This action might not be possible to undo. Are you sure you want to continue?

Ioan Sârbu

1

Abstract

There are three methods for analyzing flow and pressure distribution in looped water

distribution networks (the loop method, the node method, the element method) taking into

consideration hydraulic parameters chosen as unknown. For all these methods the non-

linear system of equations can be solved by iterative procedures.

The paper presents a different approach to this problem by using the method of varia-

tional formulations for hydraulic analysis of water distribution networks. This method has

the advantage that it uses a specialized optimization algorithm which minimizes directly an

objective multivariable function without constraints, implemented in a computer program.

The paper compares developed method to the classic Hardy–Cross method. This shows

the good performance of the new method.

1 Introduction

There are three methods for analyzing flow and pressure distribution in looped water

distribution networks (the loop method, the node method, the element method) taking into

consideration hydraulic parameters chosen as unknown.

For all these methods the non-linear system of equations can be solved by iterative

procedures: Hardy–Cross method (Abramov, 1976; Divénot, 1980) or Newton–Raphson

method (Chandrashekar and Stewart, 1975; Gofman and Rodeh, 1981; Sârbu, 1987;

Wood and Rayes, 1981).

This paper shows a different approach to this problem by using the method of variational

formulations for functional analysis of water distribution networks. This method has the

advantage that it uses a specialized optimization algorithm which minimizes directly an

objective multivariable function without constraints, implemented in a computer program.

2 The fundamentals of hydraulic computation

In the case of a complex topology for a looped network, with reservoirs and pumps at the

nodes, the total number of independent loops (closed-loops, possibly containing pumps

mounted in the pipes, and pseudoloops) M is given by the following formula:

M T N N

RP

· − + (1)

in which: T is the number of pipes in network; N – number of nodes in network; N

RP

–

number of reservoirs and pump at the nodes, equal to the number of nodes with know

hydraulic grade.

_________________________________

1

Professor, Department of Building Services, “ Politehnica” University of Timisoara, Romania

Each open-loop (pseudoloop) makes the connection between a node with a known

piezometric head (reservoir) or with a determined relation discharge – piezometric head

(pump station), and another node with a known piezometric head or a determined relation

discharge – piezometric head.

In classical analysis of looped networks in permanent water flow regime, fundamental

equations of the computing model express:

– discharge continuity at nodes:

Q q j N N

ij j RP

i

i j

N

+ · · −

·

≠

∑

0 1

1

( ,..., ) (2)

in which: Q

ij

is the discharge through pipe ij, with the sign (+) when entering node j and (–)

when leaving it; q

j

– concentrated discharge at node j with the sign (+) for node inflow and

(–) for node outflow.

– energy conservation in loops:

ε

ij ij m

ij m

ij

T

h f m M − · ·

∈

·

∑

0 1

1

( ,..., ) (3)

in which: h

ij

is the head loss of the pipe ij; ε

ij

– orientation of flow through the pipe, having

the values (+1) or (–1) as the water flow sense is the same or opposite to the path sense

of the loop m, and (0) value if ij∉m; f

m

– pressure head introduced by the potential

elements of the loop m, given by the relations:

• simple closed-loops:

f

m

· 0 (4)

• closed-loops containing pumps mounted in the pipes:

f H

m ij p ij

ij m

ij

T

·

∈

·

∑

ε

,

1

(5)

• open-loops with pumps and/or reservoirs at nodes:

f Z Z

m I E

· − (6)

where: Z

I

, Z

E

are piezometric heads at pressure devices at the entrance or exit from the

loop; H

p,ij

– the pumping head of the pump mounted in the pipe ij, for the discharge Q

ij

,

approximated by parabolic interpolation on the pump curve given by points:

H A Q B Q C

p ij ij ij ,

· + +

2

(7)

the coefficients A, B, C can be determined from three points of operating data.

The head loss is given by the Darcy–Weisbach functional relation:

h

g

L

D

Q

ij ij

ij

ij

r

ij

·

8

2

2

π

λ (8)

in which: g is the gravitational acceleration; λ

ij

– friction factor of pipe ij which can be

calculated using the Colebrook-White formula; D

ij

, L

ij

– diameter and the length of pipe ij;

r – exponent having the value 5.0.

Equation (8) is difficult to use in the case of pipe networks and therefore it is convenient to

write it in the following general form:

h R Q

ij ij ij

·

β

(9)

where R

ij

is the hydraulic resistance of pipe ij, having the succeeding relation:

R

g

L

D

ij

ij

ij

r

·

8

2

λ

π

*

(10)

The variation of hydraulic parameters λ

∗

and β has been determined for different pipe

materials and water temperatures, using a computer program (Sârbu, 1997).

Formula (9) can be written as follows:

h Z Z R Q Q

ij i j ij ij ij

· − ·

− β 1

(11)

or:

( )

Q R h R Z Z Z Z

ij ij ij ij i j ij i j ij

· · − + − +

− −

−

1 1 1

1

β β β

β

β

Π Π (12)

in which: Z

i

and Z

j

are the piezometric heads at nodes i and j; Π

ij

– active pressure

introduced by the intermediate pump on the pipe ij.

If it is associated to each loop m a circulation flow ∆Q

m

and if is choosed initial flow

distribution

ij

Q

( ) 0

which have to satisfy equation (2), then it can be written:

Q Q Q ij T

ij ij ij m

ij m

m

M

· + ·

∈

·

∑

(o)

( ,..., ) ε ∆ 1

1

(13)

and for simple loops (f

m

= 0) the system (2), (3), (9), (13) is equivalent to the following:

ε ε ε

β

ij ij ij ij m

ij m

m

M

m

ij ij m

ij m

m

M

R Q Q Q Q m M

(o)

( )

(o)

( ,..., ) +

¸

¸

_

,

+ · ·

∈

·

∈

·

−

∑ ∑ ∑

∆ ∆

1 1

1

0 1 (14)

Substituting equation (12) in equation (2) one gets a system of N–N

RP

equations at nodes

with N–N

RP

unknown:

( )

R Z Z Z Z q j N N

ij

i j

i

N

i j ij i j ij j RP

−

≠

·

−

∑

− + − + + · · −

1

1

1

0 1

β

β

β

Π Π ( ,..., ) (15)

Specific consumption of energy for water distribution w

sd

, in kWh/m

3

, is obtained by

referring the hydraulic power dissipated in pipes to the sum of node discharges:

w

R Q

q

sd

ij ij

ij

T

j

j

q

N

·

+

·

·

<

∑

∑

000272

1

1

1

0

.

'

β

(16)

where q

j

'

is the outflow at the node j.

3 The variational formulation of the problem

The hydraulic analysis of looped networks can be achieved, according to the element

method, by using a conditioned optimization model named "the content model" (Coolins

and Cooper, 1978; Coolins, 1979).

The discharges at equilibrium, in loop method, can be determined by using the criterion of

minimization of the energy content in time (power) for the whole of the network. For

networks containing potential elements (reservoirs and pumps at nodes, pumps mounted

in the pipes) objective function can be expressed as follow:

( )

F Z Z dQ H dQ Z dq

e i j ij

Q

ij

ij

T

p ij ij

Q

ij

ij

T

j j

q

j

j

N

RP

· −

¸

1

]

1

1

−

¸

¸

_

,

−

¸

¸

_

,

→

∫

∑

∫

∑

∫

∑

· ·

∗

·

0

1

0

1

0

1

,

min (17)

where Z

j

∗

is the piezometric head of node source j.

Equation (17) is subject to continuity constraints (2) and to constraints of non-negativity for

discharges (Q

ij

≥ 0).

The functionals in the first term of equation (17) represent energy loss through the network

pipes when the network carries the discharges that satisfy continuity requirements, and the

functionals in the second and third terms represent the external power input to the system.

Thus substituting equations (11) and (13) in the objective function (17) as well as

considering equation (7) after one calculates all the integrals one eliminate all the

constraints. In order to determine the discharges at which hydraulic equilibrium network

occurs, we have only to find the minimum of a function with M variables (∆Q

m

) without

constraints:

F R Q Q A Q Q

B Q Q C Q Q Z q

e ij ij ij m

ij m

m

M

ij

T

ij ij ij m

ij m

m

M

ij

T

ij ij ij m

ij m

m

M

ij ij ij m

ij m

m

M

j j

j

N

RP

·

+

+ − + +

¸

¸

+ + +

_

,

− →

∈

·

+

· ∈

·

·

∈

·

∈

·

∗

∑ ∑ ∑ ∑

∑ ∑ ∑

1

1

1

3

1

2

1

1

1

1

3

1

1

2

1

1

β

ε ε

ε ε

β

(o) (o)

(o) (o)

=

+

∆ ∆

∆ ∆ min

(18)

By using the extremum requirements in (18), ∂F

e

/∂∆Q

m

= 0 (m = 1,...,M), one gets the

system (14) for a simple network in the classical formulation of the problem.

This formulation is advantageous because it can make use of a minimization algorithm for

function (18) like the “ conjugate gradient algorithm” recommended by literature (Sârbu,

1994; Todini and Pilati, 1987). The objective function form is independent from the initial

solution, for a certain loop system. Process convergence depends of loop configuration

mode and the minimum value that remains unchanged. Function has a much favorable

form if common loop part is shorter.

The discharges through pipes Q

ij

are obtained using equation (13), in which are introduced

the circulation flow ∆Q

m

obtained as those values that minimize the objective function.

After the head losses have been calculated with equation (9), the piezometric heads Z

j

can

be determined starting from a node of known piezometric head. Then the residual

pressure head H

j

at the node j is calculated from the relation:

H Z ZT

j j j

· − (19)

where ZT

j

is the elevation head at the node j.

One can find a variational formulation for the hydraulic analysis of looped networks also in

node method, using the following objective function:

F Q dh q dZ

e ij ij

h

ij

ij

T

j j

Z

j

j

N

·

¸

¸

_

,

−

¸

¸

_

,

→

∫

∑

∫

∑

· ·

0

1

0

1

min (20)

and the energy conservation in loops constraints (3).

After substituting the functional relation (12) in (20) and after the integrals have been

calculated the constraints can be eliminated and the problem can be simplified to the

finding of the minimum of a function with N–N

RP

variables (Z

i

, Z

j

) without constraints:

F R Z Z q Z

e ij i j ij

ij

T

j j

j

N

·

+

− + − →

−

·

+

·

∑ ∑

β

β

β

β

β

1

1

1

1

1

Π min, (21)

which can be achieved by using the conjugate gradient algorithm.

Using the extremum requirements ∂F

e

/∂Z

j

= 0 (j = 1,...,N–N

RP

) one gets the system of node

equations (15).

In order to determine an initial approximation of the piezometric head at nodes one must

solve the following linear associated system:

R Z Z q j N N

ij j

i j

i

N

i ij j RP

−

≠

·

∑

− − · · −

1

1

1 ( ) ( ,..., )

(o) (o)

Π

β

(22)

Once the piezometric head at nodes have been determined one can calculate the residual

pressure head using equation (19) and afterwards the discharges through pipes with

equation (12) as well as other hydraulic parameters of the network.

Two computer programs ACIREV and ANOREV were elaborated (Sârbu, 1997), based on

the optimization models developed above. It are realized in the FORTRAN 5.1 program-

ming language and implemented on IBM–PC compatible computers.

4 Numerical application

The looped distribution network with the topology from Figure 1 is considered. It is made of

cast iron and is supplied with a discharge of 0.50 m

3

/s. The following data are known: pipe

length L

ij

, in m, pipe diameter D

ij

, in m, elevation head ZT

j

, in m, industrial concen-trated

discharges in nodes q

j

, in m

3

/s, piezometric head at the “ critical node” Z

1

= 124 m, and the

exponent β = 1.936.

It is required to determine the pumping head, discharges and pressures distribution using

the classic HARDY–CROSS procedure and the two optimization models (ACIREV,

ANOREV) developed above. Results of the numerical solution performed by means of an

IBM–AT 586 computer, referring to the hydraulic characteristics of the pipes and nodes are

presented in tables 1 and 2.

Fig.1 Schema of the analyzed distribution network

Table 1 shows the discharges and head losses through pipes established by using the

three mentioned models of computation (the iterative tolerance imposed is 10

-5

). It can be

seen that the results are very close.The difference between the discharges obtained with

HARDY–CROSS and these given by ACIREV vary between 0.08 % (pipe 3–2) to 1.8 % (pipe

3–7), and the difference between discharges obtained with HARDY–CROSS and ANOREV

varies from 0 % (pipe 9–3) to 2.7% (pipe 9–8). Specific consumption of energy for water

distribution is 0.00705 kWh/m

3

, for all three models of computation used.

Table 1. The discharges and head losses trough pipes

Pipe Computation model

i −j HARDY-CROSS ACIREV ANOREV

Q [m

3

/s] h [m] Q [m

3

/s] h [m] Q [m

3

/s] h [m]

0 1 2 3 4 5 6

2−1 0.01209 3.949 0.01219 4.007 0.01204 3.915

3−2 0.03608 0.890 0.03616 0.894 0.03605 0.889

4−3 0.06917 1.203 0.06902 1.198 0.06916 1.203

13−4 0.11373 1.359 0.11358 1.356 0.11371 1.356

6−5 0.03771 1.443 0.03761 1.437 0.03776 1.447

7−6 0.08199 0.906 0.08174 0.900 0.08206 0.907

8−7 0.13520 0.521 0.13496 0.519 0.13531 0.521

9−8 0.17547 0.262 0.17542 0.261 0.18034 0.276

13−9 0.25299 1.067 0.25317 1.069 0.25266 1.065

11−10 0.01597 2.518 0.01613 2.568 0.01610 2.558

12−11 0.04802 2.160 0.04799 2.158 0.04806 2.163

13−12 0.09508 1.123 0.09505 1.123 0.09501 1.122

5−1 0.01656 3.127 0.01647 3.094 0.01661 3.147

7−2 0.02626 1.588 0.02628 1.590 0.02625 1.587

9−3 0.02579 1.495 0.02602 1.521 0.02579 1.495

6−10 0.01666 3.092 0.01649 3.034 0.01661 3.075

8−11 0.01332 1.958 0.01351 2.014 0.01327 1.945

Table 2 presents the values for the piezometric head Z

j

and the residual pressure head H

j

at nodes determined by using the classic procedure and the two new models of

computation. The piezometric head at the node 13 has the following values: 131.402 m,

131.455 m and 131.363 m which give a residual pressure head of 29.402 m, 29.455 m and

29.363 m that is sufficient for the supply water to the consumers. The divergence of

piezometric line on network contour is 0.122 m for HARDY–CROSS, 0.109 m for ACIREV

and only 0.001 m for ANOREV.

Table 2. The piezometric head and the residual pressure head at nodes

Node Computation model

j HARDY–CROSS ACIREV ANOREV

Z

j

[m] H

j

[m] Z

j

[m] H

j

[m] Z

j

[m] H

j

[m]

0 1 2 3 4 5 6

1 124.000 24.000 124.000 24.000 124.000 24.000

2 127.949 27.449 128.007 27.507 127.915 27.415

3 128.839 27.839 128.901 27.901 128.804 27.804

4 130.042 29.042 130.099 29.099 130.007 29.007

5 127.127 26.627 127.094 26.594 127.147 26.647

6 128.571 28.571 128.530 28.530 128.594 28.594

7 129.537 28.537 129.597 28.597 129.502 28.502

8 130.058 28.558 130.116 28.616 130.023 28.523

9 130.335 28.335 130.386 28.386 130.299 28.299

10 125.478 25.478 125.497 25.497 125.520 25.520

11 128.118 27.118 128.175 27.175 128.078 27.078

12 130.278 28.778 130.332 28.832 130.241 28.741

13 131.402 29.402 131.455 29.455 131.363 29363

5 Conclusions

The mathematical model expressed by the objective functions (18) and (21) constitutes a

new way of hydraulic analysis of complex looped networks based on unconditioned

optimization techniques.

The new method replaces the solving of the non-linear system of equations (2), (3), (9)

with the direct minimization of a multivariable function, without constraints that express the

energy consumption across the network.

The computer programs ACIREV and ANOREV include this particular aspect and contain

the conjugate gradient algorithm, which give it efficiency especially in functional analyzing

of complex distribution networks. This new method is computationally more efficient and

consequently helps the designer to get the best design of water distribution systems with

fewer efforts.

References

1. ABRAMOV N. N., Rascet vodoprovodnîi setei, Stroizdat, Moskva, 1976.

2. CHANDRASHEKAR M. and STEWART K., Sparsity oriented analysis of large pipe

networks, Journal of the Hydraulics Division, ASCE, nr. HY4, 1975.

3. COOLINS M. and COOPER L., Solving the pipe network analysis problem using

optimization techniques, Management Science, nr. 7, 1978.

4. COLLINS M. et al., Multiple operating points in complex pump network, Journal of the

Hydraulics Division, ASCE, nr. HY3, 1979.

5. DIVÉNOT A., Une nouvelle méthode de calcul des réseaux maillés, La Houille Blanche,

nr.6, 1980.

6. GOFMAN E. and RODEH M., Loop equation with unknown pipe characteristics, Journal

of the Hydraulics Division, ASCE, nr. HY9, 1981.

7. LEKANE T., Modele de calcul de l'écoulement en régim permanent dans réseau d'eau

maillés, Journal de Recherches Hydrauliques, nr.2, 1979.

8. SÂRBU I., Model de calcul al regimului hidraulic în reþele complexe de distribuþie a apei,

Rev. Hidrotehnica, nr.8, 1987.

9. SÂRBU I., Energetical Optimization of Water Distribution Systems, Ed. Academiei

Române, Bucureº ti, 1997.

10. SÂRBU I., Numerical and Optimizing Methods in Building Equipment Design, Ed.

Tehnicã, Bucureº ti, 1994.

11. SIMA V. and VARGA A., Practica optimizãrii asistatã de calculator, Ed. Tehnicã,

Bucureº ti, 1986.

12. TODINI E. and PILATI S., A Gradient Method for the solution of looped Pipe Networks,

Int. Conf. on Computer Applications in Water Supply and Distribution, V(1), 1987.

13. WOOD D. and RAYES A., Reliability of algorithms for pipe network analyses, Journal

of the Hydraulics Division, ASCE, nr. HY10, 1981.

– energy conservation in loops: ij ∈m ij =1 ∑ε T ij hij − f m = 0 ( m = 1. f m – pressure head introduced by the potential elements of the loop m. Dij. ε ij – orientation of flow through the pipe. λij – friction factor of pipe ij which can be calculated using the Colebrook-White formula. The head loss is given by the Darcy–Weisbach functional relation: L 8 2 hij = 2 λij ij Qij r π g Dij (8) in which: g is the gravitational acceleration. qj – concentrated discharge at node j with the sign (+) for node inflow and (–) for node outflow..ij = A Qij + B Qij + C (7) the coefficients A...ij – the pumping head of the pump mounted in the pipe ij. ZE are piezometric heads at pressure devices at the entrance or exit from the loop.Each open-loop (pseudoloop) makes the connection between a node with a known piezometric head (reservoir) or with a determined relation discharge – piezometric head (pump station). M ) (3) in which: hij is the head loss of the pipe ij. having the values (+1) or (–1) as the water flow sense is the same or opposite to the path sense of the loop m. Equation (8) is difficult to use in the case of pipe networks and therefore it is convenient to write it in the following general form: β hij = Rij Qij (9) where Rij is the hydraulic resistance of pipe ij.. for the discharge Qij . and (0) value if ij∉m.. and another node with a known piezometric head or a determined relation discharge – piezometric head. with the sign (+) when entering node j and (–) when leaving it. ij ij∈m ij =1 T (5) • open-loops with pumps and/or reservoirs at nodes: f m = ZI − ZE (6) where: ZI . N − N RP ) (2) in which: Qij is the discharge through pipe ij. B. Lij – diameter and the length of pipe ij. having the succeeding relation: .. In classical analysis of looped networks in permanent water flow regime. approximated by parabolic interpolation on the pump curve given by points: 2 H p . Hp.0.. r – exponent having the value 5. given by the relations: • simple closed-loops: fm = 0 (4) • closed-loops containing pumps mounted in the pipes: f m = ∑ ε ij H p . C can be determined from three points of operating data.. fundamental equations of the computing model express: – discharge continuity at nodes: ∑Q i =1 i≠ j N ij + qj = 0 ( j = 1.

. in kWh/m 3.. then it can be written: (o) Qij = Qij + ∑ εij ∆Qm ij ∈m m =1 M ( Z i − Z j + Π ij Z i − Z j + Π ij ) 1− β β (12) ( ij = 1..00272 ∑R ij = 1 T ij Qij ' j β +1 ∑q j =1 q <0 N (16) where q 'j is the outflow at the node j. (9). 1979). Coolins.. Rij = Formula (9) can be written as follows: β −1 hij = Z i − Z j = Rij Qij Qij or: Qij = Rij − 1 β 1 β ij (11) h = Rij − 1 β in which: Zi and Zj are the piezometric heads at nodes i and j.. using a computer program (Sârbu.8 λ* Lij (10) r π2 g Dij The variation of hydraulic parameters λ∗ and β has been determined for different pipe materials and water temperatures. by using a conditioned optimization model named "the content model" (Coolins and Cooper. (13) is equivalent to the following: β −1 M M (14) ∑) εij Rij Qij(o) + ij∑ εij ∆Qm Qij(o) + ∑ εij ∆Qm = 0 (m = 1. according to the element method. Π ij – active pressure introduced by the intermediate pump on the pipe ij. M ) (m ∈m ij ∈m m =1 m =1 Substituting equation (12) in equation (2) one gets a system of N–NRP equations at nodes with N–NRP unknown: ∑ R (Z N − β ij i≠ j i =1 1 i − Z j + Π ij Zi − Z j + Π ij ) 1 −β β + qj = 0 ( j = 1. 1978. N − N RP ) (15) Specific consumption of energy for water distribution wsd. (3). 1997). ... T ) (13) and for simple loops (f m = 0) the system (2).... If it is associated to each loop m a circulation flow ∆Q m and if is choosed initial flow (0) distribution Qij which have to satisfy equation (2).. 3 The variational formulation of the problem The hydraulic analysis of looped networks can be achieved.. is obtained by referring the hydraulic power dissipated in pipes to the sum of node discharges: wsd = 0.

The functionals in the first term of equation (17) represent energy loss through the network pipes when the network carries the discharges that satisfy continuity requirements. . pumps mounted in the pipes) objective function can be expressed as follow: Q T Qij N RP q j T ij Fe = ∑ ∫ Zi − Z j dQij − ∑ ∫ H p . in loop method.. for a certain loop system. Then the residual pressure head Hj at the node j is calculated from the relation: H j = Z j − ZTj (19) where ZTj is the elevation head at the node j. and the functionals in the second and third terms represent the external power input to the system.. After the head losses have been calculated with equation (9).ij dQij − ∑ ∫ Z ∗ dq j → min (17) j j =1 0 ij =1 0 ij =1 0 ∗ where Z j is the piezometric head of node source j. The discharges through pipes Qij are obtained using equation (13). Process convergence depends of loop configuration mode and the minimum value that remains unchanged. the piezometric heads Zj can be determined starting from a node of known piezometric head.M). This formulation is advantageous because it can make use of a minimization algorithm for function (18) like the “conjugate gradient algorithm” recommended by literature (Sârbu. can be determined by using the criterion of minimization of the energy content in time (power) for the whole of the network. Thus substituting equations (11) and (13) in the objective function (17) as well as considering equation (7) after one calculates all the integrals one eliminate all the constraints. Todini and Pilati. ∂Fe /∂∆Q m = 0 (m = 1.. 1994. in which are introduced the circulation flow ∆Q m obtained as those values that minimize the objective function.. For networks containing potential elements (reservoirs and pumps at nodes. ( ) Equation (17) is subject to continuity constraints (2) and to constraints of non-negativity for discharges (Qij ≥ 0).The discharges at equilibrium. Function has a much favorable form if common loop part is shorter. one gets the system (14) for a simple network in the classical formulation of the problem. 1987). we have only to find the minimum of a function with M variables (∆Q m) without constraints: β +1 3 T M T M 1 1 Fe = ∑ Rij Qij(o) + ij∑ εij ∆Qm − ∑ 3 Aij Qij(o) + ij∑ εij ∆Qm + β + 1 ij =1 ∈m ij =1 ∈m m =1 m =1 (18) 2 N M M RP 1 (o) (o) + Bij Qij + ∑ εij ∆Qm + Cij Qij + ∑ εij ∆Qm − ∑ Z ∗ q j → min j 2 ij ∈m ij ∈m j =1 m =1 m =1 By using the extremum requirements in (18). The objective function form is independent from the initial solution. In order to determine the discharges at which hydraulic equilibrium network occurs.

Fe = T − 1 (21) Using the extremum requirements ∂Fe/∂Zj = 0 (j = 1. Two computer programs ACIREV and ANOREV were elaborated (Sârbu.. It are realized in the FORTRAN 5..50 m3 /s. referring to the hydraulic characteristics of the pipes and nodes are presented in tables 1 and 2.. It is made of cast iron and is supplied with a discharge of 0. It is required to determine the pumping head. Results of the numerical solution performed by means of an IBM–AT 586 computer.N–NRP) one gets the system of node equations (15)..1 programming language and implemented on IBM–PC compatible computers.. using the following objective function: h T ij N Zj Fe = ∑ ∫ Qij dhij − ∑ ∫ q j dZ j → min (20) j =1 ij =1 0 0 and the energy conservation in loops constraints (3). in m. pipe diameter Dij... industrial concen-trated discharges in nodes qj. 1997). . in m3 /s.936. based on the optimization models developed above. in m. After substituting the functional relation (12) in (20) and after the integrals have been calculated the constraints can be eliminated and the problem can be simplified to the finding of the minimum of a function with N–NRP variables (Zi . The following data are known: pipe length Lij . β + 1∑ ij = 1 j =1 which can be achieved by using the conjugate gradient algorithm. N − N RP ) (22) Once the piezometric head at nodes have been determined one can calculate the residual pressure head using equation (19) and afterwards the discharges through pipes with equation (12) as well as other hydraulic parameters of the network. ANOREV) developed above. In order to determine an initial approximation of the piezometric head at nodes one must solve the following linear associated system: ∑R i≠ j i =1 N −1 ij ( Z (o) − Zi(o) − Π ij ) = q β j j ( j = 1. and the exponent β = 1. elevation head ZTj.. Zj) without constraints: β +1 N β β Rij β Zi − Z j + Π ij − ∑ q j Z j → min . piezometric head at the “critical node” Z1 = 124 m.One can find a variational formulation for the hydraulic analysis of looped networks also in node method. 4 Numerical application The looped distribution network with the topology from Figure 1 is considered. in m. discharges and pressures distribution using the classic HARDY–CROSS procedure and the two optimization models (ACIREV.

06917 1.02579 1.356 0. It can be seen that the results are very close.01610 2.067 0.04802 2.02626 1.02628 1.09501 1.123 0.13531 0.03605 0.02579 1.01656 3.06916 1.443 0.01327 1.11371 1.262 0.03771 1.06902 1.03776 1. Table 1.The difference between the discharges obtained with HARDY–CROSS and these given by ACIREV vary between 0.518 0.01613 2.01661 3.08199 0.949 0.906 0.075 0.147 0.945 .08206 0.8 % (pipe 3–7).092 0.08174 0. Specific consumption of energy for water distribution is 0.02625 1.03761 1.261 0.25266 1.11373 1.7% (pipe 9–8).03608 0.495 0.069 0.01351 2.13520 0. The discharges and head losses trough pipes Pipe i −j 0 2−1 3−2 4−3 13−4 6−5 7−6 8−7 9−8 13−9 11−10 12−11 13−12 5−1 7−2 9−3 6−10 8−11 Computation model ACIREV Q [m3/s] h [m] 3 4 0.034 0.160 0.894 0.521 0.11358 1.04806 2.521 0.065 0.17542 0.13496 0.568 0.558 0.276 0.014 HARDY-CROSS Q [m3/s] h [m] 1 2 0.094 0.900 0.01649 3.00705 kWh/m3 .25317 1.08 % (pipe 3–2) to 1.588 0.04799 2.907 0.158 0.09508 1.163 0.01647 3.03616 0.17547 0.356 0.889 0.203 0.198 0.437 0.359 0.495 0.587 0.122 0.127 0.915 0.007 0. for all three models of computation used.01661 3.447 0.1 Schema of the analyzed distribution network Table 1 shows the discharges and head losses through pipes established by using the three mentioned models of computation (the iterative tolerance imposed is 10-5 ).01204 3.521 0.01219 4.958 ANOREV Q [m3/s] h [m] 5 6 0.01332 1.02602 1.890 0. and the difference between discharges obtained with HARDY–CROSS and ANOREV varies from 0 % (pipe 9–3) to 2.01666 3.519 0.01597 2.123 0.18034 0.Fig.590 0.203 0.01209 3.09505 1.25299 1.

363 m which give a residual pressure head of 29.058 28.520 128. (3).778 131.520 25.594 128. The piezometric head and the residual pressure head at nodes Node j 0 1 2 3 4 5 6 7 8 9 10 11 12 13 HARDY–CROSS Zj [m] Hj [m] 1 2 124. Sparsity oriented analysis of large pipe networks. Management Science. ASCE. 3. which give it efficiency especially in functional analyzing of complex distribution networks.597 130.571 129.000 127.455 m and 131. (9) with the direct minimization of a multivariable function.078 27.558 130.537 130. COOLINS M.299 28. Moskva.839 130. The new method replaces the solving of the non-linear system of equations (2).042 127.571 28.627 128. Solving the pipe network analysis problem using optimization techniques. References 1.530 28.000 127.594 129. 0.. and COOPER L.094 26.949 27.007 127.001 m for ANOREV. N.402 29. 29.804 27.839 27.023 28.116 28.007 29.478 128. Table 2. without constraints that express the energy consumption across the network..497 128.594 28. 1978. CHANDRASHEKAR M.078 130.118 27.241 28.402 m.507 128. 1976.402 Computation model ACIREV Zj [m] Hj [m] 3 4 124.127 26.099 29.Table 2 presents the values for the piezometric head Zj and the residual pressure head Hj at nodes determined by using the classic procedure and the two new models of computation. nr.175 27.000 24.363 29363 5 Conclusions The mathematical model expressed by the objective functions (18) and (21) constitutes a new way of hydraulic analysis of complex looped networks based on unconditioned optimization techniques.455 29. nr.497 25.000 24.647 128. The divergence of piezometric line on network contour is 0.299 125.616 130.386 28. 131.523 130.537 28.449 128.478 25.386 125.901 27. ABRAMOV N.099 127.000 128. The computer programs ACIREV and ANOREV include this particular aspect and contain the conjugate gradient algorithm.502 28.455 m and 29. 2.335 125.502 130.109 m for ACIREV and only 0.147 26.402 m.122 m for HARDY–CROSS.597 28. This new method is computationally more efficient and consequently helps the designer to get the best design of water distribution systems with fewer efforts.804 130.042 29.007 27. .832 131.278 28.332 28. HY4. and STEWART K.175 130. Rascet vodoprovodnîi setei.118 130.000 24. 1975.741 131.455 ANOREV Zj [m] Hj [m] 5 6 124.363 m that is sufficient for the supply water to the consumers.915 27. The piezometric head at the node 13 has the following values: 131.335 28.415 128.530 129. 7..901 130. Journal of the Hydraulics Division. Stroizdat.

SÂRBU I. 1994. 13. Academiei Române. Modele de calcul de l'écoulement en régim permanent dans réseau d'eau maillés. 1987. 10. A Gradient Method for the solution of looped Pipe Networks. and PILATI S. 7... Ed. Practica optimizãrii asistatã de calculator. DIVÉNOT A. Ed. Int. 6. La Houille Blanche. Reliability of algorithms for pipe network analyses. 11. . Journal of the Hydraulics Division.4. Numerical and Optimizing Methods in Building Equipment Design.. Journal of the Hydraulics Division. nr.8. HY9. Conf. HY3. Tehnicã. nr. ASCE. Multiple operating points in complex pump network.2. V(1). Energetical Optimization of Water Distribution Systems. nr. and RAYES A. Bucureºti. 1979.. et al. and RODEH M. 1979. SIMA V. Hidrotehnica.. Journal de Recherches Hydrauliques. and VARGA A. Bucureºti. nr. 1987. nr. Loop equation with unknown pipe characteristics. 1981.. SÂRBU I. WOOD D. SÂRBU I. ASCE. 1986. Bucureºti. ASCE. 1981. Model de calcul al regimului hidraulic în reþele complexe de distribuþie a apei. LEKANE T.. 5.6. 1997. TODINI E. on Computer Applications in Water Supply and Distribution. COLLINS M. nr. 12. 1980. Tehnicã.. GOFMAN E.. Journal of the Hydraulics Division. Rev. Ed. 9. Une nouvelle méthode de calcul des réseaux maillés. 8. HY10..

Sign up to vote on this title

UsefulNot useful- Security Issues in Drinking Water Distribution
- Water as the Fix by Daniel Finneran
- Optimal-Robust Design of Water Distribution Networks
- Uncertainty in Water Distribution Network Modeling
- Optimal Cost Design of Water Distribution Networks
- Water Filtration Manual
- Water Pumping Manual
- Soap, toilets and taps
- lect23.pdf
- Digital Speech Processing
- Optimal Control Theory
- Dynamic Optimization Book
- Mixed Model Assembly Line Balancing
- Journal of Electromagnetic Analysis and Applications
- review of algebra.pdf
- Dynamic Optimization Primer
- MPHSpeedUp
- The Shannon Law for Non-Periodic Channels
- HO7_duality and SA
- Homework 2 Solutions
- Abstracts
- A4Contrl
- 181648916 Adjoint Tutorial PDF
- optimization
- 2 Curve Fitting
- Series So Ln Notebook
- Acemoglu, Autor - 2012 - What Does Human Capital Do a Review of Goldin and Katz's the Race Between Education and Technology
- Sq Preview
- 22881_An Analytic Solution of the Stagnation Point Flow
- Analysis of Looped Water Distribution Networks

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd