

and so on for the calculation of the discharges in the other sections.

DISCHARGE [l/s] (1st Clement' formula)

SEVERAL FLOW REGIMES M ODEL (SFR M ODEL)

Model for generating random discharges

-1 -1 Nodes without hydrant

The RG Model may be used for two different purposes:

OPTIMIZATION OF PIPE DIAMETERS WITH OFRM

Labye's Iterative Discontinuous Method (LIDM) for OFR

The minimum cost variation, dP, of the elementary scheme FIGURE 32

By applying the Darcy's equation, the following values of u are obtained:

The following flow velocities are obtained for each section:

Section 1 Section 2 Section 3

Section Number Discharge max min

The friction losses are:

Section 1 Section 2 Section 3

Section 1 Section 2 Section 3

(mm) /1000 (mm) /1000 (mm) /1000

160 110 160

Z0-2 = 150 + J110,2 * L 2 + J 160,1 * L 1 = 150 + 37.88 + 28.74 = 216.63 m a.s.l.

Z0-3 = 152 + J160,3 * L 3 + J 160,1 * L 1 = 152 + 9.39 + 28.74 = 190.13 m a.s.l.

H3 = 216.62 28.74 9.39 152 = 26.49 m

Y by changing the diameter of the section 2 from = 110 mm to = 160 mm:

Y = 37.88 5.28 = 32.60 m

- The solution at the first iteration is:

Z0,1 = 216.62 26.49 = 190.13 m a.s.l.

Note that 1000 / 32.60 = X / 26.49 and

Y by changing the diameter of the section 1 from = 160 mm to = 200 mm:

Y = 28.74 8.93 = 19.81 m

Therefore, the solution at the second iteration is:

Z0,2 = 190.13 19.81 = 170.32 m a.s.l.

Y by changing the diameter of the section 1 from 200 mm to 225 mm:

Y = 8.93 4.83 = 4.10 m

Y by changing the diameter of the section 3 from = 160 mm to = 200 mm:

Y2 = 32.60 26.49 = 6.11 m

= 110 mm in the section 2 to =

Mixage on the section 3:

Note that 1000 / 6.47 = X / 1.22, so

Note that 187 / 6.11 = X / 1.22, so

The final solution is:

The cost of the network is: 126 162 300 ITL