ANALYSIS of

FLOW in WATER
DISTRIBUTION
NETWORKS

Pramod R. Bhave, Ph.D.

Professor of Civil Engineering
Visvesvaraya Regional College of Engineering
Nagpur, India

TECHNOMIC
^PUBLISHING CO., INC.,
T ANCASTER • BASET ,
Analysis of Flow in Water Distribution Networks
aTECHNOMICpublication

Published in the Western Hemisphere by

Technomic Publishing Company, Inc.
851 New Holland Avenue
Box 3535
Lancaster, Pennsylvania 17604 U.S.A.

Distributed in the Best of the World by

Technomic Publishing AG

Copyright © 1991 by Technomic Publishing Company, Inc.

All rights reserved

No part of this publication may be reproduced, stored in a

retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of the publisher.

Printed in the United States of America

10 9 8 7 6 5 4 3 2 1

Main entry under title:

Analysis of Flow in Water Distribution Networks

A Technomic Publishing Company book

Bibliography: p. 441
Includes index p. 453

Library of Congress Card No. 91-52908

ISBN No. 87762-732-0
 To My Parents
Kantutai and Kaka (Raghunathrao)
 Table of Contents

Preface xv

t INTRODUCTION
1.1 Introduction 1
1.2 Historical Development 2
1.3 Fluid Properties 3
1.3.1 Density, Specific Weight, Specific Gravity
1.3.2 Compressibility 5
1.3.3 Viscosity 6
1.3.4 Other Properties 8
1.3.5 Properties of Water 8
1.4 Continuity Principle 8
1.5 Energy Principle 14
1.5.1 Forms of Energy 15
1.5.2 Bernoulli's Theorem 16
1.5.3 Hydraulic and Energy Grade Lines 20
1.6 Momentum Principle 20

FRICTIONAL HEAD LOSS IN PIPES 25

2.1 Introduction 25
2.2 Darcy-Weisbach Formula 26
2.3 Nikuradse's Experiments on Artificially Roughened Pipes 27
2.4 Moody Diagram 28
2.5 Friction Coefficient Relationships 31
2.5.1 Laminar Flow 31
2.5.2 Transitional Flow 31
2.5.3 Smooth Turbulent Flow 31
2.5.4 Transitional Turbulent Flow 32
2.5.5 Rough Turbulent Flow 32
vii
vjii Table of Contents

2.6 Explicit Relationships for Friction Coefficient 33

2.61 Smooth Turbulent Flow 33
2.62 Transitional Turbulent Flow 33
2.6 3 All Flow Regimes 36
2.7 Empirical Formulas 36
2.7.1 Hazen-Williams Formula 38
2.7.2 Modified Hazen-Williams Formula 41
2.7.3 Manning Formula 42
2.8 Comparison of Head Loss Formulas 43
2.9 General Head Loss Formula 44
2.10 Simple Pipe Flow Problems 45
2.10.1 Determination of Head Loss 45
2.10.2 Determination of Discharge 48
2.10.3 Determination of Diameter 49
2.11 Head Loss Due to Uniformly Decreasing Discharge 51
2.12 Reduction of Carrying Capacity with Age 53
2.13 Problems 57

S MINOR HEAD LOSS IN PIPES 59

3.1 Introduction 59
3.2 Sudden Enlargement 60
3.3 Gradual Enlargement 60
3.4 Exit 65
3.5 Sudden Contraction 65
3.6 Gradual Contraction 66
3.7 Entrance 66
3.8 Bends and Elbows 67
3.9 Tees 67
3.10 Obstructions 68
3.11 Flow Meters 68
3.12 Valves 68

4 EQUIVALENT PIPES 69
4.1 Introduction 69
4.2 Pipes in Series 69
4.2.1 Discharge Known 70
4.2.2 Total Head Loss Known 73
4.3 Pipes in Parallel 75
4.3.1 Head Loss Known 11
4.3.2 Total Discharge Known 78
 Table of Contents ix

4.4 Pipes in Series-Parallel 80

4.5 Minor Loss Elements 81
4.5.1 Darcy-Weisbach Formula 82
4.5.2 Hazen-Williams Formula 82
4.5.3 Manning Formula 83
4.5.4 Equivalent Pipe Lengths 83
4.6 Problems 86

& RESERVOIRS, PUMPS, AND SPECIAL VALVES 89

5.1 Reservoirs 89
5.7./ Impounding Reservoirs 89
5.1.2 Service and Balancing Reservoirs 89
5.1.3 Three-Reservoir System 92
5.1.4 Multiple-Reservoir System 94
5.2 Pumps 96
5.2.1 Introduction 96
5.2.2 System Head-Discharge Curve 97
5.2.3 Pump Head-Discharge Curve 97
5.2.4 Head-Discharge Relationship 99
5.2.5 Characteristic Curves 102
5.2.6 Pump Combinations 103
5.3 Special Valves 110
5 . 1 / Check Valves 110
5.3.2 Pressure Reducing Valves 112
5.4 Problems 117

i. TYPES AND PARAMETERS 119

6.1 Definitions 119
6.2 Types of Networks 121
6.2.1 Serial Network 121
6.2.2 Branching Network 121
6.2.3 Looped Network 122
6.3 Parameters 123
6.3.1 Configuration 123
6.3.2 Pipe Lengths 123
6.3.3 Pipe Diameters 123
6.3.4 Pipe Roughness Coefficients 123
6.3.5 Minor Appurtenances 124
6.3.6 Demand Pattern 124
63.7 Source Supply Pattern 124
6.3.8 Hydraulic Gradient Levels at Demand Nodes 124
6.3.9 Hydraulic Gradient Levels at Source Nodes 124
x Table of Contents

6.4 Labelling Network Elements 125

6.4.1 Branching Networks 125
6.4.2 Looped Networks 125
6.5 Parameter Interrelationships 125
6.5.1 Pipe Head Loss Relationship 126
6.5.2 Node Row Continuity Relationship 128
6.5.3 Loop Head Loss Relationship 128
6.6 Rules for Solvability of Pipe Networks 129
6.6.1 Rules Proposed by Shamir and Howard 129
6.6.2 Rules Proposed by Gofinan and Rodeh 130
6.6.3 Rules Proposed by Bhave 130
6.6.4 Comparison of Rules 131
6.7 Problems 137

?/„ FORMULATION OF EQUATIONS 139

7.1 Introduction 139
7.2 States of Parameters 141
7.3 Basic Unknown Parameters 142
7.4 Single-Source Networks with Known Pipe Resistances 142
7.4.1 Branching Networks 142
7.4.2 Looped Networks 142
7.5 Multisource Networks with Known Pipe Resistances 148
7.5.1 Branching Networks 148
7.5.2 Looped Networks 149
7.6 Networks with Unknown Pipe Resistances 156
7.6.1 Branching Networks 156
7.6.2 Looped Networks 157
7.7 Inclusion of Pumps 168
7.7.1 Supply Pumps 168
7.7.2 Booster Pumps 171
7.8 Inclusion of Check Valves 172
7.8.1 Check Valve in an External Pipe 172
7.8.2 Check Valve in an Internal Pipe 173
7.8.3 Check Valves in Several Pipes 175
7.9 Inclusion of Pressure Reducing Valves 175
7.9.1 PRVin an External Pipe 175
7.9.2 PRV in an Internal Pipe 179
7.9.3 PRVs in Several Pipes 181
7.10 Problems 181

S HARDY CROSS METHOD 185

8.1 Introduction 185
 Table of Contents xi

8.2 Method of Balancing Heads 185

8.2.1 Basic Concept 185
8.2.2 Single-Source Networks with Known Pipe Resistances 189
8.2.3 Multisource Networks with Known Pipe Resistances 191
8.2.4 Networks with Unknown Pipe Resistances 195
8.2.5 Networks with Pumps 202
8.2.6 Networks with Valves 206
8.3 Method of Balancing Flows 209
8.3.1 Basic Concept 209
83.2 Networks with Known Pipe Resistances 211
8.3.3 Networks with Unknown Pipe Resistances 211
83.4 Networks with Pumps 214
83.5 Networks with Valves 217
8.4 Modified Hardy Cross Method 220
84.1 Method of Balancing Heads 220
84.2 Method of Balancing Flows 222
8.5 Selection of Initial Values 224
85.1 Selection of Pipe Flows 225
8.5.2 Selection of Nodal Heads 227
85.3 Selection of Unknown Resistances 227
8.6 Convergence Problems 227
8.7 Problems 229

NEWTON-RAPHSON METHOD 231

9.1 Introduction 231
9.2 Basic Concepts 231
9.2.1 Single-Variable Function 232
9.2.2 Multiple-Variable Function 233
9.3 Head Equations 235
9.3.1 Networks with Known Pipe Resistances 235
9.3.2 Networks with Unknown Pipe Resistances 239
9.3.3 Networks with Pumps 242
9.3.4 Networks with Valves 244
9.4 Loop Equations 246
9.4.1 Networks with Known Pipe Resistances 246
9.4.2 Networks with Unknown Pipe Resistances 248
9.4.3 Networks with Pumps 253
9.4.4 Networks with Valves 254
9.5 Labelling Network Elements 256
9.5.1 Algorithm of Epp and Fowler 257
9.5.2 Algorithm of Jeppson and Davis 258
9.6 Problems 261
xii Table of Contents

i t LINEAR THEORY METHOD 263

10.1 Introduction 263
10.2 Pipe Discharge Equations 264
10.2.1 Networks with Known Pipe Resistances 264
10.2.2 Networks with Unknown Pipe Resistances 267
10.2.3 Networks with Pumps IB
10.2.4 Networks with Valves 276
10.3 Nodal Head Equations 278
10.3.1 Networks with Known Pipe Resistances 278
10.3.2 Networks with Unknown Pipe Resistances 286
10.3.3 Networks with Pumps 290
10.3.4 Networks with Valves 293
10.4 Relationship between Newton-Raphson and
Linear Theory Methods 295
10.4.1 Pipe Discharge Equations 296
10.4.2 Nodal Head Equations 297
10.5 Problems 301

t OTHER ANALYSIS METHODS 303

11.1 Introduction 303
11.2 Analysis Using Electrical Analyzers 303
11.2.1 Hydraulic and Electrical Analogies 303
11.2.2 Electrical Analyzers with Linear Resistors 305
11.2.3 Electrical Analyzers with Nonlinear Resistors 306
11.3 Analysis through Unsteady Behaviour during Start-Up 308
11.3.1 Basic Principles 308
11.3.2 General Discussion 312
11.4 Analysis through Optimization 312
11.4.1 Introduction 312
11.4.2 System Energy Approach 313
11.4.3 Content Model Approach 318
11.4.4 Co-Content Model Approach 322
11.5 Problems 325

;?: DYNAMIC ANALYSIS 327

12.1 Introduction 327 •
12.2 Iterative Method 328
12.2.1 Introduction 328
12.2.2 Procedure 329
12.3 Direct Method 337
12.3.1 Theory 338
 Table of Contents xiii

12.3.2 Procedure 341

12.4 Problems 350

NODE FLOW ANALYSIS 353

13.1 Introduction 353
13.2 Two-Node Serial Network 355
13.2.1 Increase in Upstream Nodal Flow 356
13.2.2 Increase in Downstream Nodal Flow 357
13.2.3 Increase in Both Nodal Flows 358
13.3 Node Classification 366
13.3.1 Classification According to HGL 366
13.3.2 Classification According to Flow 366
13.4 Node-Category Compatibility 367
13.5 NFA Theory 368
13.5.1 Problem Formulation 368
13.5.2 Problem Solution 370
13.5.3 Solution Procedure 371
13.6 General Observations 383
13.7 Problems 384

CALIBRATION 385
14.1 Introduction 385
14.1.1 Definition 385
14.1.2 Reasonable Agreement 386
14.2 Data Collection and Preparation 387
14.2.1 Skeletonization of Network 387
14.2.2 Heads at Source Nodes 389
14.2.3 Supply Rates at Source Nodes 389
14.2.4 Consumption at Demand Nodes 389
14.2.5 Pipe Details 389
14.3 Calibration Methods 391
14.3.1 Methods Adjusting Pipe Coefficients 392
14.3.2 Methods Simultaneously Adjusting Pipe Coefficients and
Nodal Demands 395
14.4 Practical Considerations 405
14.4.1 Fine-Tuning the Model 405
14.4.2 Sources of Error in Field Data Collection 406
14.4.3 Field Equipment 408

MISCELLANEOUS TOPICS 409

15.1 Comparison of Common Analysis Methods 409
 xiv Table of Contents

15.2 Sensitivity Analysis 409

15.3 Pitfalls in Network Analysis 418
15.3.1 Error Criteria and Convergence 418
15.3.2 Finding Unique Solutions 419
15.4 Reliability Analysis 419
15.5 Unsteady Flow Analysis 421
15.6 Computer Application 422

15.7 Problems 423

Nomenclature 425

Answers 435

References 441

Index 453

About the Author 461

 Preface

T is taken for granted that water will always be available in the desired
I quantity when we open faucets, without our bothering to know how it is
brought to us. It is only when we do not get a sufficient quantity of water
that we start grumbling about the water supply engineers. This is the
natural human tendency and it is observed in all walks of life. It is there-
fore necessary for water supply engineers to have a good knowledge of
water supply engineering in general and water distribution systems in par-
ticular. This will help in the proper planning, analysis and design of water
distribution systems. The availability of water at different localities under
different demand patterns and at different water levels in the reservoirs, the
dependability of the system under fire flow requirements or pipe breaks,
and the expansion of the existing water distribution systems to accom-
modate new residential localities or industries are some examples where
the proper understanding of network behaviour would be necessary. It was
with this intention that a two-semester course on water distribution sys-
tems was introduced at Nagpur University—an elective subject in the final
year of the undergraduate civil engineering curriculum, and a compulsory
one for the post graduate courses in hydraulic engineering and environ-
mental engineering disciplines. The present text is the outgrowth of the
lecture notes prepared for them over the past several years.
The emphasis throughout the text has been on explaining the fundamen-
tal concepts. Numerous examples have been provided in the text. The ex-
amples are simple enough so that they can be tackled by calculators but are
complex enough to explain the finer points. Problems are provided, wher-
ever necessary, for the readers to test their knowledge. Answers are listed
towards the end of the book (pp. 435-440). No reference to computers or
computer programs is made in the text (except in the end) because the
reader, having understood the fundamental concepts, can easily formulate
programs in BASIC or FORTRAN languages. It is however emphasized
that the distribution networks encountered in practice are not simple

xv
xvi Preface

enough to be tackled just by calculators, and therefore sophisticated com-

puter programs that are commercially available must be used for their
analysis.
Thefirstfivechapters provide the necessary background before the anal-
ysis of the distribution systems can be considered. The reader can rapidly
go through them and skip certain portions. Chapter 1 deals with the histor-
ical development of water distribution systems and the essential fundamen-
tal concepts from fluid mechanics. Chapter 2 deals with the frictional head
loss in pipes. Even though a reader may have studied this under fluid me-
chanics, he is advised to go through this chapter because some useful ma-
terial can be found herein. Chapters 3 and 4 deal with minor loss elements
and equivalent pipes, respectively, and the reader may skip these chapters.
Chapter 5 covers reservoirs, pumps, check valves, and pressure reducing
valves. The reader is advised to go through this chapter and understand
their behaviour, especially that of valves under different conditions.
Chapters 6 to 10 cover different types of networks and the methods com-
monly employed for their analysis. Chapter 6 describes the types of net-
works, different parameters involved in their analysis, relationships among
these parameters and certain rules that must be satisfied for network analy-
sis to be feasible. Chapter 7 describes the formulation of different types of
equations that are used in network analysis. The illustrative networks
gradually become complex—single-source, multisource, networks with
unknown pipe resistances, networks with pumps, networks with check
valves, and finally networks with pressure reducing valves. Chapters 8 to
10 describe the Hardy Cross method, Newton-Raphson method and the
linear theory method of network analysis, respectively. These methods are
considered in their chronological order of development. Typical networks
selected for illustrative examples are solved by all three methods.
Chapters 11 to 15 mainly describe the developments in recent years.
Chapter 11 briefly deals with other analysis methods that are either ob-
solete as those using electrical analyzers or the ones which are not com-
mon as those based on unsteady behaviour during start-up and based on
optimization principles. Chapter 12 describes dynamic analysis that helps
in predicting the network behaviour over an extended period of time.
Chapter 13 describes node flow analysis used to predict the behaviour of
the network under deficient conditions. Chapter 14 deals with calibration
of distribution networks, a step essential before their modification or ex-
pansion can be undertaken. Chapter 15 covers miscellaneous topics-
topics that are useful in network analysis, for example, sensitivity analysis
and pitfalls to be guarded against in network analysis; topics under devel-
oping stages, for example reliability analysis; and complicated topics just
briefly introduced such as the unsteady-flow analysis. The reader can refer
to the given references for detailed information on these topics. As men-
tioned earlier, computer applications are just mentioned in the end.
 Preface xvii

I am indebted to Professor Roland W. Jeppson for his book Analysis of

Flow in Pipe Networks, first published more than a decade ago, and the
only book available on this topic for quite some time. I am also indebted
to the numerous research workers whose publications have been referred
to herein. I would like to thank Professor Paul N. Cheremisinoff for his
help in getting the book published. Thanks are also due to R. K. Pal for typ-
ing the manuscript practically without any errors, Murlidhar A. Dharaskar
for drawing the figures, and Narayan S. Nimbokar for his help in getting
the photocopies of the typed manuscript. In the end I would also like to
thank Technomic Publishing Co., Inc. and its staff for undertaking the
printing of this book and doing it excellently.

PRAMOD R. BHAVE
 n

Introduction

1.1 INTRODUCTION

EXT to air, water is the most essential commodity required to

N maintain life. Of the several words for water in Sanskrit, one is
"Jeewanam" which also means life. Initially, mankind used water for sim-
ple domestic purposes such as drinking, cooking, bathing and washing.
However, the present day uses of water are varied and they may be
classified as domestic, public, commercial, and industrial. Domestic use
includes water required not only for drinking, cooking, bathing and wash-
ing; but also for heating, cooling, air-conditioning, sanitary purposes,
private swimming pools, and watering lawns and gardens. Public use in-
cludes water required at public places and buildings such as public
gardens, parks and fountains, public swimming pools, hospitals, schools
and other educational institutions, hostels, prisons, public sanitary places,
street and sewer flushing and fire fighting. Commercial use includes water
used in office buildings, hotels and restaurants, for car washing, in laun-
dries, at golf courses, shopping centres, bus, railway and air terminals.
Industrial use for manufacturing and processing purposes consists mainly
of heat exchange, cooling and manufacturing [1].
A modern water supply system may include facilities for collection and
storage, transportation, pumping, treatment and distribution. Works for
collection and storage include development of watersheds, dams, im-
pounding reservoirs, intakes, wells, infiltration galleries and springs.
Transportation works include canals, aqueducts, and pumping mains for
transporting water from storage reservoirs to consumer centres. Pumping
works include pumps and other ancillary units for pumping water. Treat-
ment includes aeration, screening, sedimentation, flocculation, filtration,
disinfection, softening and removal of harmful elements. Distribution
works include distribution and equalizing reservoirs, pipes, valves, fire
hydrants and other appurtenances.
1
2 INTRODUCTION

Of the total expenditure incurred on different facilities of a water supply

system the expenditure incurred on the distribution network is quite large
and may even exceed 70 percent. Therefore, it is essential to know the be-
haviour of a distribution network under different conditions such as varia-
tion in demand including fire flow requirements, variation in reservoir
water levels, partial or full valve openings, closure of pipes during repairs
or replacement. This knowledge is obtained by carrying out an analysis of
the flow in the distribution networks.

1.2 HISTORICAL DEVELOPMENT

The history of water supply and distribution is as old as the history of

civilization. All the earlier civilizations developed on the banks of rivers
such as Sindhu (Indus) and Ganga, Tigris and Euphrates, Yellow, and Nile.
Mankind also developed means to store water in reservoirs. Excavations at
Shringaverapura in Uttar Pradesh, India, have revealed a 2,000-year-old
tank for storing water for water supply purposes. It is more than 250 m in
length, built of brick masonry parallel to the river Ganga [2]. The means
for transportation and distribution of water for water supply and irrigation
were also developed in earlier times. As observed from the excavations at
Harappa and Mohenjo-daro, the Sindhu-valley civilization had, more than
5,000 years ago, baths in many of the houses with ceramic pipes for water
supply and brick conduits under the centre of the streets for drainage [3,4].
Artificial conduits were also constructed in earlier times for the convey-
ance of water over reaches in which the natural topography did not permit
the use of open cuts. There are traces of remarkable aqueduct systems,
built by the Phoenicians in Syria and Cyprus, involving tunnels through
rocks and pressure conduits over valleys. Jerusalem was supplied water,
some 3,000 years ago, through two conduits, one of them more than 30 km
long and crossing the valley of Hinnon on arches [4].
More than 2,000 years ago, the city of Rome had a well-developed water
supply system. Water was collected from several sources in a reservoir and
transported through aqueducts or pressure conduits to a castellum or dis-
tribution reservoir. Frequently, one conduit supplied several castella. An
extensive distribution system was laid in the streets with numerous public
fountains. The distribution system within the city and the pressure con-
duits were built of several materials depending upon the pressures they had
to withstand. Lead and clay were the most common, but copper, bronze
and bored-stone pipes were also used [4,5].
In 1544, the British Parliament passed an act ". . . to provide for clear
water for the residents of London" and began to issue charters for private
water works companies. Boston, Massachusetts is credited with the earli-
 Fluid Properties 3

est recorded public water supply in the U.S.A.—1652. However, by 1800,

there were only sixteen public water supply systems in the U.S.A.; most
of them were in New England or in the larger cities near the Atlantic-
coast—originally built for "fire protection and the laying of dust" with little
thought given to domestic service. The number of water supply systems in
the U.S.A. grew to 83 by 1850, 600 by 1880 and 3350 by 1897 [6].
Even though distribution systems came into existence in the U.S.A.
when the first conduit was laid in Boston in 1652, Winston-Salem, North
Carolina is credited with the first complete domestic distribution system in
1776. It used wooden mains, that were primarily bored and charred logs.
Wood stave piping was introduced later and was used as late as the 1940s.
However, by 1850, iron pipes were preferred. After the development of
asbestos-cement pipes by an Italian waterworks inventor in 1913 and the
plastic pipes—polyvinyl chloride (PVC) and acrylonitrile, butadiene, and
styrene (ABS)—after World War II, these pipes were also used in modern
water distribution systems [6].
Water has been and still is the main commodity that is transported and
distributed through pipelines. However, in recent times, other materials
are also transported and distributed through pipelines and conduits. Trans-
portation and distribution of domestic gas to the consumers; movement of
crude oil from offshore fields to onshore separation and compressor
plants; conveyance of different chemicals in chemical plants; transporta-
tion of coal and ores from mines, fly ash from thermal power plants, etc.,
in the form of slurry; and ventilation of tunnel systems in underground
mines are some such examples [7,8]. Even though water is primarily con-
sidered in this text, the general principles developed herein would also be
applicable to such materials.

1.3 FLUID PROPERTIES

1.3.1 DENSITY, SPECIFIC WEIGHT, SPECIFIC GRAVITY

Density

The mass per unit volume of a substance is termed its density or specific
mass, and is generally denoted by Q (rho). Its dimensions are ML"3 and
unit is g/cm3 or kg/m3 (1 g/cm3 = 1,000 kg/m3).

Specific Weight

The weight per unit volume of a substance is termed its specific weight,
usually denoted by y (gamma). As it is a force per unit volume, its dimen-
4 INTRODUCTION

sions are FL~3 or ML~2T~2 and the unit is dyn/cm3 or N/m 3 (1 N = 1

kg-m/s 2 = 10s dyn; 1 dyn/cm3 = 10 N/m 3 ).
The specific weight y and the density Q of a substance are interrelated
through the gravitional acceleration g by a relationship

1 = Qg (1-1)

Specific Gravity

The ratio of the weight of a substance to the weight of an equal volume

of water at 4°C is termed specific gravity. Naturally it is dimensionless.
Temperature and pressure affect these properties to a remarkable extent
in the case of gases, whereas to a small extent in the case of liquids. For
example, the density of water at 100°C is about 4 percent less than that
at 4°C. Therefore, the effect of temperature on these properties of liquids
is neglected in practice. For water, the standard value for density is taken
as 1 g/cm 3 or 1,000 kg/m 3 , which is its maximum value at 4°C. Similarly,
the standard value for specific weight of water is 981 dyn/cm 3 or 9,810
N/m3—for specific gravity it is 1.

Example 1.1

An empty container weighing 1.287 N weighs 20.327 N when com-

pletely filled with a liquid, and weighs 21.760 N when completely filled
with water. Determine the density, specific weight, and specific gravity of
the liquid.

Solution

The weight of water in the container = 21.760 - 1.287 = 20.473 N.

Since 1 m3 of water weighs 9,810 N, the

volume of water in the container = ' = 2.087 X 10~3 m3

9,810
which is also the capacity of the container. The weight of liquid in the con-
tainer = 20.327 - 1.287 = 19.040 N.

19 040
.'. Mass of the liquid = ' = 1.9409 kg
9ol

= 93 °
 Fluid Properties

19 040
Specific weight = 2Q87 ' X 1Q_3 = 9,123 N/m 3

19 040
Specific gravity = 2Q'473 = 0-93

1.3.2 COMPRESSIBILITY

All matter is compressible to some extent. The compressibility of a

substance is expressed through its bulk modulus of elasticity, K. If a
substance with an initial volume V suffers a volume change dV when sub-
jected to an increased pressure dp, then

[Since a rise in pressure causes a decrease in volume, dp and dKare of op-

posite sign. Therefore, a minus sign is included in Equation (1.2) to get a
positive value of K.\
As AVIV is dimensionless, K has the dimensions and units of dp, i.e.,
pressure. Thus, the dimensions of K are FL~2 or ML~ ir T 2 and the unit is
N/m 2 or Pa. (1 Pa, pascal = 1 N/m 2 ). For water, the average value of A"
is 2.07 x 109 Pa. As the value of Kfor all liquids is very large, barring sit-
uations where sudden or great changes in pressure are involved, the liq-
uids are considered incompressible.

Example 1.2

Find the specific weight of water at a depth of 100 m from the water sur-
face. Take K = 2.07 x 109 Pa.

Solution

The increase in pressure intensity above standard atmospheric pressure

at a depth of 100 m from water surface can be approximately taken as
9,810 X 100 = 981,000 Pa.
Now consider 1 N of water. Let its volume at standard atmospheric
pressure be V0(V0 = 1/9810 m3) and at 100-m depth be V. Then,

V = Vo + dV
-*('•?)
6 INTRODUCTION

But from Equation (1.2), dV/V = - dpIK

981
• v= vi\ -*P\ =-*—(} '°°° \
°\ Kj 9,810 \ 2.07 x 109j
=
m X °" 9 526 ^
.'. Specific weight of water at a 100-m depth

9,810
= 9,814.7 N/m 3 .
0.999 526

(Even at a 100-m depth from the water surface, the increase in the
specific weight of water is hardly 0.05 %. Therefore, considering water as
incompressible is valid in practice.)

1.3.3 VISCOSITY

Viscosity is the property of a fluid that offers resistance to shear stress.

According to Newton's law of viscosity, the relationship between the shear
stress T (tau) and the rate of angular deformation dv/dy for a one-
dimensional flow of a fluid is given by:

dv
T = „ - (1.3)

in which \i (mu) is the proportionality constant, known as the coefficient

of viscosity, absolute viscosity, dynamic viscosity, or simply viscosity of
the fluid. The angular deformation dv/dy is also known as the velocity
gradient as it gives the rate of change of velocity v with respect to the dis-
tance y.
From Equation (1.3) the dimensions of viscosity gan be obtained as
FL~2T or ML"'T"' and the units are N • s/m2, Pa • s, or kg/m • s. Viscos-
ity is also sometimes expressed in Poise, P (1 P = 1 dyn • s/cm2 =
0.1 N • s/m2) and centipoise, cP (1 cP = 0.01 P).
The effect of pressure on viscosity is negligible under ordinary condi-
tions. The temperature, however, affects viscosity appreciably. A rise in
temperature reduces the viscosity of all liquids but increases the viscosity
of gases.
Viscosity plays an important role when the fluid flow is laminar. In tur-
bulent flow, however, the apparent shear stress is due to the exchange of
momentum between adjacent layers of the fluid, and the viscous effects are
negligible.
 Fluid Properties 7

The ratio of the dynamic viscosity /* to the density Q of a fluid is known

as kinematic viscosity, v so that
E
v = (1.4)
e
From the dimensions of /i and Q, the dimensions of v can be obtained
as L 2 T~\ The unit of kinematic viscosity is m2/s. As this unit is large,
mm2/s ( = 10"6 m2/s) is also taken as a unit. In practice, v is sometimes ex-
pressed in stokes (1 stoke = 1 cmVs = 10"4 m2/s) and centistokes (1 cen-
tistoke = 0.01 stoke = 1 mm 2 /s).
Example 1.3

The velocity distribution for laminar flow of a liquid (n = 0.04 Pa • s)

in a 0.25-m diameter pipe is given by

v = 0.5y - If
in which v = velocity in m/s at a point y meters away from the inner sur-
face of the pipe wall. Find the shear stress at the pipe wall and at 0.05 m
from the pipe wall. Also determine the total resistance in a 100-m length
of the pipe.
Solution

The velocity distribution is parabolic and v = 0 at y = 0 and 0.25 m,

points on the inner surface of the pipe wall.

dv
From Newton's law of viscosity, Equation (1.3), T = /*—

As v = 0.5y - f, ^ = 0.5 - 2y

•• ( ? ) = 0.5 s-
\dy/(^o)
At pipe wall, T0 = 0.04 X 0.5 = 0.02 Pa.

Aty = 0.05 m, I - T H = 0.3 s"1

•'. Tiy=o.os) = 0.04 X 0.3 = 0.012 Pa

8 INTRODUCTION

The total resistance in a 100 m length of the pipe = r0 X irDL

0.02 X TT X 0.25 x 100 = 1.571 N.

1.3.4 OTHER PROPERTIES

Other fluid properties such as cohesion, adhesion, capillarity, surface

tension, and vapour pressure are mostly insignificant for flow through
pipes and pipe networks. These properties, therefore, are not discussed.

1.3.5 PROPERTIES OF WATER

Since water is the primary fluid that is considered in this text, some
properties of water at different temperatures are given in Table 1.1.

1.4 CONTINUITY PRINCIPLE

In the absence of nuclear reactions, matter can neither be created nor

destroyed. This is the principle of the conservation of mass and gives the
continuity equation: "The net mass rate leaving a control volume plus the
rate of accumulation of mass within the control volume must equal zero."

TABLE 1.1. Some Properties of Water at Different Temperatures.

Bulk Modulus Dynamic Kinematic
Temperature, Density, of Elasticity, Viscosity, Viscosity,
°C kg/m3 • 109 Pa 10-3 Pa-s mm2/s
(1) (2) (3) (4) (5)

0 999.86 2.04 1.792 1.792

4 1000 2.06 1.568 1.568
10 999.72 2.11 1.308 1.308
15 999.10 2.14 1.138 1.139
20 998.23 2.20 1.002 1.004
25 997.1 2.22 0.894 0.897
30 995.7 2.23 0.798 0.802
35 994.1 2.24 0.720 0.724
40 992.2 2.27 0.653 0.658
50 988.1 2.30 0.547 0.554
60 983.2 2.28 0.468 0.476
80 971.8 2.21 0.353 0.363
100 958.4 2.07 0.282 0.294
 Continuity Principle

Figure 1.1 Control volume for continuity equation.

Consider a control volume bounded by a stream tube and two cross sec-
tions 1-1 and 2-2 as shown in Figure 1.1. Let the fluid enter through section
1-1 and leave through section 2-2. If the flow is steady, the rate of accu-
mulation of the fluid mass between sections 1-1 and 2-2 is zero. Thus, the
fluid mass rates passing through sections 1-1 and 2-2 are equal. Summing
up the mass flow rate over a collection of stream tubes, the continuity
equation for mass flow rate for steady flow in a pipe can be written as

M = I qvdA = 1 QvdA (1.5)

J A, J A2

in which M = mass flow rate; Q = density; v = velocity; and

Ax,A2 = the cross-sectional areas of sections 1-1 and 2-2, respectively.
If the flow is incompressible, Equation 1.5 can be written as

M= QQ = Q I vdA =
= Q
Q \\ vdA (1.6)
«Mi J

in which Q = volumetric flow rate or the discharge having dimensions

L3T~' and units m3/s.
From Equation (1.6), the continuity equation for discharge becomes

Q = I vdA
vdA == I\ vdA (1.7)
J At J/A2

If F, and V2 are the average velocities at sections 1-1 and 2-2 with cross-
sectional areas At and A 2 , respectively, the continuity equations for mass
10 INTRODUCTION

flow rate and volumetric flow rate in incompressible flow become, respec-
tively

M = QQ = QAiVi = QA2V2 (1.8)

Q = AlVt = A2V2 (1-9)

In distribution networks, two or more pipes may meet at a junction. The

continuity equation for steady incompressible flow at such a junction
would state that the sum of the mass flow rate (or volumetric flow rate) en-
tering a junction must equal the sum of the mass flow rate (or volumetric
flow rate) leaving it. In other words, the algebraic sum of the mass flow
rate (or volumetric flow rate) at a junction must be zero. Thus,

EM = LQQ =0 (1.10)

LQ = LAV = 0 (1.11)

Example 1.4

Pipes 1,2,3 and 4 meet at junction J of a water distribution network as

shown in Figure: Example 1.4. The diameters of pipes 1, 2, 3 and 4 are
300 mm, 100 mm, 200 mm, and 200 mm, respectively. The velocities of

.Q2

T - * • 2 m/s

pipe.1

= 0-020
 Continuity Principle 11

flow in pipes 1 and 3 are 2 m/s towards J and 3 m/s away from J, whereas
the discharge in pipe 4 is 0.020 m3/s away from J. Determine the velocity
and discharge in pipe 2.

Solution

For Pipe 1,

Qx = AM = J ( 0 . 3 ) 2 X 2 = 0.1414 nrVs

Similarly for pipe 3,

Q3 = A3V3 = J (0.2) 2 X 3 = 0.0942 m3/s

Assuming that the flow in pipe 2, i.e., Q2, is towards the junction and
taking the flow towards the junction as positive and that away from the
junction as negative, the continuity equation for the volumetric flow rate
gives

Qi + ft - Ga - QA = 0

.-. 0.1414 + Q2 - 0.0942 - 0.020 = 0

or Q2 = - 0 . 0 2 7 2 m3/s

The negative sign indicates that Q2 is not towards the junction as

assumed, but is away from the junction. Thus, the discharge in pipe 2 is
0.0272 mVs, away from J.

Velocity in pipe 2, V2 = -r-2 = —' = 3.463 m/s, away from J

f
Example 1.5

The velocity distribution for a particular flow in a 200-mm diameter

pipe is given by v = 1.2 (y/r)1'7, in which r = radius of the pipe in
meters and v = velocity, in m/s, at y meters from the pipe wall. Deter-
mine the discharge, average velocity, and the maximum velocity.
12 INTRODUCTION

Solution

Consider an annular ring of thickness by at a distance y from the pipe

wall as shown in Figure: Example 1.5. The area of the ring is 2ir(r - y)8y

and the velocity of flow through it is v = 1.2 (y/r)u7. Therefore, the dis-
charge 5Q passing through it is given by

2 4x
6Q = 2TT (r - y) 6y X 1.2 (ylr)"7 = —jr(ryU7 - / " ) 8y

On integration, the discharge through the pipe is

Q
" Jo '" 50

Here r = 100 mm or 0.1 m, therefore

49
Q = — ,r(0.1)2 = 0.03079 nvVs

and

O 49
the average velocity V = — = — = 0.98 m/s
y4 50

The maximum velocity 1.2 m/s occurs at y = r, i.e., at the centerline.

Example 1.6

A water distribution system is shown in Figure: Example 1.6. The con-

sumptions or the demands at the junctions are shown by outward arrows
 Continuity Properties 13

Reservoir,1
Reservoir, 2

60 TT

9
20 30 20 20

r
20 30 10
and the supplies at the junctions by inward arrows. All the discharges are
in liters per minute, L/min. Determine the supply rates from the two reser-
voirs and the discharges in the pipes.

Solution

Applying successively the discharge continuity equation, supply from

reservoir 1 = Q,. 2 = 60 L/min; Q2.3 = 60 - 20 = 40 L/min; Q3.4 =
40 + 30 = 70 L/min; Q4.s = 70 - 40 = 30 L/min; Q6.7 = 20
L/min; Qs.6 = 10 L/min; Qs-6 = 20 + 30 - 10 = 40 L/min; Q9-s =
40 - 30 = 10 L/min; Ql0-9 = 10 + 20 = 30 L/min; and Qlt-10 =
30 + 20 = 50 L/min = supply rate from reservoir 2.

Example 1.7

Determine the supply rate at junction 6 and the discharges in pipes of the
distribution network shown in Figure: Example 1.7.

0-1 -
14 INTRODUCTION

Solution

By adding the continuity principle to the entire network, the supply rate
at node 6 = 0.2 + 0.3 + 0.1 + 0.3 - 0.5 = 0.4.
Assuming the flow directions in the pipes as shown in Figure: Exam-
ple 1.7, we get, 2,-4 = 0.5 - G,-2; Q4-s = Q^ - 0.1 = 0.4 - 0,- 2 ;
ft-s = Qi-2 - 0.2 - 0 2 . 3 ; 06-5 = 0.3 - 0 4 - 5 - Q2-s = 0.3 -
(0.4 - e,_2) - (e,_ 2 - 0.2 - Q2-3) = 0.1 + Q2.3; Q6.3 = 0.3 -
G2-3.
At junction 6, the supply rate = Q6-s + Q6-3 = (0.1 + Q2-3) +
(0.3 — Q2-3) = 0.4, as obtained from the consideration of the entire net-
work. In this example:
(1) The supply rate at junction 6 can be found either by considering the
network as a whole or by determining the discharge in pipes 6-5 and
6-3 and then considering junction 6.
(2) The absolute values for the pipe discharges cannot be determined by
simply applying the continuity principle. They can be expressed
only in terms of discharges in some pipes. In the solution herein, the
pipe discharges are expressed in terms of the discharge in pipes 1-2
or 2-3 or both.

1.5 ENERGY PRINCIPLE

In the absence of nuclear reactions, energy can neither be created nor

destroyed. However, energy may be converted from one form to another.
A flowing mass of a fluid may contain energy in different forms such as
mechanical, thermal, and chemical. For water distribution networks, how-
ever, only the energy that is available and therefore useful is considered.
Since other forms of energy cannot be used, if the available energy is con-
verted to any other form, the energy is considered as lost. For example, in
a flowing mass of water, part of the available energy may be converted into
thermal energy due to internal friction. Such a conversion is considered a
loss of energy.
The available energy of a fluid is proportional to its mass, and therefore
it is common to express energy, per unit mass or per unit weight. As energy
is the capacity to do work and as work is force multiplied by distance, the
dimensions of energy are FL or ML2T~2 and the units are N • m or Joule,
J (1 J = 1 N • m). Naturally, the dimensions of energy per unit mass are
FL/M or L2T~2 and the units are 1 N • m/kg or 1 J/kg. When the energy
is expressed per unit weight, the dimensions are FL/F, i.e., L and the unit
is meter, m.
 Energy Principle 15

1.5.1 FORMS OF ENERGY

The available energy may be in one or a combination of three forms of

energy, namely (1) potential energy, (2) flow (or pressure) energy, and (3)
kinetic energy.

Potential Energy

The energy possessed by a fluid mass due to its position or elevation

above datum is termed potential energy. A unit fluid mass (weight equal
to g units) with its centre of mass Z units of height above datum has a ca-
pacity of doing gZ units of work in falling through a height Z. Thus, the
potential energy of unit mass is gZ.

Flow Energy

The capacity of a fluid mass to do work on its surroundings when the

fluid is flowing is termed as flow energy. Consider a piston having cross-
sectional area A, fitted to a reservoir as shown in Figure 1.2. The piston
is subjected to a pressure p on its centre so that the force acting on it hpA.
If the water pressure moves the piston through a distance L, the work done
on the piston is pAL. The mass of water on which the work is done is QAL
so that the work done per unit mass is p/g. Thus, the term p/g corre-
sponds to the work which would be done on a unit mass of water by
pressure forces if water moved from a point having pressure p to another
point having pressure zero.
As the flow energy termplQ contains the term for pressure/?, it is com-
monly (but erroneously) termed as pressure energy. It should, however, be
remembered that pressure energy is different from elastic energy that is
given to a fluid when it is compressed.

Reservoir

PA-
d L
Figure 1.2 Flow energy.
16 INTRODUCTION

Kinetic Energy

The energy possessed by a fluid mass due to its velocity is termed kinetic
energy and is equal to mass times V2/2. Thus, the kinetic energy possessed
by a unit mass is V2/2.
It can be easily seen that each of the above forms of the available energy
per unit mass has the dimensions of FL/M or L2T~2 and units of N • m/kg
or J/kg. These forms of energy are interchangeable and their sum consti-
tutes the total energy E per unit mass of a fluid. Thus,

p V2
£ per unit mass = gZ + — + -y- (1-12)

1.5.2 BERNOULLI'S THEOREM

If the steady flow of a frictionless, incompressible fluid along a stream-

line is considered, the total energy of a unit fluid mass would remain the
same. This is Bernoulli's theorem. Thus, Bernoulli's theorem can be ex-
pressed as
p V2
£ per unit mass = gZ + — + — = constant (1.13)
e 2
Considering two sections 1-1 and 2-2 in a flowing fluid, Bernoulli's
equation, Equation (1.13) can be written as

in which subscripts 1 and 2 represent sections 1-1 and 2-2, respectively.

Bernoulli's equation is applicable to the steady, incompressible flow
along a streamline of a frictionless fluid. The steady flow of liquids
(mostly water) that can be considered as incompressible for all practical
purposes is considered primarily by this text. Furthermore, the flow in a
pipe can be considered one-dimensional so that the velocity at all points of
a cross section is taken equal to the average velocity. Thus, the three
assumptions, i.e., steady, incompressible and along a streamline, can be
considered to be valid for flow of liquids in pipes. However, since a real
liquid is not frictionless, some loss of energy takes place when a liquid
flows from section 1-1 to section 2-2. Representing such energy loss per
unit mass by ELi_2, Bernoulli's equation for real fluids can be modified to

gZi + iL + n = gZ2+ p± + n + ELI2 (L15)

 Energy Principle Yl

Instead of considering the energy per unit mass, many times the energy
is considered per unit weight. As weight is g times mass, by dividing each
term of Equation (1.15) by g, Equation (1.15) can be written as

Z, + ^ + ^ = Z2+ ^ + ^ + hL,_2 (1.16)

y 2g y 2g
in which

7 = Qg; and hL - —

Each term in Equation (1.16) represents energy per unit weight and
therefore has the dimensions of FL/F, i.e., L and unit of N-m/N, i.e.,
m. Each term represents a height of liquid column and is therefore called,
in practice, a head. Thus, Z represents the potential head, position head,
or elevation head; ply represents the pressure head; and V2/2g the velocity
head. As the liquid would rise to a height of Z 4- {ply) in a piezometer
when it is introduced at right angles to the direction of flow in a flowing
liquid, the sum Z + {ply) is termedpiezometric head. It is also called the
hydraulic head. The sum Z + {ply) + {V2/2g) is the total head.
Between sections 1-1 and 2-2 of a pipeline, a pump may be fitted to sup-
ply energy or a turbine may be fitted to extract energy. Representing such
externally supplied or extracted energy per unit mass by Ee{Ee = Ep for
pumps and — ET for turbines) and head by he{he = hp for pumps and — hT
for turbines), the generalized forms of the Bernoulli's equation for unit
mass and unit weight are, respectively,

gZt + ^ + y + Ee = gZ2 + ^ + - ^ + ELl_2 (1.17)

The energy supplied by a pump or given to a turbine can be evaluated

by considering the power and the efficiency. Power is the capacity to do
work per unit time and therefore has the dimensions of FLT"1 or ML 2 T~\
The unit for power is the watt, W (1 W = 1 J/s). Thus, the power P neces-
sary for a pump having efficiency rj (eta) to raise yQ newtons of liquid per
second through a height hP meters is given by

PP™P = — watts (1.19)

18 INTRODUCTION

Similarly, the power gained by a turbine of efficiency rj through yQ new-

tons of liquid per second under a head difference of hT meters is given by

/'.urhinc = jQhrV watts (1.20)

Example 1.8

A pipe carrying water tapers from 200 mm diameter at section 1 to 100

mm diameter at section 2. The velocity at section 1 is 1.8 m/s. The
pressures at sections 1 and 2 are 80 kN/m2 gauge and 24.3 kN/m2 vacuum,
respectively. If section 2 is 6 m above section 1, find the discharge through
the pipe in liters per minute, the velocity at section 2, the direction of flow,
and the loss of head.

Solution

Discharge = AXVX = | ( 0 . 2 ) 2 X 1.8 = 0.05655 m3/s

.". Discharge = 0.05655 x 1000 X 60 = 3,392.9 L/min

(0 2) X
Q \ ' ^
Velocity at section 2, V2 = ^ = = 7.2 m/s

Considering the datum through section 1, the total head at section 1,

80 X 1,000 1.82
= + +
° 9,810 Y^9Jl =8J2° m
° f Water

Similarly, for the total head at section 2,

7 2g

/ 24.3 X 1,000 7 22
= 6 + I - " ' " ~ 1' I + ~ — 1 r r ^ 7 = 6.165 m of water
\ ~ 9,810
 Energy Principle 19

Since the total head at section 1 is greater than that at section 2, the
direction of flow is from section 1 to section 2.
Loss of head = /?, - h2 = 8.320 - 6.165 = 2.155 m

Example 1.9

A pump lifts 0.15 m3 of water per second. The diameter of the pump in-
let is 300 mm and the vacuum pressure at the inlet is 150 mm of mercury.
The outlet diameter of the pump is 200 mm and the pressure at the outlet
is 137.4 kN/m2. If the exit of the pump is 1 m higher than the inlet and the
power of the motor driving the pump is 35 kW, find the efficiency of the
pump.

Solution

At inlet, V, = — = °15 = 2.122 m/s

:
Similarly, at outlet, V2 = = 4.775 m/s
I X 0.2-
The pressure head at the inlet is 150 mm of mercury, vacuum.
13 6
.'. — = - T^L X - = -2.04 m of water
7 1,000
Applying Bernoulli's equation to the inlet (section 1) and the outlet (sec-
tion 2) and taking the datum through the inlet,

y 2g 7 2g
2
2.122 137.4 4.7752
•"• ° + ( - 2 0 4 ) +
19^ +
*' = ' + W +
19^
.-. hp = 17.979 m
™ ,. . . . 0.15 X 1000 X 9.81 X 17.979
The energy supplied by the pump = Trwi

= 26.456 kW

.". efficiency of the pump = —rr— = 0.7559, i.e., 75.59 percent

20 INTRODUCTION

1.5.3 HYDRAULIC AND ENERGY GRADE LINES

If at each point along a pipeline, the pressure head ply is plotted as a

vertical distance above the centre line, the locus of end points gives the
pressure head line or the hydraulic grade line, HGL. More generally, it is
the plot of the piezometric head Z + (p/y) above a convenient datum
against length along the pipe. It is, therefore, sometimes known as the
piezometric line. When the pressure in the pipe is greater than the at-
mospheric pressure, i.e., when the gauge pressure is positive, the HGL is
above the centre line of the pipe. When the gauge pressure at a point in the
pipeline is zero, the HGL and the pipe centre line intersect. When the
gauge pressure in a pipeline is subatmospheric, ply is negative and the
HGL is below the pipeline.
When the total head Z + (p/y) + (V2/2g) at each point is plotted
above a convenient datum against length along the pipe, the locus of end
points gives the total head line. As seen previously, the total head is nu-
merically the same as the energy possessed by a unit weight and therefore
the total head line is also known as the energy grade line, EGL. The EGL
is always above the HGL by the velocity head V2/2g, and these two lines
would be parallel, a vertical distance V2/2g apart, if the pipeline is of uni-
form cross section throughout.

1.6 MOMENTUM PRINCIPLE

The momentum principle for a fluid mass can be derived from Newton's
second law of motion which states that the net force acting on a body in
a fixed direction is equal to the rate of increase of the momentum of the
body in that direction.
Consider the control volume bounded by a stream tube and two cross
sections 1-1 and 2-2 as shown in Figure 1.3. Let the fluid enter through sec-
tion 1-1 and leave through section 2-2, with average velocities K, and V2,
respectively. For steady flow, the mass of fluid entering at section 1-1 is
equal to that leaving at section 2-2, and this, for unit time is QQ. Thus,
from Newton's second law of motion, the net force in the ^-direction, i .e.,
Fx is given by

F, = QQ (V2X - Vlt) (1.21)

in which the subscript x denotes the jc-direction. Equation (1.21) is the

 Momentum Principle 21

Figure 1.3 Control volume for momentum equation.

momentum equation for the ^-direction. Similarly, the momentum equa-

tions for the y- and z-directions are, respectively.

Fy = QQ Wly - Vly) (1.22)

F. = QQ(V2Z - Vlz) (1.23)

The general momentum equation can be written as

F = QQ (V2 - V.) (1.24)

in which F, V2, and V, are vectors.

The momentum equation can be used to determine the force induced by
moving fluids on pipe bends and elbows, reducers and enlargers, and pipe
junctions.

Example 1.10

Water is flowing at a rate of 0.35 m3/s through a 60° reducing bend in

a horizontal plane as shown in Figure: Example 1.10. The diameter and
pressure at the entrance of the bend (section 1-1) are 300 mm and 294 kPa,
22 INTRODUCTION

respectively, and the diameter at the exit end (section 2-2) is 150 mm. De-
termine the force required to keep the bend stable. Neglect the loss of head
in the bend.

Solution

From the continuity equation,

0.35
at the entrance, V, = = 4.951 m/s and

0.35
at the exit, V2 = = 19.806 m/s
7T
(0.15) 2

From the energy equation, since Z, = Z 2 ,

P2 = ^Q(V\ - VI)

= 294,000 + 500 (4.951 2 - 19.8062)

= 1.101 x 105 Pa
 Momentum Principle

By the momentum equation in the ^-direction,

p,At - Fx - p2A2 cos 60° = QQ (V2 COS 60° - K,)

IK \
:. 294 x 103\-r x 0.3 2 \ - Fx - 1.101 x 105

•K
x j x 0.15 | cos 60° = 1000

X 0.35(19.806 cos 60° - 4.951)

i.e., 20,782 - Fx - 973 = 1,733 or Fx = 18,076 N

In the j-direction

Fy - p2A2 sin 60° = QQ (V2 sin 60° - 0)

or F, = 1000 X 0.35 (19.806 sin 60°) + 1.101 X 105

x f e x 0.152 j sin 60° = 6,003 + 1,685 = 7,688 N

.'. The force required to keep the bend stable,

F = (18,0762 + 7,6882)1/2 = 19,643 N, acting in the direction

tan"1 7688/18,076, i.e., 23° 0 2 ' as shown in the figure

 Frictional Head Loss in Pipes

2.1 INTRODUCTION

HEN a real fluid flows through a pipe, part of the total energy of
W the fluid is spent in maintaining the flow. This used-up energy is con-
verted to thermal energy due to internal friction and turbulence. Such
a conversion, which is the loss of energy as far as its utility is concerned,
is usually expressed in the form of head of liquids and therefore termed
head loss.
The head loss in a pipe is classified into two categories:

(1) Head loss due to friction, i.e., frictional head loss

(2) Head loss due to minor appurtenances, i.e., minor head loss

The frictional head loss in a pipe is due to the viscosity of the fluid and
the turbulence of the flow and is present throughout the length of the pipe.
As the frictional head loss in a long pipe is relatively larger than the other
head losses, the frictional head loss is also termed major head loss.
When there is a sudden or gradual change in the boundaries of the fluid
or when there is a local obstruction to flow, the flow pattern changes. This
results in a change in the magnitude, direction or distribution of the veloc-
ity of flow. Such a change introduces additional head loss which is of a
local nature and is usually much less than the frictional head loss in a long
pipe. It is therefore termed minor head loss.
It must, however, be remembered that the classification into major and
minor head losses is rather relative, and therefore for a pipeline of small
length and having many minor appurtenances, the total minor head loss
can be more than the frictional head loss. In water distribution networks,
however, the pipelines are of considerable length and therefore the terms
major head loss and minor head loss can be used.
Frictional head loss is considered in this chapter and the head loss due
to other causes, i.e., minor head loss, is considered in the next chapter.
25
26 FRICTIONAL HEAD LOSS IN PIPES

2.2 DARCY-WEISBACH FORMULA

The frictional head loss ht for incompressible flow in a pipe depends on

the fluid properties such as the density Q and viscosity /JL, the flow charac-
teristic such as the average velocity V, the pipe characteristics such as the
length L and the internal diameter D, the pipewall roughness characteris-
tics such as the size of the roughness projections e (dimensions, L), ar-
rangement or spacing of the roughness projections e' (dimensions, L), and
a dimensionless form factor m depending upon the shape of the individual
roughness elements. Using dimensional analysis, it can be shown that [9]

' D ' D '

in which 0 represents a function of. As it is extremely difficult to quantita-

tively consider the effect of e' and m, their effect is either ignored or is in-
cluded indirectly by considering equivalent size e. Therefore, Equation
(2.1) can be reduced to

The term VDQ//JL is the Reynolds number Re. The dimensionless param-
eter elD is known as the relative roughness.
Introducing a coefficient of friction / such that

j , M (2.3)
Equation (2.2) can be written as

J. Weisbach [10] was the first to write a resistance equation for pipes in
the form of Equation (2.4). As H. P. G. Darcy carried out considerable
work on pipe flow, his name is also associated with Equation (2.4). There-
fore, Equation (2.4) is now commonly known as the Darcy-Weisbach for-
mula.
For noncircular pressure conduits, D in Equation (2.4) is replaced by AR
in which R is the hydraulic radius defined as the ratio of the cross-sectional
area of flow A, and the wetted perimeter P, i.e., R = A/P.
 Nikuradse's Experiments on Artificially Roughened Pipes 27

Sometimes, it is more convenient to use the Darcy-Weisbach formula in

a form in which hf is given in terms of the discharge Q rather than the aver-
age velocity V. Replacing V by QIA and taking A = irD2/4 for circular
pipes, Equation (2.4) on simplification becomes

16 fLQ2 SfLQ2
h( = (2.5)

Taking g = 9.81 m/s2, Equation (2.5) can further be simplified to

fLQ2
hf = (2.6)
12.1 D 5

2.3 NIKURADSE'S EXPERIMENTS ON ARTIFICIALLY

ROUGHENED PIPES

J. Nikuradse [11] performed experiments extensively on smooth and ar-

tificially roughened pipes. In these experiments, sand particles of uniform
size were used to obtain artificially roughened pipes. The relative
roughness e/D (e is the diameter of sand particles) varied from 1/30 to
1/1,024. The coefficient of friction/was plotted against Re for different e/D
values, on a log-log scale as shown in Figure 2.1. Even though the effect

o-io

0O8

0-06

0-04

0-02

- Laminar Transitional
—flow »i flow

0-01

Figure 2.1 Nikuradse's experiments on artificially roughened pipes.

28 FRICTIONAL HEAD LOSS IN PIPES

of spacing and shape aspects \e' ID and m in Equation (2.1)] were not stud-
ied in Nikuradse's experiments, the experiments did prove the validity of
the relative roughness concept.
The following conclusions can be drawn from Nikuradse's experiments
(Figure 2.1):
(1) When Re < 2,000, i.e., when the flow is laminar,/is independent
of elD and depends only on Re.
(2) When 2,000 < Re < 4,000, i.e., when the flow is in a state of tran-
sition from laminar to turbulent flow, / increases as Re increases.
(3) When Re > 4,000, turbulence sets in and the effect of elD may be
felt. However, as long as the pipe wall roughness is within the lami-
nar sublayer, the pipe behaves as a smooth pipe. The flow is termed
smooth turbulent flow. The value of/depends on Re only and is
given by the smooth pipe curve.
(4) Depending upon the value of elD, the curve branches out from the
smooth pipe curve. The larger the value of elD, the earlier the
branching takes place. As the value of Re increases, the effect of elD
also increases. This continues until the boundary roughness has its
complete effect on the flow. This flow is termed transitional turbulent
flow. In transitional turbulent flow, the value of/depends on both Re
and elD.
(5) For very large values of Re, the effect of viscosity is negligible, and
/becomes independent of Re and remains constant for a particular
value of elD. This flow is termed rough turbulent flow and /depends
only on the elD value.
Stanton, in collaboration with Pannell, conducted experiments on sev-
eral pipes of different diameters and materials using several fluids. There-
fore, the diagram as in Figure 2.1 is sometimes known as the Stanton dia-
gram or the Stanton-Pannell diagram.

2.4 MOODY DIAGRAM

Nikuradse carried out experiments on artificially roughened pipes hav-

ing uniform sand grain roughness. The roughness pattern of the commer-
cial pipes, however, is different. Therefore, Nikuradse's curves (Figure 2.1)
cannot be used directly in practice to evaluate the coefficient of friction/
for commercial pipes. As shown by Colebrook |12] the transition zones of
Nikuradse's curves do not match with those of actual pipes. However, by
introducing the concept of equivalent surface roughness, it is possible to
use Nikuradse's results for commercial pipes. The equivalent surface
 Moody Diagram 29

roughness is the roughness which provides the value of/that is the same
as the limiting value for the rough turbulent flow in Nikuradse's artificially
roughened pipes. The values of the equivalent surface roughness e for
pipes of some common materials are given in Table 2.1 [9,13-15].
Johnson [16] provided a diagram for commercial pipes using several
nondimensional groups. Rouse [17] presented a graph of \l\ff versus Kar-
man number K = Re \ff = D3n \l2gSlv upon which the Colebrook
transition curves [12] for various values of the relative roughness were
superimposed. Moody suggested that Rouse should convert the diagram to
the more customary / versus Re form. In the belief that the Prandtl-
Karman parameters were the more significant ones and therefore the form
used by him was superior to that suggested by Moody, Rouse refused to
convert the diagram [18]. Whereupon Moody [19] published his proposed
form in the widely read, "Transactions of the American Society of
Mechanical Engineers." This diagram, shown in Figure 2.2 is now univer-
sally known as the Moody diagram and gives the values of/ for different
values of Re and elD.
Several other investigators such as Powell [20], Ackers [21], Thiruveng-
dam [22], Ramalho, Tiller and Berry [23], Bewtra [24], Barr and Smith
[25], Asthana [26], Li [27) and Barr [28] also presented charts to tackle
different flow problems. However, the Moody diagram is more popular
and widely used in practice.
It is worthwhile to remember that the concept of equivalent surface
roughness on which the Moody diagram is based has some limitations. For
example, the concept of equivalent surface roughness implies that only the
height of surface irregularities significantly affects the flow. However, the

TABLE 2.1. Equivalent Surface Roughness for Different Pipe Materials.

Equivalent Surface
Pipe Material Roughness e, mm

Riveted steel 0.90-9.0

Concrete 0.30-3.0
Wood stave 0.20-0.90
Cast iron 0.26
Galvanized iron 0.15
Asphalted cast iron 0.13
Commercial steel, wrought iron 0.05
Uncoated asbestos cement, PVC pipes with waviness, n C\A
prestressed concrete
PVC 0.0021
Drawn tubing of aluminium, brass, copper, lead, glass and
plastic; coated asbestos-cement; spun cement-lined; spun 0.0015
bitumen-lined
 .Criti- Trqn-\
cal jsition
zone'TioneTzone Complete turbulence,'rough pipes

<u|Q

0-000 01
.8
"6 8103 2 4 6 8B 4 2 3 4 6 4 6 4 6 8»7 4 6 810

Reynolds number Re = ^ j

Figure 2.2 Moody diagram.

 Friction Coefficient Relationships 31

spacing of the irregularities is also important, especially for rough turbu-

lent flow. Similarly, the equivalent surface roughness ignores the effect of
the shape of the irregularities. Another factor which may affect the value
of/in large diameter pipes is the surface waviness produced by transverse
ridges. Further, the roughness of many materials is variable and frequently
increases with age. Thus, the Moody diagram is, at best, an approxima-
tion, and therefore, accurate prediction of the friction factor and thereby
the frictional losses is rather difficult to achieve.

2.5 FRICTION COEFFICIENT RELATIONSHIPS

Friction coefficient relationships applicable to the five regions of flow

(Section 2.3) are now described.

2.5.1 LAMINAR FLOW

For laminar flow (Re < 2,000) the value o f / i s given by

This value of/is the same as that obtained theoretically from the Hagen-
Poiseuille equation for laminar flow.

2.5.2 TRANSITIONAL FLOW

For transitional flow (2,000 < Re < 4,000) the value of/is uncertain.
However, from experiments it is found to lie between 0.03 and 0.08 for
commercial pipes.

2.5.3 SMOOTH TURBULENT FLOW

Using the results of the experiments performed by Saph and Schoder in

1903, Blasius [29] suggested a relationship

f- ^ ^ (2

which is valid in the range 4,000 < Re < 10s.

32 FRICTIONAL HEAD LOSS IN PIPES

Prandtl on the basis of his and Karman's experiments suggested, in 1911,

a relationship

•jz: = 2 log Re yff - 0.8 (2.9a)

which also can be written as

(£)
Equation (2.9) is known as Karman-Prandtl equation for smooth turbu-
lent flow. This equation is implicit and the value of/must be obtained by
trial and error.
Several investigators such as Schimemi, Barnes, Blair, Lamont [30] and
Ger and Holley [31] have suggested relationships similar to that of Blasius
for smooth turbulent flow.

2.5.4 TRANSITIONAL TURBULENT FLOW

For the transitional turbulent flow,/is a function of Re and elD. For this
flow Colebrook [12] suggested a relationship, known commonly as the
Colebrook-White equation:

(2IOa)

or

= ~2 log ( I ? + 5TD) (210b)

2.5.5 ROUGH TURBULENT FLOW

In the rough turbulent flow;/is independent of Re and is only a function

of e/D. For this type of flow Prandtl gave a relationship

(2.11a)
 Explicit Relationships for Friction Coefficient 33

or

which is known as the Karman-Prandtl equation for rough turbulent flow.

It may be noted that Equation (2.10) is nearly the same as Equation (2.9)
when the flow is smooth turbulent and nearly the same as Equation (2.11)
when the flow is rough turbulent. Therefore, the Colebrook-White equa-
tion, Equation (2.10), can be used to cover the entire turbulent flow region.

2.6 EXPLICIT RELATIONSHIPS FOR FRICTION COEFFICIENT

The Karman-Prandtl equation for rough turbulent flow, Equation (2.11),

gives the value of the friction coefficient/for rough turbulent flow directly.
However, the Karman-Prandtl equation for smooth turbulent flow, Equa-
tion (2.9), and the Colebrook-White equation for transitional turbulent
flow, Equation (2.10), are implicit and therefore a trial and error procedure
must be adopted for determining the value off. Several investigators, how-
ever, have proposed explicit relationships for determining the value of/.

2.6.1 SMOOTH TURBULENT FLOW

Techo, Tickner and James [32] suggested an explicit relationship

/ = {0.86859 In [&>/(!.964 In (Re) -3.8215)]}- 2 (2.12)

for the Karman-Prandtl equation, Equation (2.9). Chen [33] suggested a

simpler explicit relationship
/ = (2 log [(4.52/Re) log (Re/7)]]-2 (2.13)
Both these explicit relationships [Equations (2.12) and (2.13)] are good
approximations of Equation (2.9) and give/within an accuracy of ±0.1
percent [34].

2.6.2 TRANSITIONAL TURBULENT FLOW

Moody [19] suggested a relationship

[ / e 10 6 \ 1 / 3 l
/ = 0.0055 1 + 1 20,000 ^ + -j^-l (2.14)
34 FRICTIONAL HEAD LOSS IN PIPES

which has an accuracy of ± 5 percent for 4,000 < Re < 107, and
e/D < 0.01. Nahavandi and Catanzaro [35] used this relationship in pipe
network analysis.
Wood [36] suggested a relationship

f = a + b Rec (2.15)

in which a = 0.094 (e/D)° " 5 + 0.53 (e/D); b = 88 (e/D)044; and c =

1.62 (e/D)0134. This relationship is valid for Re > 10,000; and 10"5 <
(e/D) < 4 X 10~2. It gives an accuracy within ± 4 percent.
Barr [37] proposed

(2.16)

which was later modified by him [38] to

Churchill [34,39] proposed

(2.18)

which has an average accuracy of ±0.58 percent for 4000 < Re < 108
and 0 < e/D < 5 X KT2 [34].
Jain [40] proposed

(p)
Swamee and Jain [41] proposed

The Jain and Swamee-Jain relationships have an accuracy of ± 1 per-

cent for 5000 < Re < 108,and 10'6 < e/D < lO"2. Collins [42] has con-
firmed them as excellent approximation of the Colebrook-White equation.
Converting Equation (2.18) proposed by Churchill [39] in the form of
 Explicit Relationships for Friction Coefficient 35

Equation (2.20) proposed by Swamee and Jain, Equation (2.18) becomes

/ = 7 / ^ ^ C^M, (2-2D

Similarly, Equation (2.19) proposed by Jain [40] changes to

5 .72 \ V
e°>}\
Re°

Thus, the explicit relationships of Churchill [Equations (2.18), (2.21)],

Jain [Equations (2.19), (2.22)] and Swamee and Jain [Equation (2.20)] are
almost the same. Since Churchill had proposed his relationship [Equation
(2.18)] in 1973 [34], the credit of proposing this type of explicit relation-
ship first goes to Churchill. However, the form of Equation (2.20) appears
to be more convenient than that of Equations (2.18) and (2.19) for direct
use in practice. Further, Equation (2.20) is also now included in a text-
book of fluid mechanics [9].
Zigrang and Sylvester [43] proposed

e 5.02 e 5.02 e
^ log ^ ^
Re log (3 JD
•^T = -2 log ^ ^ ^ ^ ^
3 JD Re 3 JD
(2.23)

which is rather complicated but extremely accurate for 4,000 < Re <
108, andO < (e/D) < 5 X 10"2, and thus gives an excellent explicit form
of the Colebrook-White equation.
Haaland [44] proposed
6 9
• ' - " " (2.24)

which has an average accuracy of ±0.33 percent for 4,000 < Re < 108,
and 0 < (e/D) < 5 X 10"2.
Chen [34] proposed

which has an average accuracy of ±0.30 percent for 4,000 < Re < 108
and 0 < (e/D) < 5 X 10"2.
36 FRICTIONAL HEAD LOSS IN PIPES

2.6.3 ALL FLOW REGIMES

To cover all flow regimes, Churchill [39] proposed

(2.26)
(A + BY

where

1
A = 2.457 In 0 9
7\ - 0.27 e
,Re) ) D

and

B= /37,530y

Equation (2.26) is rather complicated and has an average accuracy

of ±0.45 percent.
Chen [33,341 has reviewed and compared the accuracy of these explicit
formulas. They are compared herein for {elD) = 1 x 10"2, 1 X 10~\
and 1 X 10"4 and Re - 104, 106, and 108, as shown in Table 2.2.
Equation (2.23) proposed by Zigrang and Sylvester exactly matches
the Colebrook-White equation, [Equation (2.10b)]. However, as the
Colebrook-White equation is itself in error by 3-5 percent when compared
with actual experimental data, exact matching of the/value obtained by
the explicit relationship with that from the Colebrook-White equation may
not be necessary. Therefore, in practice, any of the simpler versions such
as the ones proposed by Barr [Equation (2.17)], Churchill [Equation
(2.18)], Jain [Equation (2.19)], Swamee and Jain [Equation (2.20)],
Haaland [Equation (2.24)] or Chen [Equation (2.25)] can be used. Simi-
larly, even though Equation (2.26) proposed by Churchill encompasses all
the flow regimes, it is simpler to use separate formulas for laminar and tur-
bulent flow regimes.

2.7 EMPIRICAL FORMULAS

Even though the Darcy-Weisbach head loss formula is dimensionally ho-

mogeneous, some empirical formulas are popular and widely used in prac-
tice.
 TABLE 2.2. Comparison of Different Explicit Friction Coefficient Relationships with Colebrook-White Relationship.
Colebrook- Swamee- Zigrang-
Reynolds White Moody Wood Barr Churchill Jain Jain Sylvester Haaland Chen Churchill
number Eq. (2.10b) Eq. (2.14) Eq. (2.15) Eq. (2.17) Eq. (2.18) Eq. (2.19) Eq. (2.20) Eq. (2.23) Eq. (2.24) Eq. (2.25) Eq. (2.26)

e/D == 0.01
4
10 0.04313 0.04232 0.04235 0.04393 0.04406 0.04398 0.04404 0.04313 0.04304 0.04398 0.04404
106 0.03796 ' 0.03772 0.03872 0.03801 0.03801 0.03796 0.03801 0.03796 0.03804 0.03801 0.03799
108 0.03790 0.03766 0.03865 0.03791 0.03791 0.03785 0.03791 0.03790 0.03798 0.03791 0.03788
e/D = 0.001
104 0.03238 0.03263 0.03179 0.03249 0.03270 0.03263 0.03267 0.03237 0.03217 0.03257 0.03269
106 0.01994 0.02067 0.02099 0.02004 0.02003 0.02001 0.02003 0.01994 0.01994 0.02004 0.02002
10s 0.01964 0.02043 0.02043 0.01964 0.01964 0.01962 0.01964 0.01964 0.01968 0.01964 0.01963
e/D = 0.0001
4
10 0.03104 0.03120 0.03176 0.03096 0.03118 0.03111 0.03115 0.03102 0.03099 0.03105 0.03118
106 0.01344 0.01343 0.01415 0.01354 0.01351 0.01350 0.01351 0.01344 0.01333 0.01353 0.01351
103 0.01200 0.01244 0.01215 0.01201 0.01201 0.01200 0.01201 0.01200 0.01202 0.01201 0.01201
 FRICTIONAL HEAD LOSS IN PIPES
3o

2 7.1 HAZEN-WILLIAMS FORMULA

An empirical formula widely used in water supply engineering for the

flow of water through pipes is due to G.S. Williams and A. Hazen [45],
This relationship, known as the Hazen-Williams formula, is

V = 0.849 CHW fl°63 S° •s4 (2.27)

in which V = average velocity of flow in m/s; CHW = Hazen-Williams

(HW) coefficient; R = hydraulic radius ( = AIP) in m; and 5 = slope of
the energy line ( = hf/L). For circular pipes V = AQI-KD2; and R = D/4.
Substituting these values and taking S = hf/L, Equation (2.27) on simpli-
fication becomes

h
10.68 LQl
f =

in which L and D are in meters and Q in cubic meters per second. In prac-
tice, the pipe diameter D may also be given in millimeters or centimeters
and the discharge Q in liters per minute, million liters per day and several
other units. Instead of converting the pipe diameter to meters and pipe dis-
charge to cubic meters per second for the entire network, it may be ad-
visable to replace 10.68 in Equation (2.28) by an appropriate value. Such
values are given in Table 2.3.
The Hazen-Williams formula is derived for R = 0.3 m and 5 = 1/1000.
Therefore, the HW coefficient CHW is dependent on the R and 5 values
and also the flow conditions. The CMW values for common pipe materials
as recommended by Lamont [15,30] after examining 372 records are given
in Table 2.4. These values are for a velocity of flow of about 0.9 m/s.
These values should be corrected as shown in Table 2.5 when the veloc-
ities are considerably above or below 0.9 m/s. Lamont emphasizes that the
HW formula is not suitable when the CHW values are appreciably be-
low 100.
The HW formula is more suitable for smooth pipes and therefore for
medium to large diameter new pipes [30]. Although it is not quite suitable
for old pipes, the formula is used in practice by reducing the CHW values,
as described later in Section 2.J2.
Scobey [46] assembled results of more than 1000 determinations made
by himself and others on 147 pipelines ranging in diameters from 25 mm
to 5470 mm and suggested a relationship similar to the HW formula but
with the exponent of S as 0.53 [47]. White [48] on the other hand, from
several tests on the glazed and unglazed pipelines, found an average value
of 0.54 for the exponent of 5, thereby confirming the exponent of S in the
HW formula [48].
 TABLE 2.3. Values of Constant in Hazen-Williams Formula

Discharge, in
Pipe diameter,
D, in m3/s m 3 /min m3/h m3/d Us L/min ML/d
3 6 9 5
m 10.68 5.438 x 10" 2.769 x 10" 7.694 x 10" 2.969 x 10" 1.512 x 10-s 2.768 x 10-3
cm 5 .869 x 101° 2.988 x 107 1.522 x 104 42.28 1.631 x 10s- 83.07 1.521 x 107
6 6
mm 4 .351 x 1015 2.215 x 1012 1.128 x 109 3.134 x 10 1.209 x 1Qio 6 .158 x 10 1.128 x 1012
 TABLE 2.4. Values of Hazen-Williams Coefficient for Clean New Pipes of Different Materials.

C H W value for pipe diameter in millimeters

Pipe Material 75 150 300 600 1200

Uncoated cast iron 121 125 130 132 134

Coated cast iron 129 133 138 140 141
Uncoated steel 142 145 147 150 150
Coated steel 137 142 145 148 148
Wrought iron 137 142 — — —
Galvanized iron 129 133 — — —
Coated spun iron 137 142 145 148 148
Uncoated asbestos cement 142 145 147 150 —
Coated asbestos cement 147 149 150 152 —
PVC wavy 142 145 147 150 150
Concrete 69-129 79-133 84-138 90-140 95-141
Prestressed concrete — — 147 150 150
Spun cement-lined; spun bitumen-lined,
PVC, brass, lead, copper 147 149 150 152 153
Newly scraped mains 109 116 121 125 127
Newly brushed mains 97 104 108 112 115
 Empirical Formulas 41

TABLE 2.5. Correction Factors for Hazen-Williams Coefficients.

Velocity Below Velocity Above
0.9 m/s for Each 0.9 m/s for Each
Halving, Rehalving of Doubling, Redoubling
Velocity Relative to of Velocity Relative to
CHW Value at 0.9 m/s 0.9 m/s

CHW below 100 add subtract

5% to CHW value 5% from CHw value
CHW from 100 to 130 add subtract
3% to CHW value 3% from CHw value
CHW ^om 130 to 140 add subtract
1 % to CHW value 1 % from CHW value
CHW above 140 subtract add
1 % from CHW value 1 % to CHW value

2.7.2 MODIFIED HAZEN-WILLIAMS FORMULA

Even though Hazen-Williams formula is obtained for R = 0.3 m and

S = 1/1000, it is used by design engineers over wide ranges of diameters
and friction slopes. This may result in an error as high as ± 3 0 percent in
the evaluation of the velocity of flow [49]. Therefore, after comparing the
HW formula with the rational and dimensionally homogeneous Darcy-
Weisbach formula, Jain, Mohan, and Khanna [49] modified the HW for-
mula. For water temperature of about 20°C (v = 1 X 10"6 m2/s) and
g = 9.81 m/s2, they proposed the modified form of the HW formula as

V = 143.5 C R fl 0 6 5 7 5 S 0 5 5 2 5 (2.29)

which in terms of the head loss hf becomes

/ /}1.8099

f ~ 0,0,4 62C 1 8 0 9 9 D48099

In these formulas, CR is the coefficient of roughness given by

— 2 (2) 0 5 I e 1 78
CR = 1Og +
3.83 Re"^ i
CR = 1 for smooth pipes and CK < 1 for pipes other than smooth. The
CR value decreases as the velocity of flow in the pipe increases. Jain, Mo-
han, and Khanna have also presented tables giving CR values for cast iron,
concrete and steel pipes having diameters ranging from HX) mm to 2000
42 FRICTIONAL HEAD LOSS IN PIPES

mm and for velocities ranging from 0.3 m/s to 6.0 m/s. They have pre-
sented a chart for the effect of water temperature on CR values and have
also presented a nomograph for direct estimation of the loss of head for
flow of water at about 20°C O = 1 X 10"6 m2/s) for smooth pipes for
different diameters and discharges [49]. As the use of the modified HW
formula requires determination of CR values which in turn depend upon
the Reynolds number and relative roughness of the pipe, it is preferable to
use the rational, dimensionally homogeneous and simple Darcy-Weisbach
formula for which the coefficient of friction can be explicitly determined
as described earlier.

2.7.3 MANNING FORMULA

Another empirical relationship which is sometimes used for pipe flow is

the Manning formula, given as

V = - R2'3 S1'2 (2.32)

N
in which N = Manning roughness coefficient. Substituting for V, R, and
5 as before, Equation (2.32) on simplification can be written as

The usual range of the values of N for common pipe materials is given
in Table 2.6 [13]. Manning formula is more suitable for rough pipes with
N > 0.015 as stated by Lamont [30] and therefore it is more suitable for
open channel flow rather than pipe flow.
Powell [50] and Williams [51] have"shown that the Manning formula, as
is presently used, was not proposed by Manning. Gauckler in 1867, Hagen

TABLE 2.6. Values of Manning's Coefficient for Different Pipe Materials.

Pipe Material Manning's Coefficient, N

PVC 0.008-0.011
Brass, copper, glass, lead, prestressed concrete 0.009-0.012
Concrete 0.010-0.017
Wood stave 0.011 -0.013
Welded steel 0.012-0.013
Coated cast iron • 0.012-0.014
Uncoated cast iron 0.013-0.015
Galvanized iron, riveted steel 0.015-0.017
 Comparison of Head Loss Formulas 43

in 1881, Valot in 1887, Thrupp in 1888, Manning in 1891, Foss in 1894,

Crimp and Bruges in 1895, Tutton in 1896, Yarnell and Woodward in 1920,
and Strickler in 1923 provided formulas in which V was expressed as a
function of R2'3 SU2 for certain flow conditions [50,51]. Flamant first at-
tributed the formula to Manning in 1891. Orbeck et al., introduced the ex-
pression "Manning's A7" in literature in 1916. This was rather an ironic de-
velopment as Manning had clearly rejected the use of the coefficient N
[51].
The present formula is known in different parts of the world under
various names such as the Gauckler formula, Hagen formula, Strickler
formula or Manning formula. However, Hagen derived the formula only
for the Ganga canal, and the formula was well established by the time of
Strickler's 1923 paper. As Gauckler was the first to suggest the present
form of the formula, Williams advocated that Equation (2.32) should bet-
ter be named the Gauckler-Manning formula [51]. This suggestion, how-
ever, is not well received and Equation (2.32) continues to be known as the
Manning formula.

2.8 COMPARISON OF HEAD LOSS FORMULAS

The Darcy-Weisbach, Hazen-Williams, and Manning formulas are now

compared. The coefficient of friction/is obtained in terms of the Hazen-
Williams coefficient CHW and Manning's coefficient N.
From Equations (2.6) and (2.28),

_ fLQ2 _ 10-68L6 1 8 5 2
1
~ 12.ID ~ Cligf2*)4-87
5

from which
013
=
129.2P
129.2P
' /""1.8521 /TO. 148 ^ ' )
Q°

Expressing Q as irReDu/4 and taking v = 1 X 10"6 m2/s, Equation

(2.34) becomes

^ ~ C\^2Re° 148
£>° °18 (2
' -*

Similarly, from Equations (2.6) and (2.33),

_
10.29N ;
' ~ 12.ID5 ~ D1
44 FRICTIONAL HEAD LOSS IN PIPES

from which

/ = mDuT ( 2 - 36 )
From Equation (2.35) it is clear that/depends upon Re to some extent
and D to a lesser extent in addition to the CHW value. Therefore, the prac-
tice of using the same value of CHW for different diameters and discharges
is not proper. However, if the CHW value is selected using Table 2.4 and
modified for velocity using Table 2.5, better results can be obtained. Simi-
larly, the modified HW formula would also give better results.
Equation (2.36) shows that the Manning's coefficient N is a function of
the pipe diameter and is independent of Re. Therefore, it would be proper
to use the Manning formula for rough turbulent flow only.

2.9 GENERAL HEAD LOSS FORMULA

The frictional head loss in a pipe can be expressed by a general head loss
formula,

h, = RQ" (2.37)

in which R = pipe resistance constant, and n = exponent. The expres-

sions of R and n for the Darcy-Weisbach, Hazen-Williams, and Manning
formulas are given in Table 2.7. As Hazen-Williams coefficient CHW and

TABLE 2.7. Values of R and n in General Head Loss Formula.

Head Loss
Formula Ra n

Darcy-Weisbach — or
12.1 D 5

10 68L 1
Hazen-Williams • 18S2n487 -852

.. . 10.29A/2/.
Manning — 2

a
R value is for L and D in meters and Q in cube meters per second.
 Simple Pipe Flow Problems 45

Manning coefficient N are generally assumed to remain constant for a

given pipe, the value of R remains constant for a given pipe and is indepen-
dent of the discharge through it. However, for the Darcy-Weisbach for-
mula, as the coefficient of friction/is a function of the Reynolds number,
the value of the pipe resistance constant R of a pipe depends on the dis-
charge through it.

2.10 SIMPLE PIPE FLOW PROBLEMS

Three types of simple pipe flow problems are encountered in practice.

They are as follows:
(1) Determination of the head loss hf
(2) Determination of the discharge Q
(3) Determination of the diameter D

When the Hazen-Williams or Manning head loss formulas are used, the
solution is quite simple because the unknown parameter hf, Q, or D can
be determined directly by using Equation (2.28) or Equation (2.33). How-
ever, when the Darcy-Weisbach head loss formula is used, the solution may
become complicated as the coefficient of friction / depends on the dis-
charge Q and diameter D. Therefore, even though the determination of hf
is simple as Q and D and therefore/can be easily determined, the deter-
mination of Q or D is rather complicated as / is unknown, and therefore
a trial and error procedure may be necessary.

2.10.1 DETERMINATION OF HEAD LOSS

Here Q, L, D, the fluid and therefore the kinematic viscosity v, and the
pipe material and therefore e, are known, and hf is unknown. Therefore,
using the known parameters, the Reynolds number Re and relative
roughness elD are determined. Using one of the explicit formulas, the
coefficient of friction/is determined. Later, using Equation (2.5) or Equa-
tion (2.6), the head loss h, is determined.

Example 2.1

Crude oil (e = 860 kg/m3, n = 0.008 Pa • s) flows through a 100-mm

diameter, 5-km long pipeline at a rate of 5 m3/h. Find the head loss due
to friction.
46 FRICTIONAL HEAD LOSS IN PIPES

Solution

v = Q= 5/3600
A

9 .302 x 1 0 - nWs
Q 860

Since Re < 2000, the flow is laminar. Therefore,

„ 64 64

and

8 x 0.03367 x 5000 x (5/3600)2

_
hf
~ ^ X 9.81 X

Example 2.2

Water flows through a 300-mm diameter, l-km long new cast iron pipe.
If the discharge is 0.1 m3/s, determine the head loss in the pipe by all three
head loss formulas.

Solution

1. For new cast iron pipe, e = 0.26 mm. Therefore,

e 0.26
= 8.667 x 10"4
D 300

Taking v — 1 X 10~6 m2/s for water,

«' - ^ " , x 0 4 3 X x°,'x . 0 - = 424 413

'
 Simple Pipe Flow Problems 47

Using the Colebrook-White relationship, Equation (2.10b)

1 „ , / 2.51 _ 8.667 x 1Q-4

+
// 3.7

Solving by trial and error, / = 0.01974. Using the Darcy-Weisbach for-

mula,

, fLQ2 0.01974 x 1000 x (0.1) 2

hf = TTTT^ = , r i x (0_3)5 = 6-714 m

Incidentally, using Churchill's explicit relationship, Equation (2.18),

/ = 0.01987, giving hf = 6.758 m; using the Swamee-Jain relationship,
Equation (2.20), / = 0.01987, giving h, = 6.758 m; from the Zigrang-
Sylvester relationship, Equation (2.23), / = 0.01974, giving h, = 6.714
m; from the Haaland relationship, Equation ( 2 . 2 4 ) , / = 0.01968, giving
hf = 6.693 m; and from the Chen relationship, Equation (2.25),
/ = 0.01989, giving hf = 6.765 m. Thus, as stated earlier, the Zigrang-
Sylvester explicit relationship g i v e s / and therefore hh the same values as
those obtained by the Colebrook-White relationship and therefore can be
used where greater accuracy is required. However, other simpler explicit
relationships also give nearly the same result and therefore any one of
them can be used in practice.
2. For the Hazen-Williams head loss formula, from Table 2.4 for new
cast iron pipe (uncoated) CHW = 130. Further, as Q is in m3/s and D in
mm, using Table 2.3,

_ 4.351 x 10' 5 LQ 1 8 5 2
f /""1.852 r»4.87

_ 4.351 x 1015 x 1000 x O.I 1 - 8 "

130 1852 x 300 487 ~ ' m

3. Using the Manning formula, Equation (2.33), and taking N = 0.013

for new cast iron pipe from Table 2.6

_ 10.29 X N2L _ 10.29 X (0.013) 2 X 1000

= 10.69 m
It is seen that, for this example, the DW and HW formulas give nearly
the same head loss while the Manning formula gives a higher value.
48 FRICTIONAL HEAD LOSS IN PIPES

2.10.2 DETERMINATION OF DISCHARGE

For this case, hf, L, D, v and e are known and the discharge Q is un-
known. From the known e and D values, the relative roughness elD is de-
termined. Rearranging the Darcy-Weisbach formula, Equation (2.5),

(2.38)

Further, the Reynolds number ite = (4Q)/(irDv). Substituting the value

of Q from Equation (2.38),

Re =
•KDV

or

(2.39)

Substituting the value of Re \ff from Equation (2.39) in the implicit

Colebrook-White relationship, Equation (2.10b) becomes

(2.40)
3.7 D

which is now explicit and gives the value of/directly. This value of/ is
substituted in Equation (2.38) to determine the discharge Q.

Example 2.3

A 150-mm diameter, 2000-m long PVC pipe connects two reservoirs

having a difference of 30 m in their water levels. Find the discharge
through the pipe. Assume v = 1.12 x 10"6 m2/s for water.

Solution

From Table 2.1, for PVC pipe, e = 0.0021 mm = 2.1 X 10"6 m. Fur-
ther, v = 1.12 X 10"6 m2/s, L = 2000 m, D = 0.15 m, and h, = 30 m.
Substituting these values in Equation (2.40),
 Simple Pipe Flow Problems 49

1 „ , /2.51 X 1.12 X 10-6

Vf"= - 2i l ulog
M 0.15
2000 2.1 X 10'6
+
19.62 X 0.15 X 30 3.7 X 0.15

from which / = 0.01538.

From Equation (2.38)

x
* " " ~" = 0.02994, say 0.03 m3/s
* 4

2.10.3 DETERMINATION OF DIAMETER

For this case, hf, Q, L, v and e are known and the diameter D is un-
known. Since the diameter D is unknown, both the relative roughness elD
and the Reynolds number Re are unknown. Therefore, the coefficient of
friction/is also unknown.
From a rearrangement of Equation (2.5)

<241)

in which C, = SLQz/ir2ghf, a known quantity. Similarly,

to = X I - § (2.42,

in which C2 = 4(2/7TJ', a known quantity.

Having obtained d and C2 values, the following procedure can be fol-
lowed for the solution of the problem:
(1) Assume a suitable value of/. (Usually/ = 0.02 serves as a good
starting value.)
(2) Obtain D from Equation (2.41).
(3) Using the obtained value of D, determine the value of Re from Equa-
tion (2.42).
(4) Find the relative roughness elD.
(5) For the obtained Re and elD values (steps 3 and 4, respectively), ob-
tain the new value of/by the Colebrook-White or any other explicit
formula.
(6) Compare the obtained and the assumed values of/. If they are suffi-
ciently close, terminate the procedure, otherwise use the obtained
50 FRICTIONAL HEAD LOSS IN PIPES

value and go to step 2. (Usually two or three iterations are suffi-

cient.)
(7) When the procedure is terminated, the value of D is the one obtained
in step 2.

Example 2.4

Find the diameter of a new 1500-m long cast iron pipe to carry water at
a rate of 3000 L/min, if the permissible head loss is 20 m.

Solution

3000
Q = 3000 L/min = ^ ^ = 0.05 m'/s

L = 1500 m, hf = 20 m, e = 0.26 mm = 0.00026 m, and assume

v = 1 X 10~6 m 2 /s.

From Equation (2.41),

D5 _ 8 X 1500 X (0.05)2 .
7T2 X 9.81 X 20 *
or
D* = 0.01549/ (1)

and from Equation (2.42),

_ 40 1 ' _ 4 x 0.05 \_
e
~ i r i > D ~ - i r X \ x \0'b D

or

Assume / = 0.02 which gives D = 0.1987 m from Equation (1) and

Re = 320,380 from Equation (2).
Using the Swamee-Jain explicit relationship,

f
= ' I 0.00026025 5.74 V? = ° ° 2 1 8 9
8 ++
° \ 33.7
7 xx 0.1987
01987 320,380O9/J
320,
 Head Loss Due to Uniformly Decreasing Discharge 51

For / = 0.02189, D = 0.2023 m and Re = 314650, which gives / =

0.02182. Again f o r / = 0.02182, D = 0.2022 m and Re = 314850, which
g i v e s / = 0.02182. Therefore, D = 0.2022 m.

2.11 HEAD LOSS DUE TO UNIFORMLY DECREASING DISCHARGE

When water is withdrawn through different points along the length of a

pipe as in a municipal pipeline having service connections [Figure 2.3(a)],
the discharge in the pipe decreases from one end to the other. When the
position of the service connections and the rate of withdrawal are known.
the pipe can be considered to be comprised of several pipes connected in
series, each pipe having the same diameter but different flow rates. How-
ever, if it is assumed that the spacing of the service connections and the
withdrawal rate are uniform, the head loss in the pipe can be determined
by considering uniformly decreasing discharge.
Consider a pipe of length L [Figure 2.3(b)] in which the discharge Q, at
section 1-1 decreases uniformly along its length and becomes Q2 at section
2-2. Therefore, discharge in the pipe at a section distance x away from sec-
tion 1-1 is given by Qx = £ , - (£>, - Q2)(x/L). Using Equation (2.5),
the head loss in a pipe portion of length Ax is given by

l6/ r
e,-<e,-a>f
Integrating and simplifying, the head loss in the entire pipe of length L
is given by

L 16/L

If the rate of withdrawal is small in comparison with Qu for pipes of

small length Q2 = Qu and Equation (2.43) reduces to Equation (2.5). If
the discharge becomes zero at the end (<22 = 0),

which is one-third the head loss for uniform discharge <2i.

If it is assumed that the withdrawals are concentrated at the ends, i.e.
one-half of (Gi — Q2) at each end, the discharge in the pipe becomes

\ - Qi\ G. + Qi
52 FRICTIONAL HEAD LOSS IN PIPES

Laterals.

(a)
1,
Qi-K -rQ

f
(b)
Figure 2.3 Head loss in pipe with uniformly decreasing discharge: (a) pipe with service connec-
tions; (b) pipe with uniformly decreasing discharge.

the average discharge. The head loss for the pipe in this case would be

which is not much different from Equation (2.43). Therefore, it is a com-

mon practice in pipe network analysis to consider half the withdrawal at
each end and thus take the average flow rate in the pipe to determine the
head loss in it. However, where greater accuracy is required, the actual
withdrawal pattern can be considered, and the equivalent uniform dis-
charge in the pipe can be obtained as was done by Hamberg and Shamir
[52] for several demand patterns, using Manning's head loss equation.

Example 2.5

A 200-mm diameter, I000-m1ong cast iron pipe supplies water to house

connections at a rate of 20,000 L/d per 100 m length of the pipe. If the in-
flow at the entrance of the pipe is 2 ML/d, find the head loss due to fric-
tion. Find also the head loss by assuming uniform flow rate in the pipe.

Solution

<3, = 2 ML/d. Thus, Q, = (2 x 106)/(1000 X 3600 X 24) =

0.02315 m 3 /s, and Q2 = 2 - 0 . 2 = 1.8 ML/d = 0.02083 m 3 /s. The
average discharge Q = (0.02315 + 0.02083)/2 = 0.02199 m7s.
For new cast iron pipe, e = 0.26 mm. Therefore (e/D) = (0.26/200) =
0.0013, and Re = (AQ/TTDV) = (4 X 0.02199)/(x X 0.2 X 1 X 10"6) =
139,993.
 Reduction of Carrying Capacity with Age 53

Using the Swamee-Jain explicit relationship,

/
= r / o.ooi3025 5.74^p = 002277

[ °8\ 3.7 +
139,993°9 j j
Using Equation (2.43),
_ 16 X 0.02277 X 1000
' ~ 7T2 X 19.62 X (0.2) 5

[ (0.02315)2 + (0.02315)(0.02083) + (0.02083)2!

3

= 2.846 m

Taking the average flow rate in the pipe,

Herein the difference between the two head loss values is insignificant.

2.12 REDUCTION OF CARRYING CAPACITY WITH AGE

The water carrying capacity of pipes decreases with age. Cast iron (CI)
pipes are subject to corrosion and even smooth, cement-lined and bitumen-
lined pipes are affected by the formation of slime. In general, the loss in
carrying capacity is caused by (1) a decrease in the area of the cross section
due to the accumulation of deposits on the interior of the pipes, (2) an in-
crease in the roughness, or (3) by both. Biological growth, silting, in-
crustation, and tuberculation are the most common forms of such deposits,
which may vary in thickness from 1 to more than 10 mm.
For the annual growth rate of roughness in old CI pipes, Colebrook and
White [53] suggested a formula in terms of the pH of water. Their formula
in SI units is

2 log a = 6.61 - pH (2.46)

in which a = annual growth rate of roughness, in millimeters. Lamont

[30] studied the records of 60 old CI pipes and suggested a relationship be-
54 FRICTIONAL HEAD LOSS IN PIPES

tween the annual growth rate and the Langlier saturation index, which in
turn depends upon the pH, alkalinity, and the calcium content of water.
The growth rate of roughness as observed by Lamont [15,30] in coated cast
iron pipes is given in Table 2.8.
It must, however, be remembered that it is not possible to precisely pre-
dict the annual growth rate. Figure 2.4 shows the effect of age on the HW
coefficients. The solid curves show the anticipated HW coefficients as
given by the HW tables of Williams and Hazen [45]. Data from the ob-
servations of the Committee, New England Water Works Association
[54], the values on mains in the Los Angeles water system [55], and the
average curve [56] are shown in the figure. Some observations taken on
water mains in India as reported by Ghatpande and Dhekane [57], Raman
et al. [58], and Dhabadgaonkar and Thakre [59] are also shown in the
figure. The HW coefficients as calculated by Lamont [15,30] are given in
Table 2.9.
Hudson [60] studied several pipes in seven cities of U.S.A. and con-
cluded . . .
(4) Softening of water results in deposits and loss in carrying capacity.
(5) Alignment of a pipe is a major factor determining carrying capacity.

140

100mm

Committee, New England WWA

20
California Section Committee,AWWA
Average Curve A,-

10 20 - 30 40 50 60 70
Pipe Age in Years

Figure 2.4 Effect of age on Hazen-Williams coefficient ([45,54-56]: A,—150 mm CI at Alibag;

B—250 mm CI at Barsi; D—200 mm CI at Dhule; M,—200 mm CI at Mahad; R— KX) mm CI at
Rohe; and V—150 mm CI at Vadgaon [57]; A2—675 mm CI at Ahmedabad; H—550 mm CI at
Hyderabad; M2—900 mm at Madras; S—750 mm Hume Steel at Surat [58]; N—750 mm CI at
Nagpur [59]).
 TABLE 2.8. Growth Rate of Roughness in Coated Cast Iron Pipes.

Equivalent Surface Roughness, mm

Trend Degree of Corrosive Average Growth
Number Attack of Water Rate, mm/y After 30 Years After 60 Years After 100
Years

1 slight 0.025 0.89 1.65 2.67

2 moderate 0.076 2.41 4.70 7.75
3 appreciable 0.250 7.75 15.37 25.53
4 severe 0.760 22.99 45.85 76.33
56 FRICTIONAL HEAD LOSS IN PIPES

TABLE 2.9. Effect of Age on Hazen-Williams Coefficients for Coated

Cast-Iron Pipes.
HW Coefficient for Pipe Diameter in Millimeters
Trend Degree of Corrosive
Number Attack of Water 75 150 300 600 1200

30 years old
1 slight 100 106 112 117 120
2 moderate 83 90 97 102 107
3 appreciable 59 70 78 83 89
4 severe 41 50 58 66 73

60 years old
1 slight 90 97 102 107 112
2 moderate 69 79 85 92 96
3 appreciable 49 58 66 72 78
4 severe 30 39 48 56 62

100 years;old
1 slight 81 89 95 100 104
2 moderate 61 70 78 83 89
3 appreciable 40 49 57 64 71
4 severe 21 30 39 48 54

(6) The CHW value depends upon so many unknown factors that it is im-
possible to estimate it accurately.
(7) A head-loss test must be conducted on each section of main to deter-
mine the working coefficient of carrying capacity in order to have an
accurate design coefficient to use in solving hydraulic problems.

It is thus evident that the correct estimation of the CHW value (and natu-
rally the/and N values) for old pipes is extremely difficult and would re-
quire a large number of field observations to arrive at good results.
Example 2.6

A 200-mm diameter, 1-km long coated cast-iron pipe connects two res-
ervoirs that have a 20 m difference in their water levels. Estimate the dis-
charge through the pipe when the pipe is new, 30 years old, 60 years old,
and 100 years old. Use DW and HW formulas. Assume v for
water = 1 x 10"6 m2/s and the degree of corrosive attack of water as
moderate.
Solution: Darcy-Weisbach Formula

These results were obtained following the same procedure as in Exam-

ple 2.3.
 Problems 57

• When the pipe is new, from Table 2.1, e = 0.13 mm and therefore
Q = 0.06436 m3/s.
• When the pipe is 30 years old, from Table 2.8, e = 2.41 mm,
and therefore Q = 0.04368 nrVs
• When the pipe is 60 years old, from Table 2.8, e = 4.70 mm, and
therefore Q = 0.03862 m3/s
• When the pipe is 100 years old, from Table 2.8, e = 7.75 mm,
and therefore Q = 0.03482 m3/s
Solution: Hazen-Williams Formula

• When the pipe is new, from Table 2.4, by interpolation,

CHW = 135, and therefore Q = 0.06600 m3/s
• When the pipe is 30 years old, from Table 2.9, by interpolation,
CHW = 92.5, and therefore Q = 0.04523 m3/s
• When the pipe is 60 years old, from Table 2.9, by interpolation,
CHW = 81, and therefore Q = 0.03960 m3/s
• When the pipe is 100 years old, from Table 2.9, by interpolation,
CHW = 73 and therefore Q = 0.03569 m3/s

The values of Q given by the DW and HW formulas are nearly the same.
In the DW formula, the relationship between Q and e is not simple, but as
e increases, / increases, and therefore Q decreases. In the HW formula,
however, Q oc CHW as seen from Equation 2.28.

2.13 PROBLEMS

Problem 2.1: A liquid (v = 9.3 mm2/s) flows through a 200-mm

diameter, 4-km long pipe. If the Reynolds number is 1,600, find the dis-
charge through the pipe, in liters per hour, and the head loss due to fric-
tion.
What would be the head loss if the discharge becomes 10 times greater?
Problem 2.2: A 200-mm diameter, 2,000-m long new CI pipe conveys
water at a rate of 0.1 m3/s. Determine the head loss in the pipe using the
Darcy-Weisbach, Manning, and Hazen-Williams head loss formulas.
Problem 2.3: A 250-mm diameter, 500-m long pipe (e = 0.13 mm)
connects two reservoirs having a water level difference of 15 m. Find the
discharge through the pipe.
What would be the percentage change in discharge if e doubles?
Problem 2 4 : A plan to convey water through a new CI pipe at the rate
of 0.1 m3/s between two points, A and B, 800 m apart is proposed. If the
permissible head loss between A and B is 10 m, find the diameter of the
pipe using the DW and HW formulas.
58 FRICTIONAL HEAD LOSS IN PIPES

Problem 2.5: A 150-mm diameter, 500-m long new CI pipe supplies

water to house connections at a uniform rate such that the inflow of 1500
L/min at the entrance reduces to 1000 L/min at the exit. Find the head loss
due to friction. Also find the head loss by assuming a uniform flow rate in
the pipe. What is the percent difference between the two values?
What would be the answers if the flow rate decreases uniformly from (a)
2500 L/min to 2000 L/min and (b) 500 L/min to 0 L/min? What do you
conclude from these answers?
Problem 2.6: Using the DW and HW head loss formulas, estimate the
head loss in the pipe of Problem 2.2 after 15, 30, 45 and 60 years. Assume
that the degree of corrosive attack of water is appreciable.
 Minor Head Loss in Pipes

3.1 INTRODUCTION

s stated in Chapter 2, head loss occurs when the pattern of flow in a

A pipe changes due to a sudden or gradual change in the boundaries or
due to a local obstruction. Even though the source of such a loss is con-
fined to a short length of the pipe, its effect persists for a considerable
length downstream. For example, the normal velocity distribution after a
bend may not be restored for a distance as great as 100 diameters down-
stream from it. The flow in such a length is complicated, and the head loss
is slightly different from that given by the pipe friction approach of Chap-
ter 2. In practice, however, it is assumed that the effects of the usual pipe
friction and the additional turbulence can be separated, and also that the
additional loss is confined to the device causing it and is therefore local.
The total head loss in a pipe is taken as the sum of the normal friction loss
and the additional loss. As the additional head loss is small in comparison
with the frictional head loss in a long pipe, it is usually termed as a minor
head loss.
The minor head loss hm is expressed in terms of the velocity head or the
discharge in the pipe, i.e.,

in which K = minor head loss coefficient that remains practically con-

stant for high Reynolds number, and V = average velocity in the pipe
before the minor head loss element.
Several minor loss elements are considered in this chapter. Since they
are covered in textbooks of fluid mechanics and hydraulics, their theoreti-
cal derivations, if any, are omitted.
59
60 MINOR HEAD LOSS IN PIPES

3.2 SUDDEN ENLARGEMENT

By making several assumptions Borda observed that the minor head loss
due to sudden enlargement (Figure 3.1), i.e., hse, may be theoretically ex-
pressed as

(K, - v2y
K = z (3.2)
2g
in which K,, V2 = velocity in smaller and larger pipes, respectively [4].
Several investigators including Brightmore as stated by King and Brater
[13] and Archer [61] investigated this head loss experimentally. They
observed that Borda's formula gives lower head loss values than those de-
termined experimentally for lower velocities and smaller differences in di-
ameters and higher values for the opposite conditions. Therefore, Archer
[61] suggested another formula, from which King and Brater [13] estimated
values of the coefficient for sudden enlargement Kse for different pipe di-
ameter ratios D2IDX for use in the formula

K = Kse p - (3.3)

Some typical values of Kse are given in Table 3.1.

3.3 GRADUAL ENLARGEMENT

Minor head loss due to gradual enlargement in increasers (Figure 3.2),

hge, can be expressed as [13]

Vl2 — V22
K =^ 2g (3.4)

in which Kge = head loss coefficient for gradual enlargement.

—V, ~~ V2

Diameter, D,

Figure 3.1 Sudden enlargement.

 TABLE 3.1. Minor Head Loss Coefficients.

Appurtenance Head Loss Coefficient8

1. Sudden Enlargement = K —

D,:D2 Range of V,: 0.6-13 m/s

1:1.2 0.11-0.08
1:1.4 0.26-0.20
1:1.6 0.40-0.32
1:1.8 0.51-0.40
1:2 0.60-0.47
1:2.5 0.74-0.58
1:3 0.83-0.65
1:4 0.92-0.72
1:5 0.96-0.75

/ \l\ - V\
Gradual Enlargement hqe = Kg
\ " 2g )

G 2° 4° 6° 8° 10° 12° 15°

KqC! 0.033 0.039 0.046 0.055 0.078 0.10 0.16
O 20° 30° 40° 50° 60° 75° 90°
Kge 0.31 0.49 0.60 0.67 0.72 0.72 0.67

Exit
\ ••" ~ " 2g

3ell-mouthed 0.1
Sharp-cornered 1
Pipe into still water or air 1
(free discharge) (V2 - 0)

4. Sudden Contraction /),.„ -- K c _ _

D 1: D2 Range of V?: 0.6-13 m/s

1.2:1 0.07-0.11
1.4:1 0.17-0.20
1.6:1 0.26-0.24
1.8:1 0.34-0.27
2:1 0.38-0.29
2.5:1 0.42-0.31
3:1 0.44-0.33
4:1 0.47-0.34
5:1 0.48-0.35

(continued)

61
 TABLE 3.1. (continued).

Appurtenance Head Loss Coefficient8

5. Gradual Contraction = K :
2g

Ordinary 0.25
Bell-mouthed 0.10
Streamlined 0.04

6. Entrance hsn = Ken —

Bell-mouthed 0.04
Slightly rounded 0.23
Sharp edged 0.5
Re-entrant (pipe projecting into tank) 1.0
Strainer and foot valve 2.5

7. Bends and Elbows hh = Kh —

a. Elbows—45°
Flanged—regular 0.20-0.30
Flanged—long radius 0.18-0.20
Screwed—regular 0.30-0.42
b. Elbows-90°
Flanged—regular 0.21-0.30
Flanged—long radius 0.18-0.20
Intersection of two cylinders
(welded pipe- not rounded) 1.25-1.8
Screwed—short radius 0.9
Screwed—medium radius 0.75
Screwed—long radius • 0.6
c. Miter Bends
Angle of bend, 9
5° 0.016- 0.024
10° 0.034- 0.044
15° 0.042- 0.062
22.5° 0.066- •0.154
• 30° 0.130- •0.165
45° 0.236- 0.320
60° 0.471- 0.684
90° 1.129- 1.265
d. Return Bend (2 Nos. 90°)
Flanged—regular 0.38
Flanged—long radius 0.25
Screwed 2.2

62
 TABLE 3.1. (continued).

Appurtenance Head Loss Coefficient 8

8. Tees

Standard—bif u rcating 1.5-1.8

Standard—90° turn 1.8
Standard—run of tee 0.6
Reducing—run of tee (in terms of
velocity at smaller end)
2:1 0.9
4:1 0.75

/ V2 \
9. Obstructions h0 - Ko —
I 2gj
Pipe-area to flow-area-at-
obstruction ratio
1.1 0.21
1.4 1.15
1.6 2.40
2.0 5.55
3.0 15.0
4.0 27.3
5.0 42.0
6.0 57.0
7.0 72.5
10.0 121.0

0. Flow Meters

a. Venturimeters

Throat to Inlet Diameter Ratio Long Tube Type Short Tube Type

1:3 1.0-1.2 2.43

1:2 0.44-0.52 0.72
2:3 0.25-0.30 0.32
3:4 0.20-0.23 0.24
b. Orifice Meters
Orifice to pipe diameter ratio
1:4 4.8
1:3 2.5
1:2 1.0
2:3 0.4
3:4 0.24
(continued)

63
 TABLE 3.1. (continued).

Appurtenance Head Loss Coefficient*

1 V 2 \
11. Valves KV
T)
a. Gate Valves
fully open 0.19
three-fourths open 1.15
one-half open 5.6
one-quarter open 24.0
b. Butterfly Valves
Closure angle, 9
0° (fully open) 0.3
10° 0.46
20° 1.38
30° 3.6
40° 10
50° 31
60° 94
c. Diaphragm Valves
fully open 2.3
three-fourths open 2.6
one-half open 4.3
one-quarter open 21.0
d. Plug Globe or Stop Valve
fully open 4.0
three-fourths open 4.6
one-half open 6.4
one-quarter open 780.0
e. Check (Reflux) Valves
Swing check (fully open) 2.5
Ball type (fully open) 2.5-3.5
Horizontal lift type 8-12
f. Foot Valve with Strainer 2.5
g: Pressure Reducing Valve 10.0

Subscripts 1 and 2 represent earlie> and later ends, respectively.

64
 Sudden Contraction 65

Figure 3.2 Gradual enlargement.

Based on the experimental study of Gibson [62] and Parker [63], King
and Brater [13] estimated the values of Kge for different values of the angle
of enlargement 0 . Some typical values are given in Table 3.1.

3.4 EXIT

The head loss at an outlet or exit, hex, is expressed as

V1 — V2
hex = Kex^~ (3.5)

in which K,x = head loss coefficient for exit. The value of Kex depends on
the type of exit, as shown in Table 3.1. When a pipe discharges into still
water or air, the kinetic energy of the outgoing water is lost, and therefore,
K.x = 1 and V2 = 0 in Equation (3.5), giving

(3.6)

3.5 SUDDEN CONTRACTION

Even though a sudden contraction is geometrically the reverse of a sud-

den enlargement, the flow pattern is not just the reverse. A vena contracta
(section of greatest contraction of the jet) is formed after the sudden con-
traction (Figure 3.3), and eddies are formed between the vena contracta
and the pipe wall. These eddies are the main cause of the head loss.
Head loss due to sudden contraction, hsc, is given by

in which Cc — coefficient of contraction (=AJA2), and Ksc = head loss

66 MINOR HEAD LOSS IN PIPES

—
' --.
• ^
rVena contracta,

—v, —V2

A2
m—"
Area, A,
Figure 3.3 Sudden contraction.

coefficient for sudden contraction. The value of C and therefore of Kst de-
pends mostly on the diameters of the small and large pipes.
Using the experimental results of Brightmore. and suggestion of Parker
[63], King and Brater [13] estimated the values of K,c for different values
of diameter ratios. Some typical values are given in Table 3.1.

3.6 GRADUAL CONTRACTION

Minor head loss due to gradual contraction in reducers. hKC. is given by

V\
K = *„, ~ (3-8)
2H
in which Kgc = head loss coefficient for gradual contraction, and V2 =
velocity at the small end. The values of KKU for different types of reducers
are given in Table 3.1.

3.7 ENTRANCE

Minor head loss at the pipe entrance. /?,.„. is given by

V2
/»,„ = K,,,, . (3.9)
2
#

in which Ken = head loss coefficient for the pipe entrance, and V = aver-
age velocity in the pipe. The head loss at the entrance is due to separation
and therefore depends on the shape of the entrance. Head loss coefficients
for different entrance shapes are given in Table 3.1.
 Tees 67

3.8 BENDS AND ELBOWS

Even though the magnitude of the average velocity remains the same in
bends and elbows, the direction of velocity changes. This change requires
a radial inward force to provide radial inward acceleration. This causes
distortion in the velocity distribution which in turn produces additional
turbulence as well as double spiral or secondary motion that may persist
for a downstream distance up to 100 pipe diameters until dissipation due
to viscous forces takes place.
Minor head loss in bends and elbows. hh. is given by

hb = K,X (3.10)

in which Kb = head loss coefficient for bends and elbows, and V = aver-
age velocity in the pipe. The value of K,, depends upon the angle of the
bend 9 , the ratio of the radius of the bend R to the diameter of the pipe
D (i.e., on relative radius RID), and the relative roughness of the pipe elD.
The value of Kb increases with the angle of the bend and the relative rough-
ness, but decreases with the RID ratio. The Reynolds number has prac-
tically no influence on the coefficient.
Several investigators such as Fuller. Beij. Anderson, and Straub as
stated by King and Brater [13] have carried out experiments and provided
the values of Kb. Ito [64,65] provided information for flow and head loss
in bends. Jeppson [14] provided curves showing the effect of the relative
radius RID and the angle of bend 9 on the value of K,,. Values of Kh for
some typical bends and elbows as compiled by Amirtharajah [66] are
given in Table 3.1.

3.9 TEES

The head loss in tees depends on the division of flow and the extent of
the change in direction of flow. The head loss in tees, /?,, can be expressed
as
V1
h, = K,— (3.11)
2#

in which K, = head loss coefficient for lees. Values of A", for some typical
tees are given in Table 3.1 [66].
68 MINOR HEAD LOSS IN PIPES

3.10 OBSTRUCTIONS

The head loss due to obstructions, ho, can be expressed as

V2
K K (3.12)

in which Ko = head loss coefficient for obstructions and V = average

velocity in the pipe before the obstruction. The value of Ko would depend
on the shape and size of the obstruction and also on its position relative to
the centre line of the pipe. Some typical values of Ko are given in Table 3.1
[66].

3.11 FLOW METERS

Venturimeters and orifice meters are usually installed in pipelines to

measure the discharge through them. Even though they are designed to
minimize the head loss through them, flow meters cause head loss depend-
ing upon their constriction and their angle of divergence. Head loss
through venturimeters is less than that through corresponding orifice
meters and therefore venturimeters are preferred to orifice meters where
head loss is to be minimized. Head loss through flow meters, hfm, can be
expressed as

h/m = K/m^ (3.13)

in which Kfm = head loss coefficient for the flow meters. Values of Kfm for
some typical venturimeters and orifice meters are given in Table 3.1 [66].

3.12 VALVES

Valves are provided in pipelines to control the rate of flow, the direction
of flow, or the pressure. The head loss through valves, hv, can be expressed
as
V1
hv = K.— (3.14)
2
S
in which Kv = head loss coefficient for valves. Values of Kv for gate
valves, butterfly valves, diaphragm valves, plug globe or stop valves,
check valves, foot valves, and pressure reducing valves are given in Table
3.1 [66-68].
 Equivalent Pipes

4.1 INTRODUCTION

ATER distribution networks often contain pipes connected in series,

W parallel, or series-parallel combinations. Further, they may also
contain minor loss elements such as bends, valves, etc. In network analy-
sis, it is preferable to reduce such a combination to a simple pipe by the
equivalent pipe concept. An equivalent pipe is a pipe of uniform diameter
such that the head loss and discharge in it are the same as the head loss and
discharge in the multiple pipe combination. The equivalent pipe may be
of some specified length, specified diameter, or specified roughness coef-
ficient.

4.2 PIPES IN SERIES

When two or more pipes of different diameters or different friction

coefficients or both are joined end to end, a series pipe system (Figure
4.1) is obtained. Using x, x = 1, 2, . . . for the individual pipes, and e
for the equivalent pipe, for a series pipe system:
(1) The discharge through all the pipes is the same. Therefore,

Q. = a , x = 1, 2, . . . (4.1)

(2) The head loss in the equivalent pipe is the sum of the head losses in
the individual pipes. Therefore, neglecting minor head losses,

K = K + K + ••• = £ K (4-2)
X

69
70 EQUIVALENT PIPES

Li L2 U
U U U
ht, hf, K
Oi = O2 = O3

Figure 4.1 Pipes in series.

Using the Darcy-Weisbach formula, Equation (2.5), and simplifying

feLe y fxLx
-QS - L -fir (4-3)

When the Hazen-Williams formula, Equation (2.28) is used

i-'X /A A-.

For Manning's formula, Equation (2.33),

' Wk _ y mk

For the general head loss formula, Equation (2.37),

Two types of problems arise in series pipe systems: (1) determining the
total head loss and the equivalent pipe when the discharge is known and
(2) determining the discharge through the system and the equivalent pipe
when the total head loss is known.

4.2.1 DISCHARGE KNOWN

Since the discharge is known, the head loss in individual pipes and
flierrfore the total head loss can be easily obtained. Similarly, the length
 Pipes in Series 71

of an equivalent given-diameter pipe can also be easily obtained. However,

the diameter of an equivalent given-length pipe requires a trial and error
procedure if the DW head loss formula is used.

Example 4.1

A 200-m long, 300-mm diameter new cast iron pipe is connected in

series to a 100-m long, 150-mm diameter new cast iron pipe. If the series
combination carries water (v = 1.004 mm2/s) at a rate of 0.05 m3/s, find
(1) the head loss in the system, (2) the length of an equivalent new cast iron
pipe of 200 mm diameter, and (3) the diameter of a 300-m long equivalent
new cast iron pipe. Use the Zigrang-Sylvester explicit relationship for esti-
mating/in the Darcy-Weisbach formula. Also find all the answers using
Hazen-Williams formula. Assume CHW = 130 for all pipes.

Solution: Darcy-Weisbach Head Loss Formula

For new cast iron pipe e = 0.26 mm

* L • i f •0M0867

v = 1.004 mnrVs = 1.004 X 10"6 m2/s

422J20
"SL
= x x 0.15 x iSL x 10- =
Using the Zigrang-Sylvester relationship, Equation (2.23), / 300 =
0.02043, and/,so = 0.02303.
(1) Head loss in the system

h, = hfl + hfl

_ 0.02043 X 200 X 0.052 0.02303 X 100 X 0.052

12.1 X (0.3)5 +
12.1 X (0.15)5

= 0.3474 + 6.2660 = 6.613 m

72 EQUIVALENT PIPES

(2) For the 200-mm diameter equivalent new cast iron pipe,

e 0.26
= 0.0013
D 200

"' - T x0.2 4 x X i° [ |£x,0- = 3I7 -° 40

/ = 0.02172

.". Using Equation (4.3),

0.02172 x U _ 0.02043 x 200 0.02303 x 100

(0.2)5 ~ (0.3)5 +
(0.15)5

Le = 471.6 m

(3) If De a n d / are the diameter and friction coefficient of the equivalent

pipe, respectively, then

/ x 300 _ 0.02043 x 200 0.02303 x 100

D*e ~ (0.3)5 +
(0.15)5

giving

D, = 0.3930/° 2 (1)

4 x 0.05 63408
•KD. X 1.004 X 10"' De

For the first trial, l e t / = 0.02. /. De = 0.1797 m; (e/De) = 0.001447;

Re = 352,855, giving/ = 0.02217. F o r / = 0.02217, De = 0.1835 m;
(e/De) = 0.001417; Re = 345,548; / = 0.02208, giving Z>« = 0.1833
m = 0.1835 m, ah earlier obtained value. .'. De = 0.1833 m, i.e., 183.3
mm.

Solution: Hazen-Williams Head Loss Formula

For new cast iron pipe CHW = 130

10.68 x 0.05 18 » / 200 100

(1) h, - 130i.8S2 \(0.3) 4 (0 15)4 8
m
 Pipes in Series 73

(2) Using Equation (4.4),

/ 200 100 \ _
- n ?487
((0.3) 4 8 7 +
(0.15) 4 8 7 / ~

300 / 200 100

V
' £>487 \(0.3) 4 8 7 (0.15) 487

giving De = 0.1854 m, i.e., 185.4 mm.

4.2.2 TOTAL HEAD LOSS KNOWN

When the total head loss in a series system is known, for the DW head
loss formula, a trial and error procedure is necessary to determine the
head loss in individual pipes, the discharge in the system, and the length
or the diameter of the equivalent pipe.

Example 4.2

In the series system of Example 4.1, the total head loss is 10.00 m. Find
(1) the discharge through the system, (2) the length of an equivalent new
cast iron pipe of 200 mm diameter, and (3) the diameter of a 300-m long
equivalent new cast iron pipe. Use the Zigrang-Sylvester explicit relation-
ship for estimating/in the Darcy-Weisbach head loss formula. Also find all
the answers using theHazen-Williams head loss formula. Assume CHW =
130 for all pipes.

Solution: Darcy-Weisbach Head Loss Formula

From Example 4.1, (e/D)300 = 0.000867, and (e/D)ls0 = 0.001733.

(1) Since the discharge is unknown, the Re and /values for the individ-
ual pipes are also unknown. For the series system,

h, = hh + hfl

• , Q _ /• X 200 x g 2 / 2 x 100 x Q1
12.1 X (0.3) 5 12.1 X (0.15) 5

= (6802/, + 108832 f2)Q2

Taking / , = / 2 = 0.02 for the first trial, Q = 0.06576 m3/s.

(/te)3oo = 277980; (Re)iS0 = 555960; / , = 0.02011; and f2 =
74 EQUIVALENT PIPES

0.02292, giving Q = 0.06165 m'/s. .'. (Re)300 = 260610; (Re)l50 =

521220;/, = 0.02018; and/ 2 = 0.02294, giving Q = 0.06162 nrVs.

(2) The equivalent pipe carries the same discharge as the series combi-
nation.
.-. Qe = 0.06162 nrVs

= 0.0013
D 200

and

/ = 0.02158

But

0.02158 x L , x (0.06162)2
= 10
12.1 x (0.2)5

which gives

• Le = 472.5 m

/ x 300 x (0.06162)2
1 _
' i2.i x m
which gives

De = 0.3933/? 2 (1)

and

4 x 0.06162 78144
_ 6 {
e
~ I x 1.004 x 10" De '

For the first trial, l e t / = 0.02. .". De = 0.1799 m; (e/D,) = 0.001446;

Re = 434,370, g i v i n g / = 0.02205. F o r / = 0.02205, De = 0.1834 m;
{elD.) = 0.001418; /?e = 426,090; / = 0.02197 giving De = 0.1833 m,
i.e., 183.3 mm.
 Pipes in Parallel 75

Solution: Hazen-Williams Head Loss Formula

10 .68 x Q..«2 / 200 100

_
1301 .852 \ ( 0 . 3 ) 4 8 7 """ (0. 15) 487

which gives

Q = 0.06864m3/s

as in Example 4.1

300 / 200 100

487 487
D \(0.3) (0.15) 487

giving D« = 0.1854 m, i.e., 185.4 mm, as in Example 4.1

4.3 PIPES IN PARALLEL

When two or more pipes of uniform diameter, carrying uniform dis-

charge throughout their lengths, start at one point and meet again at some
other point, they are said to be in parallel, giving a parallel-pipe system
(Figure 4.2). Using x = 1, 2, . . . for individual pipes and e for the
equivalent pipe, for a parallel-pipe system:
(1) The head loss in all the pipes is the same. Therefore,

hfm = h,x,x = 1 , 2 , . . . (4.7)

(2) The discharge in the equivalent pipe is the sum of the discharges in
the individual pipes. Thus,

a (4-8)

Using the Darcy-Weisbach formula and simplifying,

76 EQUIVALENT PIPES

Pi. Li, f,, Q,,h f ,

». L 2 , f 2 , Q 2 , h f ,

Da, L3. Qs. h f 3 .

— •
)
h h
f,= V f3
Figure 4.2 Pipes in parallel.

For the Hazen-Williams formula,

C H w,.£>, 21
£0.54 -E L?-
(4.10)

For Manning's formula,

D8X'3
NM n =E
~ u
NxLxn
(4.11)

For the general head los's formula,

-m, (4.12)

Using the Darcy-Weisbach formula, Jeppson [69] provided several for-

mulas for equivalent pipes to replace a set of pipes. He recommended
lumping the effects of the length, diameter and head loss coefficient into
the pipe resistance constant R, i.e., his treatment is based on the general
head loss formula. He has shown that time and effort are considerably re-
duced if equivalent pipes are used to replace all sets of parallel pipes in the
network analysis. Walski [70] recommended the use of the length and di-
ameter of the largest diameter pjpe and the change of its head loss coeffi-
cient to account for the other pipes to obtain an equivalent pipe for a set
of parallel pipes. He also provided formulas and nomograms for this pur-
pose.
In parallel-pipe systems two types of problems arise: (1) determining the
total discharge and the dimensions of the equivalent pipe when the head
 Pipes in Parallel 77

loss is known and (2) determining individual pipe discharges, the head
loss, and dimensions of the equivalent pipe when the total discharge is
known.

4.3.1 HEAD LOSS KNOWN

Using the known head loss and other parameters, the discharge in each
pipe is calculated. A trial and error procedure is necessary if the DW head
loss formula is used. These discharges, when added together, give the total
discharge. From this, the equivalent pipe dimensions can be obtained.

Example 4.3

In a 3-pipe parallel system as in Figure 4.2, L, = 300 m; Di = 200

mm; ex = 0.3 mm; L2 = 250 m; D2 = 300 mm; e2 = 0.2 mm;
L3 = 400 m; D3 = 250 mm; and e3 = 0.4 mm. If the head loss through
the system is 10 m, find: (1) the total discharge through the system and (2)
the length of a 300-mm diameter (e2 = 0.2 mm) equivalent pipe. Take v
for water = 1 X 10~6 m2/s.
Also find the answers using the HW formula, taking the coefficient of
the three pipes as 120, 130, and 110, respectively. Take Cm = 130 for the
equivalent pipe.

Solution: Darcy-Weisbach Head Loss Formula

(1) For the first pipe,

15
% - 1 = °<*>
Taking/, = 0.02 for the first trial,

0.02 X 300 X Q\
12.1 X (0.2) 5

giving Qt = 0.08033 m3/s

.'. Re = 511400, / , = 0.02217, Q, = 0.07630 m3/s

/ . Re = 485740, / , = 0.02219, Q, = 0.07626 m3/s

Similarly, Q2 = 0.2541 nWs; and Q3 = 0.1147 m3/s, giving total

Q = 0.4451 m3/s.
78 EOUIVAl FNT PIPFS

(2) For the equivalent pipe,

i. - i =ommi
Q = 0.4451 nrVs; Re = 1.889.000;/= 0.01806: and/.. = 82.18 m

Solution: Hazen-Williams Formula

(1) For the first pipe,

10.68 x 300 x Q\-™2 _

(120)'-"" x (0.2)'187

giving Gi = 0.07731 m3/s

Similarly,

Q2 = 0.2684 m 3 /s; and Q3 = 0.1091 mVs. giving Q = 0.4548

m3/s

(2) For the equivalent pipe, the diameter = 300 mm, and C,m = 130.
This gives Le = 94.13 m.
4.3.2 TOTAL DISCHARGE KNOWN

When the DW head loss formula is used, the solution involves a

repeated trial and error procedure as neither the discharge nor the head
loss is known for the individual pipes. Briefly the procedure is as follows:
(1) Assume a suitable value of/, say 0.02. for all the pipes
(2) Using these/values, find Q2, Qi. . . . in terms of Qx
(3) Find J J 2* and using the given Q value, determine the values of

(4) Determine Re, elD and then/for each pipe and then go to step 2
(5) Terminate the procedure when/values stabilise
(6) Determine Qu Q2, . . . ; and h,

Example 4.4

In the 3-pipe parallel system of Example 4.3, the total discharge through
the system is 18 ML/d. Find the distribution of flow and the head loss us-
ing the DW head loss formula. Take v for water = 1 mnWs.
 Pipes in Parallel 79

Find the answers using the HW formula with HW coefficients for the
three pipes as given in lixample 4.3. Also find the HW coefficient of a
250-m long equivalent pipe of 300 mm diameter.

Solution

The total discharge.

Solution: Darcy-Weisbach Formula

Assuming / t o be the same for the three pipes.

300 25 40
Q± _ ° Qj - ° @'
(0.2f ~ (0.3r ~ (0.25) s

giving

Qi = 3.019 (?,. and Q, = 1.513 £),

r5.3 = °- 0 3 7 6 5 mJ;s

Q2 = 0.11368 mVs: and Q, = 0.05697 m'/s

Using these discharges and the appropriate e'D values. /, = 0.02264:

f2 =^0.01865: and/, = 0.02282.

0.02264_x 300 ^f _ O.()[865 x_2M)_C>i _ 0.02282 x 400 Q\

~{().2r '" ~ (0.3)- ~ (0.25) s

giving

Q, = 3.326 Qt. and Q, = 1.507 Ql

:. Q, = 0.0357! m''s: Q, = 0.11877 m'/s: and Q, = 0.05382 ni^s

.-./, = 0.02268: ,/2 = 0.01862; and/, = 0.02286

/. C?. = 0.0357 mVs: Q2 = 0.1188 m'/s: and Q, = 0.0538 m'/s

and the head loss = 2.24 m

80 EQUIVALENT PIPES

Solution: Hazen-Williams Formula

300 X QlS52 250 X Q\M1 400 X Q\*

(120)1852 X (0.2) 487 (130)1852 X (0.3) 487 (HO) 18 " X (0.25) 487

from which Q2 = 3.472 Qr, and Q3 = 1.411 Qu giving Q1 = 0.0354

nV/S; Q2 = 0.1229 nWs; and Q3 = 0.0500 m3/s; and h, = 2.35 m.

For the equivalent pipe,

10.68 X 250 X (0.2083)' 852

2 4 87
C'H$? x (0.3) - ~

giving CHW = 220.5

4.4 PIPES IN SERIES-PARALLEL

In pipe networks, pipes connected in series may also be subsequently

connected in parallel to give a series-parallel system. Pipes connected in
series are initially replaced by equivalent pipes to give a system composed
of equivalent pipes connected in parallel. This parallel system is replaced
by an equivalent pipe which is equivalent to the entire series-parallel
system.

Example 4.5

Find the diameter of a 900-m long equivalent pipe (CHW = 100) to re-
place the series-parallel system shown in Figure: Example 4.5. The length,
diameter and CHW coefficient values for the pipes are as follows: Pipe
1—300 m, 250 mm, 120; Pipe 2 ^ 0 0 m, 300 mm, 130; Pipe 3—200 m,
200 mm, 100; Pipe 4—500 m, 400 mm, 130; and Pipe 5—300 m, 250
mm, 80.
 Minor Loss Elements 81

Solution

For the series-pipe system consisting of pipes 1, 2, and 3, the diameter

of 900-m long equivalent pipe with CHW = 100 can be obtained from
Equation (4.4).

900 300
(120) 1M2
,852/n "><;\4.87
(0.25)

400 200
(130)1 8 " (0.3) 487 (100)' 852 (0.2) 4 8
' s * /** y\ v< oci /A ^\ \ A on •

which on simplification gives De = 0.2495 m.

Similarly, for pipes 4 and 5,

9 0 0 5 0 0 300
(100)1-852 Dt*7 ~ (130) 1852 (0.4) 4 87 ++
(80) 1852 (0.25) 4 87

giving De = 0.2839 m

Now two pipes, each 900-m long, having CHW = 100 and of diameters
0.2495 m and 0.2938 m are connected in parallel. Therefore from Equa-
tion (4.10),

100 x Dj63 _ 100 X (0.2495)2 63 100 X (0.2839)2 63

(900) 054 ~ (900) 054 +
(900) 054

or
63
m = (0.2495) 2 " + (0.2839) 2 •"

giving De = 0.3483 m

.'. Equivalent pipe diameter for the entire system = 0.3483 m, or

348.3 mm.

4.5 MINOR LOSS ELEMENTS

When minor loss elements such as valves, meters, and bends are present
in a pipeline, the head loss in them can be considered by the methods
82 EQUIVALENT PIPES

shown in Chapter 3. However, in pipe network analysis it is a common

practice to consider the head loss effect of such elements through equiva-
lent pipe lengths. The equivalent pipe length has the same diameter, dis-
charge, and friction coefficient as the pipe in which the minor element is
present, and the head loss in it is equal to the head loss in the minor ele-
ment. Therefore if AL is the equivalent pipe length for a minor element,
then its value can be obtained using different head loss formulas.

4.5.1 DARCY-WEISBACH FORMULA

Equating the head loss in a minor element [Equation (3.1)] to that in an

equivalent pipe of length AL [Equation (2.4)],

KYL _ f^k YL
2g ~ D " 2g

which gives

AL = f (4.13)

As the friction coefficient/depends upon Re and therefore upon the dis-

charge Q, the equivalent length AL would also depend upon Q. When Q
is known, the actual value of./and therefore that of AL can be calculated.
When Q is unknown, a trial and error procedure would be necessary.

4.5.2 HAZEN-WILLIAMS FORMULA

From Equations (3.1) and (2.28),

SKQ2 (10.68) AL(2 1852

or

AL =

Thus, for the HW formula, AL is also a function of Q.

 Minor Loss Elements 83

4.5.3 MANNING FORMULA

From Equations (3.1) and (2.33)

SKQ2 _ IO.29N2ALQ2
ir2gir ~ D 16/3

giving

KD*/3
A L < 4 I 5 )

It is thus seen, that for Manning's head loss formula, the equivalent pipe
length is independent of the discharge.

4.5.4 EQUIVALENT PIPE LENGTHS

Even though Equations (4.13)-(4.15) give equivalent pipe lengths for

minor loss elements, for simplicity, standard equivalent pipe lengths may
be taken in pipe network analysis. Barnard [71] has given values for
equivalent LID ratios for different minor loss elements. Zimke [72] carried
out experiments on pipe bends and elbows and provided charts giving
equivalent pipe length, pipe diameter ratios versus the bend radius, and
pipe diameter ratios. Typical values of equivalent pipe lengths for some
minor loss elements are given in Table 4.1 [71-74]. For a particular minor
element, an appropriate length-diameter ratio is selected and multiplying
this value by the pipe diameter gives the equivalent pipe length for the
minor element.
It has been observed that

AL = 35KD (4.16)

provides a good approximation for equivalent pipe lengths for minor loss
elements.

Example 4.6

A 300-m long, 200-mm diameter new cast iron pipe contains a globe
valve (K = 10) and a meter (K = 2). If the discharge in the pipe is 3000
L/min, determine the equivalent pipe length using DW (e = 0.26 mm),
 TABLE 4.1. Equivalent Pipe Lengths.

Ratio of Equivalent
Pipe Length
Appurtenance to Diameter

1. Sudden Enlargement
D,ID2 = 3:4 7
1:2 22
1:4 31
2. Sudden Contraction
D,ID2 = 2:1 12
3. Entrance: Sharp-Edged 18
4. Bends and Elbows
a. Elbows—45°
Flanged—regular
Flanged—long radius
Screwed—regular
b. Elbows—90°
Flanged—regular 24
Flanged—long radius 18
Cast—regular 32
Cast—long radius 20
c. Welded-90°
bend radius
= 0.5 38
pipe diameter
1.0 17
1.5 12
2.0 10
3.0 9
d. Return bend—180°
regular 73
large radius 39
Tee
Line flow 60
Branch flow 300
6. Valve
a. Gate valve
Fully open 6.7
Three-fourths open 39
One-half open 190
One-quarter open 800
b. Globe valve 320
c. Swing check 80
 Minor Loss Elements 85

HW (CHW = 130), and Manning's (N = 0.010) formulas. What would be

the answers if the discharge doubles?

Solution

The combined loss coefficients for both the minor elements = 10 +

2 = 12.

3000
Q = 3000 L/min = ^ ^ = 0.05 mVs

For the Darcy-Weisbach equation, taking v = 1 x 10"6 m2/s,

and

Re = 4X0.05 =
7T x 0.2 x 1 x 10-6

.". / = 0.02172, and from Equation (4.13),

/ . The equivalent pipe length = 300 + 110.5 = 410.5 m

For the Hazen-Williams formula, from Equation (4.14),

12 X (130) 1852 X 0.2 0 8 7 X (0.05) 0148 ,_o

AL = ^ = 120.8 m

so that equivalent length = 420.8 m. For Manning's formula, from Equa-

tion (4.15),

A , 12 X 0.2 4 ' 3
AL = =u2Jm
124.5 x (omoy
so that the equivalent length = 412.7 m. When the discharge doubles, the
corresponding answers are 412.4 m, 433.8 m, and 412.7 m, respectively.
Using Equation (4.16), L = 35 X 12 X 0.2 = 84 m, so that the
86 EQUIVALENT PIPES

equivalent pipe length = 384 m. Using Table 4.1, and assuming that the
equivalent pipe lengths are proportional to the K values. /. = (320 +
64) X 0.2 = 76.8 m, so that the equivalent pipe length = 376.8 ni.

4.6 PROBLEMS

Problem 4.1: Three new CI pipes with the following lengths and di-
ameters: 200 m, 200 mm; 300 m. 150 mm; and 400 m, 250 mm, are con-
nected in series. If this pipeline carries water at a rate of 4,000 L/min.
find: (1) head loss in the system. (2) the length of an equivalent new CI
pipe 300 mm in diameter, and (3) the diameter of a 900-m long equivalent
new CI pipe. Solve the example using the DW, HW, and Manning formu-
las.
Problem 4.2: In the series system of Problem 4.1, the total head loss
is 20 m. Find: (1) the discharge through the system, (2) the length of
an equivalent new CI pipe 300 mm in diameter, and (3) the diameter of a
900-m long equivalent new CI pipe. Solve the example using the DW, HW,
and Manning formulas.
Problem 4.3: The three pipes of Problem 4.1 are connected in parallel.
If the head loss through the system is 15 m, find: (1) the total discharge
through the system, (2) the length of an equivalent new CI pipe 300 mm
in diameter, and (3) the diameter of a 400-m long equivalent new CI pipe.
Solve the example using the DW and HW formulas.
Problem 4.4: In the 3-pipe parallel system of Problem 4.3, the total dis-
charge is 0.18 m 3 /s. Determine the distribution of flow and the head loss
using the DW and HW formulas.
Problem4.5: The series-parallel system shown by Figure: Problem 4.5
consists of 6 pipes having the following length, diameter and HW coeffi-
cient: Pipe 7—300 m, 250 mm, 120: Pipe 2—400 m, 200 mm, 110; Pipe
3—600 m, 200 mm, 100: Pipe 4—200 m. 200 mm, 90; Pipe 5—300 m,
250 mm, 110; and Pipe 6—250 m. 200 mm, 120. Find: (1) the diameter of
an equivalent 600-m long pipe with an HW coefficient of 130 and (2) the
HW coefficient of a 600-m long, 300-mm diameter equivalent pipe.

Problem 4.6: Solve Problem 4.5 if the six pipes are connected as
shown in Figure: Problem 4.6.
 Problems 87

Problem 4.7: In the series-pipe system of Problem 4.1, the first pipe
contains a globe valve (K = 10), the second pipe contains a meter (K =
2). and third pipe contains some minor loss elements (total A' = 25). If
the pipe line carries water at a rate of 4000 I./min. rind: (I) the head loss
in the system and (2) the diameter of a 900-m long equivalent new CI pipe.
Solve the example using the DW and HW formulas.
 Reservoirs, Pumps, and Special Valves

5.1 RESERVOIRS

5.1.1 IMPOUNDING RESERVOIRS

ATER supply projects that withdraw water directly from streams

W may be unable to satisfy the demands of the consumers during low
or no flow stream conditions. However, such streams may carry large
quantities of water during floods. Impounding reservoirs, constructed
directly on such streams store water during periods of high runoff and
release it for use during periods of low runoff. These reservoirs take care
of the seasonal variations in supply and demand of water. Water from the
impounding reservoirs is transported to the consumption centre by grav-
ity, if topology permits, or is pumped through pumping mains.

5.1.2 SERVICE AND BALANCING RESERVOIRS

Reservoirs are provided in water distribution systems to ensure suffi-

cient quantity of water at required pressures. Service reservoirs, storage
reservoirs, or distribution reservoirs are provided as close to the con-
sumption centres as possible to take care of the daily fluctuations in de-
mand. The reservoir fills when the rate of filtration or pumping exceeds
the rate of demand and empties under the reverse conditions. This allows
the pumps and the treatment plants to operate at a constant rate thereby re-
ducing their capacity. These reservoirs also provide fire storage for im-
mediate use in large quantities for fire-fighting purposes and also provide
emergency storage during failure of intake, supply conduit, or power.
Equalizing reservoirs or balancing reservoirs are provided to equalize
pressures in the distribution system and thereby to reduce fluctuations in
pressure caused by fluctuating demand. They also raise pressures at
remote points far from the storage reservoirs and pumping stations and
89
90 RESERVOIRS, PUMPS, AND SPECIAL VALVES

thus provide improved service during periods of peak demand. When

located near pumping stations, they equalize the head on pumps. This
results in better pump selection and allows them to operate at higher effi-
ciencies. Incidentally, service reservoirs also serve as balancing reser-
voirs and balancing reservoirs as service reservoirs.
Service and balancing reservoirs may be surface reservoirs or elevated
reservoirs. When the height of a surface reservoir exceeds its diameter, it
is known as a standpipe.
A reservoir may be connected to a distribution network by separate in-
let and outlet pipes or by a single pipe such that water enters and leaves
through the same pipe. In the latter case the reservoir is said to be floating
on the system. When the supply rate exceeds the demand rate, water en-
ters into the reservoir, and when the demand rate exceeds the supply rate
water flows out of the reservoir.
When a distribution network receives water from the reservoirs through
gravity only, the network is a gravity network. When pumps are provided,
the network is a pumped network.

Example 5.1

Pipes 1 and 2 connect reservoirs 1 [water level (W.L.) = 120.00 m],

and 2 (W.L. = 100.00 m) respectively, to a junction point J as shown in
Figure: Example 5.1. A third pipe, pipe 3, starts from the junction pointy

JW.L. 120.00 m
_ W . L 100.00 m

and discharges into the atmosphere at point 3. The length, diameter and
CHW coefficient values for the pipes are as follows: Pipe 1—300 m, 300
mm, 100; Pipe 2—150 m, 200 mm, 130; and Pipe 3—200 m, 300 mm,
100. Determine the discharges in pipes 1 and 2 and the hydraulic gradient
level (HGL) values at points J and 3, when the outflow at point 3 is: (1) 0.2
m3/s and (2) 0.4 nrVs.
 Reservoirs 91

Solution

For the pipe diameter in millimeters and discharge in cubic meters per
second, using Table 2.3, the resistance constant for pipe 1,

_ 4.351 X 1015 X L _ 4.351 x 1015 X 300

1
~ C}jvv52 X D 4 8 7 ~ (100)' 852 x (300) 487

Similarly, R2 = 493.90; and R3 = 148.61.

When the HGL at 7 is 100.00 m, reservoir 2 neither supplies nor
receives water. Therefore, Q2 = 0; and (?i = Q3.

•«*•-*

Thus, when the demand at point 3, i.e., q3 is 0.2720 m3/s, reservoir 2

neither supplies nor receives water. When q3 > 0.2720 m3/s, reservoir 2
supplies water and the flow in pipe 2 is from 2 to 7. Similarly, when
q3 < 0.2720 m3/s, reservoir 2 receives water and the flow in pipe 2 is from
7 to 2. Here, reservoir 2 receives or supplies water through the same pipe
2 and thus reservoir 2 is floating on the system.

Case 1

Outflow at 3, q3 = 0.2 mVs

As q3 = 0.2 m3/s < 0.2720 m J /s, reservoir 2 receives water. Let the
discharge in pipe 1 be Q. From the flow continuity at 7, the discharge in
pipe 2, from 7 to 2, i.e., Q2 = Q - 0.2. Let the HGL at 7 be //,.
For pipe 1, 222.910 1852 = 120 - //,; and for pipe 2, 493.90 X
(Q - 0.2) 1852 = Hj - 100. Adding these two equations, 222.910 1852 +
493.90(0 - 0.2) 1852 = 20. (This equation can also be directly obtained
by considering path 1-7-2, so that the algebraic sum of the head losses
in pipes 17 and 72 is equal to the difference in water levels in reservoirs
1 and 2.)
Solving by trial and error, Q = 0.2548 m3/s.

.-. g , = 0.2548 m 3 /s; and Q2 = 0.2548 - 0.2 = 0.0548 nWs

Hj = 120 - 222.91 X (0.2548) 1852 = 102.28 m; and
H3 - 102.28 - 148.61 X (0.2) 1 - 8 " = 94.74 m
92 RESERVOIRS, PUMPS, AND SPECIAL VALVES

Case 2

Outflow at 3, q3 = 0.4 nrVs

As q3 = 0.4 m3/s > 0.2720 mVs, reservoir 2 supplies water. Let the
discharge in pipe 1 be Q so that Q2 (from 2 to 7) = 0.4 — Q. Con-
sidering path 1-7-2, 120 - 222.91 G 1852 + 493.90(0.4 - Q) 1 8 5 2 = 100.
Solving by trial and error, Q — 0.3109 m3/s.

.-. <2, = 0.3109 nrVs; Q2 = 0.4 - 0.3109 = 0.0891 m3/s;

Hj = 120 - 222.91 X 0.3109' 852 = 94.39 m; and
H3 = 94.39 - 148.61 X 0.4' 852 = 67.16 m

5.1.3 THREE-RESERVOIR SYSTEM

Consider the three-reservoir system shown in Figure 5.1. This system is

similar to the one considered in Example 5.1, except that the end 3 of pipe
73 is now connected to reservoir 3. In Example 5.1 the flow in pipe 73 was
known while here the flows in all three pipes are unknown.
Let Hi, H2, H3 (//, > H2 > H3) and //, be the HGL values at reser-
voirs 1, 2 and 3, and the junction 7, respectively. As Hi is maximum, the
flow in pipe 1 will always be along 17. Further, as H3 is minimum, flow
in pipe 3 will always be along 73. However, the direction of flow in pipe
72 depends upon the relative values of Hj and H2. If //, = H2 (a rare
case), Q2 = 0 and Qx = Q3. If //, > H2, Q2 ( > 0) will be from 7 to 2
and therefore Ql = Q2 + Q3. If //, < H2, Q2 ( > 0) will be from 2 to 7;
and therefore Qi + Q2 = Q3. Which of these conditions exists is decided

IT

Figure 5.1 Three-reservoir system.

 Reservoirs 93

by assuming ft = ft and finding whether Qi > Qs, Q\ — fi», or

Qi < Q3. Having decided this, the value of ft is successively assumed
until the appropriate condition is satisfied. The problem can aiso be solved
as illustrated in Example 5.2.

Example 5.2

For the three-reservoir system shown in Figure 5.1, /A = 120.00 m,

H2 = 92.00 m, and H3 = 80.00 m. If the length, diameter, and CHW
coefficient for these pipes are as given in Example 5.1, find the discharge
in each pipe and the HGL at J.

Solution

From Example 5.1, /?, = 222.91, R2 = 493.90, and R3 = 148.61. Tak-

ing ft = H2 = 92.00 m,

/120 92\1/1852
ft
- hasr) = a3262 mVs
and

= ° 2 5 7 0 m"/
As g , > <93, / 6 > ft and Q, = Q2 + Q3. Let Q3 = Q and Q2 = mQ
so that <2, = (1 + w)(2. Considering path 1-7-2,

852
222.911(1 + m)QY + 493.9O(m0' 8 5 2 = 28 (1)

Similarly, for path 1-7-3,

222.91[(1 + /w)GJl 852 + 148.6K21852 = 40 (2)

Dividing Equation (1) by Equation (2),

222.91(1 + my 852 + 493.90m 1852 _

222.91(1 + my 852 + 148.61 ~ '

On simplification, 66.873(1 + m) 1852 + 493.90m 1852 - 104.027 = 0.

94 RESERVOIRS, PUMPS. AND SPECIAL VALVES

Solving by trial and error, m — 0.1582, which when substituted in Equa-

tion (2) gives Q = 0.27356 m3/s and mQ = 0.04328 m3/s.

.-. Qx = (1 + 0.1582)0.27356 = 0.31684 nrVs; Q2 = 0.04328 m 3 /s;

Q3 = 0.27356 m 3 /s; and

Hj = 120.00 - 222.91(0.31684)'852 = 93.47 m

5.1.4 MULTIPLE-RESERVOIR SYSTEM

Complex systems are formed when several reservoirs are connected by

pipelines having one or more junctions. A four-reservoir system is shown
in Figure 5.2. In Figure 5.2(a) the reservoirs are connected at a common
junction point 7, while in Figure 5.2(b), there are two junction points Jy
and 7 2 .
Multiple reservoir systems can be solved by a trial-and-error procedure.
For the four-reservoir system in Figure 5.2(a), if//, > H2 > H3 > H4,
Hj is successively assumed so that it is equal to the HGL at the interme-
diate reservoirs, i.e., H2 and H3, to determine whether //, lies between //,
and H2, H2 and H3, or H3 and H4, by considering the flow continuity at
junction 7. Later, finer adjustment of Hj is made so that the flow continuity
equation is exactly satisfied at 7. For the system in Figure 5.2(b), HGL at
one of the junctions, say 7,, is assumed and the flows in pipes 1 and 2 are
determined. Applying the flow continuity equation at 7,, the flow in pipe
7i72 and HGL at 7 2 are determined. Flows in pipes 3 and 4 are subse-
quently determined to check the continuity equation at 7 2 . If the continuity
equation is not satisfied at 72, a new assumption for HGL at 7, is made.
The HGL value at 7, is increased if the net inflow at 72 is positive and vice
versa. The procedure is continued until the flow continuity equation at 7 2
is satisfied.

(a) (b)
Figure 5.2 Four reservoir system: (a) with one junction point, (b) with two junction points.
 Reservoirs 95

This procedure can be extended to systems having more than four reser-
voirs. However, such systems form a particular case of general pipe net-
works, and therefore network analysis procedures, described in later chap-
ters, can better be employed instead of the trial-and-error procedure.

Example 5.3

In the four-reservoir system of Figure 5.2(b), //, = 120.00 m, H2 =

90.00 m, H3 = 130.00 m, and H4 = 80.00 m. If /?, = 200, R2 = 400,
R3 = 150, R4 = 300, and RJlJ2 = 250, and the head loss is given by
hf = RQ2 in which hf is in meters and Q is in cubic meters per second,
find the discharge in each pipe and the HGL values at Jx and J2.

Solution

For the first trial let HJl = 100.00 m. Therefore

/ ioo - 90 \ 0 5
Q2 (from J, to 2) = — = 0 . 1 5 8 1 nWs

\ 400 J

:. QJlJ2 = 0.3162 - 0.1581 = 0.1581 m 3 /s; and

Hj2 = 100 - 250 (0.1581) 2 = 93.75 m

/ 130 - 93 75 \ 0 5
Q3 (from 3 to J2) = " ^ 4 ^ ~ ^ = 0 4 9 1 6 m3/s; and

AtJ 2 , net inflow = 0.1581 + 0.4916 - 0.2141 = 0.4356 nrVs

As the net inflow at J2 is positive, ///, should be increased. For the next
trial, let HJl = 110.00 m.

.-. C = 0.2236 nrVs, Q2 (from Jt to 2) = 0.2236 mVs

QjlJ2 = 0, Hj2 = 110.00 m, Q3 = 0.3651 mVs and Q4 = 0.3162 nvVs

At7 2 , net inflow = 0 + 0.3651 - 0.3162 = 0.0489 nxVs ( * 0)

96 RESERVOIRS, PUMPS, AND SPECIAL VALVES

Proceeding in the same manner and using interpolation, ultimately, Qi =

0.2065 m 3 /s, Q2 = 0.2317 m 3 /s, G,2y, = 0.0252 rrrVs, Q3 = 0.3500
nrVs, Q4 = 0.3247 m 3 /s, //,, = 111.47 m and HJ2 = 111.63 m.

5.2 PUMPS

5.2.1 INTRODUCTION

A pump is provided in a pipeline to supply the extra head to lift water

from a lower level to a higher level. In pipe networks, pumps may be pro-
vided externally to serve as supply pumps to provide water from external
sources to the pipe networks, or may be provided internally within the net-
work as booster pumps to boost up pressure at some points within the
system.
A typical pump installation lifting water from a lower level reservoir 1
to higher level reservoir 2 is shown in Figure 5.3. Writing Bernoulli's equa-
tion [Equation (1.18)] for points 1 and 2,

7 2#

+ E (friction losses)

+ £ (minor losses) (5.1)

in which hp — net head delivered by the pump. The velocity heads are

head ~i z ...

6'7 8
Figure 5.3 Typical pump installation.
 Pumps 97

-o y
o
QJ /
x: / /
£ .^fTbtal
V)
______^_—-""hea Gloss
>*
CO
Static head
t
Discharge Qp-»-
Figure 5 4 Typical system head-discharge eur\c.

small and therefore they can be neglected. Further, p, = /j 2 (equal to the

atmospheric pressure). Therefore, from Equation (5.1)
hp = (Z2 — Z,) + E (friction and minor losses), or

K = //< + /;, (5.2)

in which hs = static head (equal to Z2 - Zx) and /?., = total head loss.
The sum of the static head and total head loss is termed the system head.

5.2.2 SYSTEM HEAD-DISCHARGE CURVE

As long as the water level in the two reservoirs [Figure (5.3)] remains
unchanged, the static head remains constant. However, the total head loss
which comprises frictional head losses and minor head losses varies with
the discharge Qp through the pump. Thus, the system head /;, can be ex-
pressed as

hj = /?, + f(Qp) (5.3)

A curve showing the relationship between the system head hx and the
discharge Qp is known as the system head-discharge curve. A typical
system head-discharge curve is shown in Figure 5.4.

5.2.3 PUMP HEAD-DISCHARGE CURVE

At a constant rotational speed, a pump has a unique relationship be-

tween the net delivered head hp and its capacity, i.e.. the discharge Q,,,
supplied by it. The curve showing this relationship is known as pump
head-discharge curve. A typical pump head-discharge curve is shown in
98 RESERVOIRS, PUMPS, AND SPECIAL VALVES

Pump head-discharge curve

O
o
a> ^-Actual h p
si
ex.

System head-discharge curve

t
Actual Qp—
Capacity Qp-*-
Figure 5.5 Typical pump head-discharge curve.

Figure 5.5 in which the system head-discharge curve of Figure 5.4 is also
shown. The point of intersection of the system-head curve and the pump-
head curve is the actual operating point having discharge Qp that gives
K = hT.
Head-discharge curves for different types of pumps are shown in Figure
5.6. The pump head-discharge curve in Figure 5.6(a) is strictly mono-
tonically decreasing, i.e., the head decreases as the discharge increases.
Thus, for a given head there is only one value for the discharge (point a).
Such a curve is a stable head-discharge curve and is obtained for screw
pumps. The head-discharge curves in Figures 5.6(b) and 5.6(c) which are
obtained for centrifugal pumps and half-axial pumps, respectively, are not
monotonically decreasing. Thus, for the same head, two discharges
[points b and c in Figure 5.6(b)l or three discharges [points d, e, and f in
Figure 5.6(c)] may exist. Such curves having two or more discharges for
the same head are unstable head-discharge curves.
Two typical system curves are also shown in Figure 5.6. In Figure 5.6(a),

Discharge Qp Discharge Qp Discharge QP

(a) (b) (0
Figure 5.6 Head-discharge curves for different types of pumps: (a) stable, (b) unstable with two
discharges, (c) unstable with three discharges.
 Pumps 99

there is always one intersection point and thus only one discharge exists for
a particular head. For pump curves in Figure 5.6(b) and 5.6(c), even
though system curve 2 has one intersection point, system curve 1 has two
and three intersection points, respectively. Thus, the condition hp = /?,,
which gives the operating points, gives no insight as to which of these
operating points is more likely to occur. Therefore, a distribution network
having several pumps having unstable head-discharge curves would have
multiple operating points, and the analysis and operation of such networks
would be uncertain as stated by Collins et al. [75,76]. In practice, however,
improvement in the system design with provision of pumps having stable
head-discharge curves (except perhaps at very small percentage of dis-
charge) can ensure a single operating point and thus a unique steady-state
flow condition. Therefore, it is quite common in pipe network analysis to
assume that the pumps have stable head-discharge curves. Therefore, it is
assumed throughout in this text that the head-discharge curves for pumps
are stable giving unique operating points.

5.2.4 HEAD-DISCHARGE RELATIONSHIP

Several types of relationships have been suggested to mathematically ex-

press the pump head-discharge curves. Some relationships express hp as a
function of Qp while others express Qp as a function of hp.

5.2.4.1 hp as a Function of Qp

It is common to use a relationship

hp = AQP + BQP + Ho (5.4)

in which A, B and Ho = constants, determined by fitting Equation (5.4) to

three points taken from the expected working range of the pump head-
discharge curve. To improve the convergence characteristics in network
analysis, Jeppson and Travallaee [77] suggested a transformation

B_
P = QP + — (5.5)
2A

so that Equation (5.4) becomes

hp = AGl + \ H o - ^ \ (5.6)

The convergence is improved because the exponent 2 of G;, in Equation

100 RESERVOIRS. PUMPS, AND SPECIAL VALVES

(5.6) is close to the value of n in general head loss relationship of a pipe,

Equation (2.37).
A relationship of the form

hp = Hp- RPQ"P (5.7)

in which Hp, Rp = constants can be used. Here Hp can be considered as

the shut-off head, i.e., the maximum head that can be provided by the
pump as Qp approaches 0 and Rp as the resistance constant of the pump.
The exponent n may be taken the same as the one in the pipe head loss
relationship used in the analysis (Table 2.7). By selecting points in the ex-
pected working range of the pump, the values of Hp and Rp are obtained by
regression analysis.

5.2.4.2 Qp as a Function of hp

Donachie [78J used a fourth degree polynomial

Qp = A + Bhp + Ch2p + Dhp + Ehp (5.8)

in which A, . . . , E = constants obtained from regression analysis.

Chandrashekar [79] suggested a relationship

QP = Go - ahbp (5.9)

in which a, b = constants, and Qo = a constant discharge.

For better convergence, if the exponent b is taken as 1/n, Equation (5.9)
becomes

QP = QO- ahy (5.10)

However, this changes the nature of the curve, i.e., Qp becomes a convex
function of hp instead of being a concave function. Thus, fitting a curve
given by Equation (5.10) would be improper.
Using Equation (5.4) and its modified version, Equation (5.6),

or
 Pumps 101

For Equation (5.11), the nature of the curve is maintained and therefore
curve fitting is accurate and convergence is also improved.
If Equation (5.7) is used, a much simpler relationship between Qp and
hp is obtained. Thus,

HP - h,. (5.12)

which maintains the nature of the curve, is fairly accurate, and also im-
proves convergence.
By applying polynomial regression to several pump curves, Fietz [80.
67] observed that a third degree polynomial (hp as a polynomial of Q,) is
usually the best fit. However, Jeppson and Davis |68J have observed that
a second degree polynomial is usually sufficient in practice.
Typical head-discharge data and the corresponding curve are shown in
Figure 5.7(a). The discharge-head curve for the same data is shown in

20

£ . 1 5 --
O

Qp 0-0 0-1 0-2 0-3 04

5-- \
18-0 16-5 147 11-9 8-0
l 1 i

0-1 0-2 0-3 0-4

Discharge, m//s
(a)

5 10 15 20
Pump head, m
(b)
Figure 5.7 Pump data and curves: (a) head-discharge curve, (b) discharge-head curve.
102 RESERVOIRS, PUMPS. AND SPECIAL VALVES

Figure 5.7(b). It is seen that both the curves are concave. Assuming the
working range of the pump to be 0.1-0.3 m3/s the analytical relationships
described earlier are given below:

Equation (5.4): hp = 17.3 - 3Q - 50Q2p

Equation (5.6): hp = 17.345 - 50 Gj,, in which Gp = Q,, + 0.03

Equation (5.7): hp = 17.18 - 49.950J 8 6 6

hp = 17.04 - 5 7 . 3 5 0 (« = 2 for DW and Man-

ning formulas)

hp = 17.19 - 49.20G; 852 (n = 1.852 for HW for-

mula)

Equation (5.9): Qp = 0 . 3 8 1 - 6.512 X lO"6/?3,-807

Equation (5.10): Qp = 1.409 - 0.3197/?° 5 (n = 2 for DW and Man-

ning formulas)

Qp = 1.321 - 0.2664/?° M (n = 1.852 for HW for-

mula)

Equation (5.11): Qp = (0.3469 - 0 . 0 2 / g ° 5 - 0.03

Equation (5.12): Qp = (0.3439 - 0.02/g 0 - 535

Qp = (0.2971 - 0.0174/!,/ 5 (n = 2 for DW and

Manning formulas)

Qp = (0.3494 - 0.02033h,,) 054 (n = 1.852 for HW

formula)
5.2.5 CHARACTERISTIC CURVES

To completely describe the performance of pumps, a group of curves

showing the head capacity, power capacity and efficiency capacity are used
(Figure 5.8). These curves known as pump characteristic curves are ob-
tained from pumping tests and can be obtained from the manufacturers.
The power delivered by a pump is given by
 Pumps 103

Power, kW
T70
Efficiency -60
%
r100 -50
•90 -40
•80 -30
-70 -20
•60 -10
1
•-50
0-05 0-10 0-15 0-20
Discharge or Capacity, n
Figure 5.8 Typical pump characteristic curves.

in which P = power in kilowatts; y = specific weight of the liquid in

newtons per cubic meter; Q = discharge in cubic meters per second;
hp = head in meters; and rj = efficiency.
5.2.6 PUMP COMBINATIONS

Pumps may be combined so that they run in series or in parallel. When

the pumps are connected in series, they carry the same discharge but the
heads supplied by them are added [Figure 5.9(a)]. When the pumps are
connected in parallel, the heads supplied by them are the same but the dis-
charges are added [Figure 5.9(b)].
A series combination of pumps is used in high buildings where water is
raised in steps to higher floors through booster pumps at intermediate
floors. Because of the series combination, the required head is obtained

a
d. PumpBX
•"~-<\ -Combination

t PumpA\^
\
• Capacity •Capacity
(a) (b)
Figure 5.9 Two pump combinations: (a) series, (b) parallel.
104 RESERVOIRS, PUMPS. AND SPECIAL VALVES

40 -i 1 r
Tank full, source at low level •

N 2&3
1&3
1&2
Pump/1
-Tank empty, source at high level

. CD
XI

40 80 120 160 200 240

Discharge, ML/d

Figure 5.10 Three pump parallel combination.

without exerting excessive pressures in the pipes and pump casings of the
lower floors.
Parallel combinations are commonly used for lifting water that varies in
quantity (from minimum demand to maximum demand) and head (source
at high level and service reservoir empty to source at low level and service
reservoir nearly full). Figure 5.10 shows the head-discharge curves for a
three-unit parallel combination [81,82]. Pumps of different capacities are
installed so that they can be run in different combinations to obtain maxi-
mum efficiency.
Example 5.4

A pump lifts water from a sump (water level = 100.00 m) to an elevated

reservoir (W.L. = 110.00 m) through a 200-m long, 300-mm diameter
pipe (e = 0.26 mm). Using the pump data of Figure 5.7 and taking v = 1
mm2/s for water, determine the discharge to the elevated reservoir. Use the
Swamee-Jain relationship for Darcy-Weisbach formula.
Solution

Assuming the working range for the pump to be 0.1-0.3 rrrYs, from Sec-
tion 5.2.4,

hp = 17.04 - 57.35 Q2P

 Pumps 105

For the first trial, assuming the flow in the pipe to be rough turbulent,
from Equation (2.11b),

0.26
- , . \ / = 0.01897
3.7 X 300

0.01897X200
12.1 X (0.3)5

As the static lift is 120.00 - 110.00 = 10.00 m, the system head, hT =

10 + 129.030. At the operating point, hp = hT, and therefore, 17.04 -
57.350 = 10 + 129.030. This gives 0 = 0.1944 m3/s, which is
within the assumed working range of the pump.

AQ_ 4 X 0.1944
Re = = 825,060
•nDv TT x 0.3 x 1 x 10-6

From the Swamee-Jain relationship, [Equation (2.20)],/ = 0.01948, giv-

ing h, = 132.480. Therefore, 17.04 - 57.350 = 10 + 132.480,
which gives 0 = 0.1926 m3/s. For 0 = 0.1926 m3/s, Re = 817,420,
/ = 0.01984 as before, .*. 0 = 0.1926 m3/s.
Example 5.5

A pump lifts water from a low level reservoir (W.L. = 80.00 m) to an

elevated reservoir (W.L. = 90.00 m) as shown in Figure 1: Example 5.5.
W.L. 90.00 m
90° elbow
K = 0.90

Delivery Pipe
Suction pipe * - (Pipe 2)
(Pipe
90° elbow 90° elbow
K = 0.9 K = 0.90
W.L. pump Gate Valve
80.00 m fully open
Foot Valve K = 0.19
K = 2.50
The suction pipe is 30 m long, 300 mm in diameter, and the delivery pipe
is 150 m long, 250 mm in diameter. Both pipes are new cast iron with
e = 0.26 mm. Several appurtenances with their K values are located on
the pipeline as shown in the figure. The pump head-discharge data are
given in Table 1: Example 5.5. Develop the system head-discharge curve
106 RESERVOIRS, PUMPS, AND SPECIAL VALVES

TABLE 1. EXAMPLE 5.5.

Discharge, Head,
m3/s m

0 30.8
0.02 30.9
0.04 30.8
0.06 29.0
0.08 27.0
0.10 23.0
0.12 17.3
0.14 9.0

and determine the steady state head and flow. Use the Zigrang-Sylvester
relationship for determining/for the DW formula. Take v for water = 1.3
mm 2 /s.
Solution

The static head hs = 90.00 - 80.00 = 10.00 m. From Equation (2.5),

the total frictional loss

8
> < 3 0 , 8 X 150 1
2 _2 2 w 2
•K X 9.81 X {03f - " 7T n oi
X 9.81 v^
X /r\ IJ\SJ
(0.25) *^

= (1020/, + 12,691 f2)Q2

Total minor losses
=
" m "foot-valve "1~ " e l b o w "I" " g a l e valve "I" 2.<l<:]bo\>. "I" ^c\il

8 8 1
2 r-v4 ( t - v a l v e + Kcibov,) H " — (Aj,a,c.valvc + 2ft e |[, o w + A^.xil) \Q

X 9.8? X

= 97.93g 2
.-. System head, hT = 10.00 + (1020/, + 12691/2 + 97.93)g 2
 Pumps 107

The system head is obtained for different values of discharge as shown

in Table 2: Example 5.5.
The pump head-capacity curve and the system head-discharge curve are
plotted in Figure 2: Example 5.5. The intersection point of these curves

30'

25

-§20
<v ating head 15-7 m
/
"Q-15

0- 10 r

5 Operating discharge,0-124rr^7
0 002 004 0-06 0-08 0-10 0-12 0-14
Discharge.mVs
gives the operating point for which head = 15.7 m; and discharge =
0.124 m3/s.
Example 5.6

In Example 5.1, reservoir 1 is replaced by a sump with a W.L. at 90.00

m. A pump is fitted in pipe 1 as shown in Figure: Example 5.6. The pump

WL 90.00 m

characteristic curve is as given in Figure 5.7. The pipe and other data are
the same as that in Example 5.1. Determine the different values as in Ex-
ample 5.1, if the outflow at 3 is (1) 0.1 m3/s and (2) 0.4 nrVs.
108 RESERVOIRS, PUMPS, AND SPECIAL VALVES

TABLE 2. EXAMPLE 5.5.

0, £/»„ K K
m^/s h m m m m m
0 0 0 0 10.00 10.00
0.02 0.02280 0.02286 0.13 0.04 0.17 10.00 10.17
0.04 0.02177 0.02152 0.47 0.16 0.63 10.00 10.63
0.06 0.02053 0.02100 1.04 0.35 1.39 10.00 11.39
0.08 0.02018 0.02073 1.82 0.63 2.45 10.00 12.45
0.10 0.01995 0.02056 2.81 0.98 3.79 10.00 13.79
0.12 0.01980 0.02044 4.03 1.41 5.44 10.00 15.44
0.13 0.01974 0.02040 4.72 1.66 6.38 10.00 16.38
0.14 0.01969 0.02036 5.46 1.92 7.38 10.00 17.38

Solution

From Example 5.1, fl, = 222.91; R2 = 493.90; and R3 = 148.61.

Case 1

Outflow at 3, q3 = 0.1 nrVs

Assuming Hj = 100.00 m, we get Q2 = 0 and Qx = 0 . 1 nvVs. For

the Hazen-Williams formula, the head-discharge curve for the pump (Sec-
tion 5.2.4) is

hp = 17.19 - 49.20C; 8

Therefore,

Hj = 90 + [17.19 - 49.20 X (O.I) 1852 ] - 222.91 X (O.I) 1852

= 103.36 m > 100.00 m

Thus, the direction of flow in pipe 2 is from J to 2. Let the discharge

in pipe 1 be Q so that Q2 = Q — 0.1, and Q3 = 0.1 m3/s. Considering
path l-J-2, 90 + (17.19 - 49.2O0 1852 ) - 222.91 Q 1 8 5 2 - 493.90 X
(Q - 0.1)1852 = 100 or, 272.11G1852 + 493.90«2 - 0.1)1852 = 7.19.
Solving by trial and error, Q =, 0.1316 m3/s. Therefore, Q, = 0.1316 nvVs;
Q2 (from J to 2) = 0.0316 m 3 /s; //, = 100.83 m; and H3 = 98.74 m.
Case 2

Outflow at 3, <jr3 = 0.4 m3/s

 Pumps 109

Following the procedure outlined for case 1, the direction of flow in pipe
2 is found to be from 2 to J, and the required values obtained are as fol-
lows: Ql = 0.2530 m 3 /s; Q2 = 0.1470 m3/s; //, = 85.84 m; and
H3 = 58.61 m.
Example 5.7

In the three-reservoir system of Example 5.2, reservoir 1 is replaced by

a sump, and a pump is fitted in pipe 1 as shown in Figure: Example 5.7.

WL 92.00 m

Sump

The pump characteristic curve is as given in Figure 5.7. The pipe and other
data are the same as that in Example 5.2. Determine the different values
as in Example 5.2, when the sump W.L. is (1) 80.00 m and (2) 98.00 m.
Solution

From Example 5.1, /?, = 222.91, R2 = 493.90, and R3 = 148.61. From

Equation (5.7), for the pump, hp = 17.19 - 49.20 0 8 5 2 .
Case 1

Sump W.L. = 80.00 m

Let Hj = H2 = 92.00 m, so that Q2 = 0. Therefore, for pipe 1, 80 +

(17.19 - 49.200! 852 ) - 222.91(2! 852 = 92.00, which gives Q, = 0.1179
m3/s. Similarly, for pipe 3, 92.00 - 148.61 £>V8" = 80.00, which gives
Q3 = 0.2570 m3/s. As 0 , < Q3, Hj < 92.00 m so that Q, + Q2 = Q3.
L e t g i = Q, Q2 = mQ so that Q3 = (1 + m)Q. Considering path 1-7-3

80 + (17.19 - 49.20C?1852) - 222.91 £?1852 - m)QY

= 80.00, i.e.,

m)Q]18 272.11(2'852 = 17.19 (1)

110 RESERVOIRS. PUMPS, AND SPECIAL VALVES

Similarly, for path 2-7-3

92 - 493.9O(m0' 8 5 2 - 148.61|(1 + m)QY 852

= 80
852
i.e., 148.611(1 + m)QY + 493.90(m©' 8 5 2 = 12 (2)

Dividing Equation (1) by Equation (2), simplifying and solving, we get

m = 0.4290, and therefore, 0 , = 0.1524 m3/s; Q2 = 0.0654 m 3 /s;
Q3 = 0.2178 nrVs; and Hj = 88.84 m.
Case 2

Sump W.L. = 98.00 m

Following the procedure outlined for case 1, we get Qt > Q3,
Hj > 92.00 m and Q, = Q2 + Q3. Let Q, = Q, Q2 = mQ, so that
Q3 = (l — m)Q. Considering paths 1-7-2 and 1-7-3, we get a two-
equation solution which yields m = 0.0255, and Q = 0.2643 mVs.
Therefore, £>, = 0.2643 mVs; Q2 = 0.0068 nWs; Q3 = 0.2575 m3/s;
and Hj = 92.04 m.

5.3 SPECIAL VALVES

Valves and the head losses through them are considered in Chapter 3.
Though all valves cause loss of head when flow takes place through them,
some valves are simple as far as their inclusion in pipe network analysis
is concerned. For example, globe valves, gate valves, and such other
valves control the magnitude of flow through them. They can be con-
sidered in the analysis by using appropriate head loss coefficients (Table
3.1) depending upon how far open the valves are. It is assumed that the
coefficients remain the same regardless of the magnitude of the discharge
and the direction of flow through them. Some valves, on the other hand,
have multiple actions. They control the direction of flow through them and
reduce the water pressure to a preset value. Check valves and pressure re-
ducing valves are examples of this type and require special treatment in the
analysis. These valves are now considered.

5.3.1 CHECK VALVES

A check valve allows flow in one direction only. It opens when the flow
is in the desired direction and closes when the flow is in the opposite direc-
tion, thus preventing flow. Check valves are mainly used on suction and
delivery sides of pumps. Afoot valve is a check valve fitted on the inlet of
 Special Valves 111

w
(a)

Hi>Hj
(b)
Qx=O

(C)
Figure 5.11 Check valve: (a) fitted in pipe connecting nodes / and./, (b) //, > //,-, (c) //, < //,.

the suction pipe of a centrifugal pump. It holds water in the suction pipe
and thus obviates or reduces the priming effort.
Different situations arising due to the presence of a check valve are
shown in Figure 5.11. Figure 5.11(a) shows a check valve fitted in pipe x,
connecting nodes / and j , with HGL values //, and Hj, respectively. The
check valve in Figure 5.11(a) allows flow from node i to node j . When
Hi > Hj, flow takes place from / to j and Qx > 0 [Figure 5.11(b)J. When
Hi < Hj, the check valve closes and prevents flow from j to i so that
& = 0 [Figure 5. ll(c)J.
Example 5.8

Solve Example 5.1, assuming that a check valve (K = 2.5) is fitted in

pipe 2 as shown in Figure: Example 5.8.
WL120.00m

1 WLiOO.OOm
2

Solution

From Example 5.1, /?, = 222.91, R2 = 493.90, and R3 = 148.61. For

the check valve, as K = 2.5,

r>
8A: 8 x 2.5
ir2gD4 7T2 X 9.81 x (0.2) 4
= 129.10
112 RESERVOIRS, PUMPS, AND SPECIAL VALVES

From Example 5.1, reservoir 2 starts supplying water when q3 > 0.2720
rrrVs.

Case 1

Outflow at 3, q3 = 0.2 m3/s

As Q3 < 0.2720 m3/s, the check valve prevents flow from J to 2 in

pipe 2. Therefore, g , = 0.2 m3/s, Q2 = 0, Hd = 120.00 - 222.91 X
1852
(0 2)..8 S 2 = 1 0 8 6 9 m ; a n d Hj _ 1 0 8 6 9 _ i 4 8 6 1 x (0.2) = 101.14 m.

Case 2

Outflow at 3, q3 = 0.4 nvVs

As q3 > 0.2720 nvVs, reservoir 2 supplies water. Let the discharge

in pipe 1 be Q so that Q2 (from 2 to J) = 0.4 — Q. Considering path
l-J-2 and the head loss in the check valve, 120 - 222.91£) 1852 +
493.9(0.4 - QY 852 + 121.1(0.4 - Qf = 100. Solving by trial and er-
ror, Q = 0.31421 m a /s. Therefore, Q, = 0.31421 nWs; Q2 = 0.08579
m 3 /s; Hj = 120.00 - 222.91 X (0.31421)1852 = 93.88 m; and H3 =
93.88 - 148.61 X (0.4)' 852 = 66.65 m.
Because of the presence of the check valve in pipe 2, reservoir 2 is no
longer floating on the system, and supplies water only when q3 > 0.2720
nrVs.

5.3.2 PRESSURE REDUCING VALVES

A pressure reducing valve (PRV) maintains a constant preset pressure

downstream of it regardless of how large the upstream pressure becomes
[83,14,79]. Thus, several PRVs together can reduce pressure in an exces-
sive zone of a pipe network. For example, if PRVs are provided in pipes
5 and 6 of the network of Figure 5.12, they can control the pressures at the
downstream nodes 3 . . . 6.
There are two exceptions to this usual, pressure-reducing behavior of a
PRV. These exceptions are the following:
(1) If the pressure becomes less than the pressure setting of the PRV, the
PRV becomes inoperative and has no effect on the flow, except for
the head loss through it. The head loss coefficient for a PRV, i.e.,
KRRV, can be taken as 10 [67,68].
(2) If a PRV is bypassed and the downstream pressure exceeds its
pressure setting, the PRV acts as a check valve, prevents flow in the
 Special Uilves 113

" 2 5 8
Pipe,i 4 7 10

3 6 9
8 7 6
Figure 5.12 A network.

opposite direction, and allows the pressure immediately downstream

of it to exceed its pressure setting. In this case, several PRVs can ef-
fectively control the sources from which supply is withdrawn under
different demand conditions. As the network demand increases, ad-
ditional sources of supply become operative.
Consider a PRV fitted in pipe x, connecting nodes / and j with HGL
values Hi and Hj respectively [Figure 5.13(a)J. The PRV is oriented such
that flow can take place along y only. Let u and d represent the upstream
and downstream ends of the PRV, respectively. Let Hu and Hd be the HGL
values at u and d, respectively. Let //sel be the preset HGL value at the
downstream end d of the PRV.

Hset
d
(Q)

i Hi>H S e t »H d >H j
(b)
9-
i Hi>H d >Hset>Hj j
(c)

o- CU=O c
i Hi>Hj>H Se t J
(d)

9-
Hi<Hj
(e)
Figure 5.13 Pressure reducing valve: (a) fitted in pipe connecting nodes 1 and ;, (b) H, >
««, a Hd > Hj, (c) //, > Hd> HM > H,, (d) H, > H, > H^, (e) H, < Ht.
114 RESERVOIRS, PUMPS, AND SPECIAL VALVES

The following four modes of operation may occur for the PRV:
(1) Hi > Hsa > Hd > Hj [Figure 5.13(b)]: As H > Hj, flow takes
place from / toy. Further, as Hd < //sel, the PRV is inoperative. If
A"PRv is the head-loss coefficient, the resistance constant of the PRV,
i.e.,
8ATp RV
**PRV —

The PRV acts as an ordinary valve with fixed resistance constant,

/?PRV, and the head loss through it is hLpRV = /? P R V 0.
(2) Hi > Hd > Ha > Hj [Figure 5.13(c)]: As //, > Hj, flow takes
place from / to j . As Hd > //seI, the PRV is operative and Hd be-
comes equal to Ha. The head reduction through the PRV is Hu —
Hxl. The PRV has a variable resistance so that

'H — H \05
u
r» I •'set 1
"PRV = I ^ I >

(3) Hi > Hj > Hxx [Figure 5.13(d)]: It may happen that a PRV is
bypassed so that //, becomes more than //„,,. As //, > //,-, flow
should take place from / toy. However, as the intervening //se, is less
than Hj, the PRV prevents flow, i.e., Qx = 0. The pressures on the
upstream and downstream ends of the PRV increase so that Hu = //,
and Hd = Hj.
(4) Hi < Hj [Figure 5.13(e)]: As //, < //,, flow direction would have
been fromy to / in the absence of a PRV. However, the PRV shuts off
when Hi < Hj and prevents flow from j to /. The PRV behaves as a
check valve so that Qx = 0, //„ = #, and Hd = Hj.
Example 5.9

In the network of Example 5.1, PRV-1 (K = 10) is fitted in pipe 1 at a

distance of 100 m from point 1; and PRV-2 (K = 10) is fitted in pipe 2 at
a distance of 50 m from point 2 as shown in Figure: Example 5.9. The pipe

W.L. 120-00 m
JW.L. 100-00 m
 Special Valves 115

data remain the same as in Example 5.1. Determine the discharges in pipes
1 and 2 and all the HGL values for the following conditions:
(1) q3 = 0.4 m3/s; Hxll = 110.00 m; and HM.2 = 96.00 m
(2) q3 = 0.4 nWs; //set_, = 80.00 m; and //,cl2 = 75.00 m
(3) q3 = 0.3 m3/s; HM.X = 110.00 m; and Hxt2 = 80.00 m
For these PRV settings, what is the limiting value of q3 at which
reservoir 2 starts supplying water?
(4) q3 = 0.1 m3/s; HmA = 110.00 m; and Hsa2 = 80.00 m
Solution

From Example 5.1, /?, = 222.91, R2 = 493.90, and R3 = 148.61. Fur-

ther, as 1-M, = 100 m [ = (1/3) length 1-7], /?,_„, = (1/3) X 222.91 =
74.30; and Rdl.j = 148.61. Similarly, for 2-u2 [= (1/3) length 2-7],
R2.U2 = (1/3) X 493.90 = 164.63; and Rd2.} = 329.27.

10201
*

and

*PRV"2 =
^ x 9.8^ x (0.2)4

Case 1

q3 = 0.4 m 3 /s, H^ = 110.0 m, and Hsa2 = 96.00 m

Assume Hj = 96.00 m, the lower of the two PRV settings, so that

0 2 = 0. .-. For pipe 1, 222.91Q1 852 + 102.010 = 120 - 96 = 24,
which gives Qt < 0.4 nWs; /. H3 < 96.00 m and Qt + Q2 = 0.4
m 3 /s.
Assume that both the PRVs are inoperative. Considering path 1-7-2,
120.00 - 222.9K2! 852 - 102.010 + 493.90(0.4 - Ql)lM1 + 516.42 X
(0.4 - 0i) 2 = 100.00 which gives 0 , = 0.2976 nWs, and Q2 = 0.1024
m3/s. To check the assumption that both the PRVs are inoperative, for
PRV-1, Hdl = 120.00 - 74.30 X (0.2976)1852 - 102.01 X (0.2976)2 =
103.09 m < 110.00 m. .'. PRV-1 is inoperative.
Similarly, for PRV-2, Hdl = 100.00 - 164.63 X (0.1024)1852 -
516.42 X (0.1024)2 = 92.17 m < 96.00 m. .\ PRV-2 is inoperative.
.'. The assumption that both the PRVs are inoperative is correct, and
116 RESERVOIRS. PUMPS, AND SPECIAL VALVES

therefore, Q, = 0.2976 nWs; Q2 = 0.1024 m3/s. The different HGL

values are obtained as//„, = 112.13 m; Hdl = 103.09 m; HU2 = 97.58 m;
Hd2 = 92.17 m; //, = 87.34 m; and H3 = 60.11 m. Both the PRVs are in-
operative and thus they are in mode 1 of operation.

Case 2

q3 = 0.4 nWs; //seM = 80.00 m; and H,a7 = 75.00 m

Since the PRV settings are low compared to the respective reservoir
water levels, assume that both the PRVs are operative. Further, from Case
1, Gi + Qi = 0.4. .'. Hdl = 80.00 m, and Hd2 = 75.00 m. For path
dt-J-d2, 80.00 - 148.610! 8 " + 329.27(0.4 - (?,)' 852 = 75.00 which
gives £>, = 0.2669 m 3 /s; and Q2 = 0.1331 m3/s.
For these discharges, if the PRVs were inoperative, Hdt = 120.00 —
74.30 x (0.2669)' 852 - 102.01 X (0.2669)2 = 106.30 m > 80.00 m.
.'. PRV-1 is operative and reduces the pressure from 113.56 m ( = //„,) to
80.00 m. Similarly, Hd2 = 100 - 164.63 X (0.1331)1852 - 516.42 x
(0.1331)2 = 86.92 m > 75.00 m. .'. PRV-2 is also operative and reduces
the pressure from 96.07 m ( = HU2) to 75.00 m.
The assumption that both the PRVs are operative is correct, and there-
fore, £>, = 0.2669 nrVs; Q2 = 0.1331 m3/s. The different HGL values
are: HUI = 113.56 m; Hdl ( = // set ,) = 80.00 m; HU2 = 96.07 m; Hd2
( = // scl2 ) = 75.00 m; Hj = 67.13 m; and H3 = 39.90 m. In this case
both the PRVs are in a pressure reducing condition, i.e., in mode 2 of op-
eration.

Case 3

q3 = 0.3 m 3 /s; //set_, =? 110.00 m; and Hxl2 = 80.00 m

Assume that //, = //sel.2 = 80.00 m, so that Q2 = 0. Also assume that

PRV-1 is inoperative. /. 120.00 - 222.910 1 8 " - 102.01 X Q2 = 80
which gives Q = 0.3313 m3/s ( > 0.3 nWs). .'. Hj > 80.00 m ( = // s e l 2 ).
Here PRV-2 is bypassed and is in mode 3 of operation, and Q2 — 0.
.-. Ql = 0.3 m 3 /s; Q2 = 0. The different HGL values are: HUI = 112.01
m; Hdx = 102.83 m (#set., = 110.00 m); Hj = 86.84 m; HU2 = 100.00
m; Hd2 = 86.84 m ( = / / , ) ; arid H3 = 70.86 m.
In this case, as is found earlier, when Hj = // sel2 = 80.00 m,
0 , = 0.3313 m3/s, and Q2 = 0. .\ When Q3 < 0.3313 nrVs, Q2 = 0,
and thus reservoir 2 does not supply water. When Q3 > 0.3313 nWs, res-
ervoir 2 supplies water. Thus, the PRV in pipe 2 determines when supply
should be withdrawn from reservoir 2.
 Problems 117

Case 4

q3 = 0.1 nrVs; HxxA = 110.00 m; and // sel2 = 80.00 m

From case 3, g , = 0.1 m3/s and Q2 = 0; HUI = 120.00 - 74.30 X

(0.1) 1852 = 1 1 8 9 6 m ; Hdi = 1 1 8 9 6 _ 102.01 x O.I2 = 117.94 m >
110.00 m ( = // sel .,). .'. PRV-1 is operative and Hdl = 110.00 m. Hj =
110.00 - 148.61 X (O.I)1852 = 107.91 m. Here //, > 100.00 m, the W.L.
in reservoir 2. As Ht < Hs for pipe 2, PRV-2 is in mode 4 of operation,
and it prevents flow from J to 2. /. HU2 = 100.00 m, Hdl = 107.91 m
( = Hj), and H3 = 107.91 - 148.61 X (0.1) 1 8 " = 105.82 m.

5.4 PROBLEMS

Problem 5.1: The length, diameter, and CHW coefficient for the pipes
in the system of Figure: Example 5.1 are as follows: Pipe 1—400 m, 350
mm, 80; Pipe 2—200 m, 250 mm, 110; and Pipe 3—300 m, 300 mm, 90.
Determine: (1) Demand at point 3 so that the discharge in pipe 2 is zero
and (2) discharges in pipes 1 and 2 and the HGL values at J and 3 when
the demand at 3 is (a) twice the value in (1) and (b) one-half the value in
(1).
Problem 5.2: For the three-reservoir system shown in Figure 5.1,
//, = 125.00 m, H2 = 103.00 m, and H3 = 87.00 m. If the length, diam-
eter, and CHW coefficient for the pipes are as given in Problem 5.1, find
the discharge in each pipe and the HGL at J.
Problem 5.3: In the four-reservoir system of Figure 5.2(a), //, =
130.00 m, H2 = 120.00 m, H3 = 110.00 m, and H4 = 100.00 m. If the
length, diameter, and CHw values for the pipes are: Pipe 1—200 m, 300
mm, 120; Pipe 2—300 m, 250 mm, 100; Pipe 3—400 m, 250 mm, 90;
and Pipe 4—250 m, 200 mm, 110; find the discharge in each pipe and the
HGL value at J.
Problem 54: The reservoirs and pipes of Problem 5.3 are connected to
form the system shown by Figure: 5.2(b). If the water levels in the reser-
voirs are the same as those in Problem 5.3, and if pipe 7, J2 is 300 m long,
300 mm in diameter with CHW = 110, find the discharge in each pipe and
the HGL values at Jt and J2-
Problem 5.5: A sump (water level = 107.20 m) is connected to an
elevated service reservoir (W.L. = 121.20 m) through a 500-m long, 350-
mm diameter new CI pipe (e = 0.26 mm). The pump fitted in this pipe
has discharge-head characteristics as follows: 0 m 3 /s-50.2 m, 0.01
m 3 /s-50.3 m, 0.02 nr7s-50.2 m, 0.03 m 3 /s^9.5 m, 0.04 m3/s-48.1 m, 0.05
m 3 /s-46.4 m, 0.06 m 3 /s-43.5 m, 0.07 nWs^lO.l m, 0.08 m 3 /s-36.4 m, 0.09
118 RESERVOIRS, PUMPS, AND SPECIAL VALVES

m 3 /s-32.5 m, 0.10 m 3 /s-27.9 m, 0.11 m 3 /s-22.3 m, and 0.12 m3/s-15.8 m.

Determine the steady-state supply to the reservoir.
Problem 5.6: A pump lifts water from a low level reservoir (W.L. =
95.00 m) to an elevated reservoir (W.L. = 112.70 m) as shown in Figure:
Example 5.5. The suction pipe is 20 m long, 200 mm in diameter with
CHW = 90. The delivery pipe is 200 m long, 300 mm in diameter with
CHW = HO. Several appurtenances with their K values are located on the
pipeline as shown in the figure. Using the pump head-discharge data as
given in Problem 5.5, determine the steady state head and flow.
Problem 5.7: In the three-reservoir system of Problem 5.2, reservoir 1
is replaced by a sump (W.L. = 68.00 m); and a pump described in Prob-
lem 5.5 is fitted in pipe 1. Determine the discharge in each pipe and the
HGL at J.
Problem 5.8: Solve Problem 5.3 with a pump described in Problem 5.5
fitted in pipe 3 so that reservoir 3 supplies water to junction J.
Problem 5.9: Solve Problem 5.4 with a pump described in Problem 5.5
fitted in pipe 4 so that reservoir 4 supplies water to junction J2.
Problem 5.10: Solve Problem 5.1 assuming that a check valve (K =
2.5) is fitted in pipe 2 when the permissible flow direction is from (1) J to
2 and (2) 2 to /.
Problem 5.11: Solve Problem 5.2 if a PRV (HM = 109.00 m) is fitted
in pipe 1 just upstream of the junction J and a check valve (K = 2.5) is
fitted in pipe 2 if the permissible flow direction is from (1) J to 2 and (2)
2 to J.
Problem 5.12: In the network of Problem 5.1, PRV-1 (K = 10) is fitted
in pipe 1 at a distance of 100 m from the junction J; and PRV-2 (K = 10)
is fitted in pipe 2 at 80 m from reservoir 2 as shown in Figure: Example
5.9. The pipe data remain the same as in Problem 5.1. Determine the dis-
charge in pipes 1 and 2 and all HGL values for the following conditions:
(1) q3 = 0.4 m 3 /s, #*,., = 110.00 m, and tfsel2 = 92.00 m
(2) q3 = 0.3 m 3 /s, # seM = 75.00 m, and Hxt2 = 85.00 m
(3) q3 = 0.2 nWs, H^ = 100.00 m, and //sel2 = 95.00 m
(4) q3 = 0.3 m 3 /s, //set_, = 80.00 m, and H^2 = 80.00 m
 Types and Parameters

6.1 DEFINITIONS

link is a segment of a distribution network which has a constant flow

A and no branches. A pipe is a segment of a link which has a constant
diameter. If a link is of uniform diameter throughout, it consists of one
pipe only and therefore the link and pipe are identical. A p-pipe series link
has p pipes connected in series. Similarly, a p-pipe parallel link has p
pipes connected in parallel. In Figure 6. l(a), AB and BC are pipes while
ABC is a link. If there is external flow (inflow or outflow) at B, as shown
in Figure 6.1(b), pipes AB and BC become links. Even though the terms
link and pipe are distinctly defined herein, it is common in water distribu-
tion networks to use them interchangeably.
A node is a point where two or more links meet, a link begins, or a link
ends [nodes B, A, and C, respectively, in Figure 6. l(c)]. A node is a junc-
tion node if two or more links meet, a starting node if a link begins and
an end node if a link ends. A source node or simply a source is a starting
node which receives flow from outside and supplies it to the network
[node A in Figure 6.1(c)]. A source may also supply a part of the flow it
receives to an outside agency for direct consumption. A distribution node,
a consumption node, or a demand node is a node which receives flow
through one or more links and distributes it [node C in Figure 6.1 (a),
nodes B and C in Figure 6. l(b)]. If a demand node distributes its inflow
partly externally and partly internally [node B in Figure 6.1(b)], or en-
tirely internally [no external demand, i.e., demand is zero as for node B
in Figure 6. l(c)], the node is an intermediate node. If a demand node does
not distribute internally but distributes only externally, the demand node
is a sink node, or simply a sink [node B in Figure 6. l(d)].
A link is a supply link if it supplies to a node and a distribution link if
it takesflowaway from a node. [For node B in Figure 6.1 (b), AB is a sup-
ply link while BC is a distribution link.] Thus, a link is a supply link for
119
120 TYPES AND PARAMETERS

Figure 6.1 Definitions.

a downstream node and a distribution link for an upstream node [link AB

in Figure 6. l(b) for nodes B and A, respectively.] A source has distribu-
tion links only while a sink has supply links only. An intermediate node
has both supply and distribution links. All the actual links in a network are
real links. However, imaginary links, termed pseudo links are sometimes
introduced as explained later.
A loop or a circuit is a closed figure obtained by starting from a node,
travelling only once along any of the connected links, passing only once
through any of the nodes and returning back to the starting node [closed
figures ABEFA, ABCDEFA in Figure 6.1(e)]. An independent loop, a
natural loop, a basic loop or a nonoverlapping loop is a loop which does
not overlap any other loop [loops- ABEFA and BCDEB in Figure 6. l(e)J.
A loop that overlaps two or more basic loops is a nonbasic loop or an
overlapping loop. [Loop ABCDEFA in Figure 6.1(e) is an overlapping
loop, overlapping basic loops ABEFA and BCDEB.] A pseudo loop is a
loop formed by one pseudo link and one or more real links.
A path is a combination of nodes and links and is obtained by starting
at a node and terminating at another node. It has one or more connecting
links and none or some intermediate nodes. In Figure 6.1(e), ABCD is a
path from node A to node D with B and C as intermediate nodes and AB,
BC, and CD as connecting links. Similarly, AFED, ABED and AFEBCD
are also paths from A to D. All these paths have intermediate nodes.
However, AB is also a path from A to B, but it has* no intermediate node
and has only one connecting link.
 Types of Networks 121

Source

Figure 6.2 Serial network.

A pipe network, a distribution network, or simply a network is the grid

formed by the nodes, links, and loops, if any. In practice, it may also con-
tain reservoirs, pumps, valves and other accessories. All the figures in
Figure 6.1 are networks.

6.2 TYPES OF NETWORKS

6.2.1 SERIAL NETWORK

A serial network is a network having no branches or loops (Figure 6.2).

It is the simplest of all the types of pipe networks. Generally, it has one
source, one sink, and one or more intermediate nodes. The first link starts
at a source and the last link ends on a sink. All the intermediate nodes are
connected by two links, a supply link on the upstream and a distribution
link on the downstream. The direction of flow in all the links is fixed and
is along the direction from the source to the sink. A path is a serial net-
work.

6.2.2 BRANCHING NETWORK

A branching or a dead-end network is a tree-like network without any

loops (Figure 6.3). It consists of several serial networks. Usually it has one
source, one or more intermediate nodes, and more than one sink. Each in-

Source
a

Figure 6.3 Branching network.

122 TYPES AND PARAMETERS

termediate node is connected on the upstream by one supply link and on

the downstream by one or more distribution links. The direction of flow in
all the links is fixed and is along the direction from the source to several
sinks.

6.2.3 LOOPED NETWORK

A looped network contains loops (Figure 6.4). It must have at least one
source as in a single-source network, but may have more sources as in a
multisource network. A looped network must have at least one sink.
(Why?) The node that serves as the sink depends upon the nodal demands.
For example, node F in Figure 6.4 is a sink. However, if demand at D in-
creases, the flow direction in links DE and DF may reverse and node D
would behave as a sink. Although the direction of flow in a link is fixed for
a particular demand pattern, it may change for another demand pattern.
Thus, for a node, a link which behaves as a supply link for one demand
pattern may behave as a distribution link for another demand pattern. Sim-
ilarly, when the direction of flow changes in a link, the upstream node
becomes the downstream node and vice versa.
Pipe networks seldom consist of loops only. In practice, they are of com-
posite type consisting of loops, branches, and serial parts. However, it is
usually possible to reduce a composite network to a looped network for
analysis, and therefore composite networks are referred to as looped net-
works.
In serial and branching networks, only one path is available for
transporting water from a source to a particular node. Thus, if a link of
such a network is closed for repairs or replacement, all demand nodes sit-
uated downstream of this link are completely cutoff from the source.
Therefore, supply through a serial or branching network is less reliable
when compared to that through -a looped one where alternate paths exist.
However, serial and branching networks are cheaper and usually easier to
analyze than looped networks. Serial and branching networks are used for
industrial water supply purposes, sprinkler irrigation projects, and for dis-

C D
Figure 6.4 Looped network.
 Parameters 123

tribution of water to small communities in rural areas. Looped networks,

although costlier and comparatively more difficult to analyze, are more
reliable than serial and branching networks, and therefore are used for
water distribution to large communities in urban areas.
Because serial and branching networks have similar analyses, the term
branching networks will be used to include serial networks in this text.

6.3 PARAMETERS

Several parameters are involved in the analysis of pipe networks.

6.3.1 CONFIGURATION

The configuration of a pipe network involves the location of sites for

various elements such as elevated service reservoirs, pumps, pipes, several
types of valves, and accessories. The configuration is decided by taking
into consideration the existing pattern of the streets and highways, existing
and planned subdivisions, property right-of-ways, possible sites for
elevated and ground service reservoirs, location and density of demand
centres, and general topography. In the analysis of pipe networks, however,
the geometrical layout of the network and the location of the several ele-
ments are considered known.

6.3.2 PIPE LENGTHS

The pipe lengths are obtained from the known geometrical layout of the
network. When nodes are connected by links consisting of pipes in series,
in parallel, and in series-parallel combinations, such links are usually re-
placed by equivalent pipes in network analysis.

6.3.3 PIPE DIAMETERS

The pipe diameters are either known or calculated for equivalent pipes.
However, some unknown pipe diameters can be obtained from the analysis
as explained later.

6.3.4 PIPE ROUGHNESS COEFFICIENTS

The pipe roughness coefficients such as the Hazen-Williams coefficient

CHW, Manning's coefficient jVand the Darcy-Weisbach friction factor/are
suitably assumed as described in Chapter 2, and therefore, they are treated
as known parameters. It should be remembered, however, that the CHW
124 TYPES AND PARAMETERS

and N values do not depend on pipe discharges and therefore remain con-
stant during the analysis. The Darcy-Weisbach friction factor/, however, is
a function of the Reynolds number and therefore of the pipe discharge, and
thus it must be reevaluated when the pipe discharge changes.

6.3.5 MINOR APPURTENANCES

The effect of the minor appurtenances described in Chapter 3 can be in-

dividually considered with the help of Table 3.1. However, for large net-
works, it is a common practice to consider equivalent pipe lengths (Table
4.1) and correspondingly increase the pipe lengths. For simplicity, the pipe
lengths are sometimes arbitrarily increased by 5-10 percent to account for
the effect of minor appurtenances [84J.

6.3.6 DEMAND PATTERN

The nodal demands fluctuate with the time of day, day of the week, and
the season and thus are variable [1,82]. However, it is a common practice
to assume that the demands remain constant in the analysis. Such analysis
is for steady state conditions and is termed static analysis as opposed to
dynamic analysis, in which fluctuations in demand are considered.

6.3.7 SOURCE SUPPLY PATTERN

As the inflows at the source, nodes depend upon the nodal demands
which are steady in static analysis, the source supply pattern is also steady
in the static analysis.

6.3.8 HYDRAULIC GRADIENT LEVELS AT DEMAND NODES

The hydraulic gradient levels or simply the heads at the demand nodes
are mostly unknown and obtained from the analysis. Therefore, they are
generally treated as unknown parameters.

6.3.9 HYDRAULIC GRADIENT LEVELS AT SOURCE NODES

When service reservoirs are the source nodes, the network is gravity-fed
and the heads at the source nodes are known. Thus, the source heads are
known and remain fixed in the static analysis.
In case of pumped source nodes, the water level in the sumps is known
and is assumed to remain fixed. However, the head supplied by the pump
to the network is a function of the discharge and therefore is not known.
As stated in Section 2.9, the length L, diameter D and the roughness
 Parameter Interrelationships 125

characteristic CFIW, / , or N can be grouped into one parameter, the pipe

resistance constant R. When L, D, and CMW (or/, or N) are known, R is
a known parameter. When any one of them is unknown, R is an unknown
parameter. Therefore, the parameters involved in hydraulic network analy-
sis are the (1) pipe resistance constants R, (2) nodal inflows or outflows q,
(3) nodal heads H, and (4) pipe discharges Q.

6.4 LABELLING NETWORK ELEMENTS

The main elements of a hydraulic network are the nodes, pipes, and
loops. Even though they can be labelled arbitrarily, a procedure for their
labelling is suggested herein.

6.4.1 BRANCHING NETWORKS

6.4.1.1 Single-Source Networks

In a single-source branching network, the source is labelled 0 (zero) and

the demand nodes are labelled j,j= 1 . . . J. Since each demand node
has only one upstream pipe, a pipe is labelled by its downstream node
label.

6.4.1.2 Multiple Source Networks

Let a multiple source branching network have M source nodes and N de-
mand nodes. It is preferable to label the source nodes initially from 1 to M
and label the demand nodes later from M + I to M + N (= J). The
pipes are labelled in any convenient order.

6.4.2 LOOPED NETWORKS

Even though the procedure suggested by Bhave [85,86] can be adopted

for labelling the various elements of looped networks, it would be suffi-
cient to label the source nodes from 1 to M and demand nodes from
M + \ to M + N (= J). The loops or circuits are labelled by Roman
numerals I, II, . . . C (Capital alphabet and not Roman numeral denoting
hundred), the total number of circuits or loops in the network.

6.5 PARAMETER INTERRELATIONSHIPS

Consider a looped network shown in Figure 6.5. Let the network have
M source nodes, labelled 1 . . . M, and N demand nodes, labelled
M + 1, . . . , M + N ( = / ) , so that the node labels are j , j = 1 . . . J
126 TYPES AND PARAMETERS

Source node, Demand node,

' "2 M+1 M+2 '"
I II

D ' D

Figure 6.5 Labelled looped network.

( = M + N). Let the pipe labels be x, x = 1 . . . X; and the loop labels

be c, c = I, II, . . . C. According to graph theory [87],

X = J + C - 1 (6.1)

which is also valid for branching networks for which C = 0.

The several parameter relationships available for network analysis are as
follows.

6.5.1 PIPE HEAD LOSS RELATIONSHIP

The head loss in a pipe can be expressed by a general head loss relation-
ship, Equation (2.37). Thus, for all pipes in the network,

h , = Ht - Hj = RXQ"X, x = 1 X (6.2)

in which hx = head loss in pipe x; //,-, //, = HGL values at the upstream
node / and downstream nodej of pipe x, respectively; Qx = discharge in
pipex; Rx = resistance constant of pipe x (Table 2.7); and n = an expo-
nent (Table 2.7). Since n * 1 in most cases, Equation (6.2) represents a
nonlinear relationship between hx and Qx.
The direction of flow in a pipe is assumed during the network analysis.
However, when the direction of flow changes, Qx becomes - Qx. There-
fore, to evaluate hx when the flow direction reverses and Qx becomes — Qx,
Equation (6.2) is written as

hx = H, - Hj = 'RX\ Qx (6.3)

Thus, | Qx |"-' is always positive while Qx can be positive or negative de-

pending upon the direction of flow in the pipe. When the flow in a pipe oc-
curs in the assumed direction, Qx, and therefore hx, are positive, and
 Parameter Interrelationships 111

Hi > Hj. When the flow direction reverses, Q,, and therefore hx, become
negative, and //, < //,.
Equation (6.3) is sometimes expressed as

K = RL Qx, x = 1 . . . X (6.4)

in which Rx = modified resistance constant of pipe x and is given by

RL = R,\ G x | - \ J C = 1 . . . X (6.5)

Equation (6.4) is the linearized form of the general head loss relation-
ship, expressing hx as a linear function of Q,. In Equations (6.2) and (6.3)
the resistance constant Rx has a fixed value for a pipe when the Hazen-
Williams or the Manning's head loss formulas are used. However, the
modified resistance constant Rx depends upon Q and therefore it must be
reevaluated when the value of Qx changes.
Equation (6.2) can also be expressed as

To take care of the change in the flow direction, Equation (6.6) can be ex-
pressed as

Qx = UuiJL _ x= i ,_X (6.7)

R¥"\ //,• - Hj P ' ' ~'

in which the denominator is positive while the numerator may be positive

or negative depending upon the direction of flow in the pipe.
Equation (6.7) can also be expressed as

Q, = C' (//, - Hj), x = 1 . . . X (6.8)

in which Cx = modified conductance of pipe x, and is given by

,x = 1. . .X (6.9)
R\'

Equation (6.8) is the linearized form of Equation (6.7), but Cx depends

upon Hi and //, values, and therefore must be reevaluated when Ht, Hj or
both change.
128 TYPES AND PARAMETERS

6.5.2 NODE FLOW CONTINUITY RELATIONSHIP

For steady incompressible flow in a network, the algebraic sum of the

inflows to and outflows from a node must be zero. Thus,

Q, + qj = 0,j=l...J (6.10)

in which qs = external flow, i.e., supply (inflow) or demand (outflow), at

nodey.
Equation (6.10) gives a set of J linear relationships in terms of the pipe
flows Qx. However, as any one of these relationships can be expressed in
terms of other relationships, Equation (6.10) gives a set of J — 1 indepen-
dent linear relationships in terms of Qx.
Using Equation (6.6), Equation (6.10) can also be expressed in terms of
Hi and //,, as

J
' R ) + qj = O,j= I . . .J (6.11)
i connected
toy through x

Using Equation (6.7) the node flow continuity relationship becomes

Hi Hj
~ , , , +qj = 0J=l...J (6.12)
/connected Dl/nl LJ U \
J
t o ; through j: ' !

Equation (6.10) which is linear in terms of Qx has become nonlinear in

terms of//, and //, in Equations (6.11) and (6.12). Therefore, using Equa-
tion (6.8), Equation (6.10) can also be expressed linearly in terms of//, and
Hi as

C (H - Hj) + qj = 0 , j = 1 . . . J (6.13)
/ connected
toy through x

6.5.3 LOOP HEAD LOSS RELATIONSHIP

For all the loops of a network, the algebraic sum of the head losses in
the pipes forming a loop must be zero. Thus,

2 K = 2 R*Qi = 0, c = I, II . . . C (6.14)
 Rules for Solvability of Pipe Networks 129

Using Equation (6.4) the linearized version of Equation (6.14) becomes

R
E ^ = E 'Q' = 0, c = I, II . . . C (6.15)

in which /?,' is given by Equation (6.5).

6.6 RULES FOR SOLVABILITY OF PIPE NETWORKS

The unknown parameters in a pipe network can be nodal heads, the

nodal flows, and the pipe resistances. For network analysis to be feasible,
the number of unknown parameters and their distribution over the entire
network must be such that an adequate number of independent equations
can be formulated. If the unknown parameters are concentrated in one part
of the network and the known parameters in the other, it is not possible to
determine all the unknown parameters. Therefore, the number and distri-
bution of the unknown parameters should obey certain rules so that the
network analysis is feasible. Shamir and Howard [88-90], Gofman and
Rodeh [91], and Bhave [92] have considered the problem of solvability of
pipe networks.

6.6.1 RULES PROPOSED BY SHAMIR AND HOWARD

Shamir and Howard were perhaps the first to consider the problem of
solvability of pipe networks when all three types of unknowns, nodal
heads, nodal flows, and pipe characteristics, are present. They initially
suggested certain rules [88] and later updated them [89-90]. Their rules
are as follows:
(1) The total number of the unknowns must be equal to the total number
of the nodes in the network.
(2) At least one nodal head must be known.
(3) A node having an unknown consumption must be connected to at
least one other node with a known consumption.
(4) A subsystem consisting of a pipe with unknown resistance and its
two end nodes, must have no more than one additional unknown—
one of the heads or consumptions at the two end nodes.
(5) Considering any node, at least one of the following must be un-
known: the consumption at the node, the head at the node itself or at
any adjacent node, or the resistance of a pipe connected at the node.
These rules were developed heuristically by considering different com-
binations of unknowns, and no rigorous mathematical proof was provided
for their validity.
DO TYPES AND PARAMETERS

6.6.2 RULES PROPOSED BY GOFMAN AND RODEH

Using graph theory, Gofman and Rodeh [91] also developed certain rules
for solvability of networks having pipes with unknown characteristics.
(They termed such pipes as head generators.) However, in developing their
rules they assumed that a pipe network "virtually always has a unique solu-
tion in the absence of head generators." Therefore, their rules are not gen-
eral as they do not deal with the number and distribution of the unknown
nodal heads and unknown nodal flows. Their rules are as follows:
(1) The number of nodes with known heads must exceed the number of
head generators by one.
(2) For a given network G, there exists a connected network G' such
that:
a. All head generators and nodes with known heads belong to G'
b. G' is loopless
c. A path in G' connecting two nodes of known nodal heads con-
tains at least one head generator

Gofman and Rodeh have rigorously proved the validity of these rules.
However, these rules are not general as stated earlier. If a network is
solvable in the absence of head generators, it would still remain solvable
when the head generators are introduced such that they satisfy these rules.

6.6.3 RULES PROPOSED BY BHAVE

Bhave [92] developed the following rules:

(1) Unknown-Number Rule—The total number of the unknowns must be
equal to the total number of the nodes in the network.
(2) Minimum-Number Rule—At least one nodal head must be known
and one nodal flow unknown.
(3) Assignment Rule—Considering any node, at least one of the follow-
ing must be unknown: the flow at the node, the head at the node itself
or at any adjacent node, or the resistance of a pipe connected to the
node. To satisfy this, two of more nodes must not use the same un-
known. Every node of the network must have an assigned unknown,
and no unknown must be assigned to more than one node.
(4) Path Rule—Each pipe with an unknown resistance must lie on an in-
dependent path connecting two nodes of known heads, and each such
path must not contain more than one pipe with an unknown resis-
tance. All these paths must form branching configurations without
any loops.
 Rules for Solvability of Pipe Networks 131

The unknown-number rule guarantees that the number of the available

equations is equal to the number of the unknown parameters in the net-
work.
The minimum-number rule guarantees, at least, (1) one known nodal
head and (2) one unknown nodal flow. The known nodal head serves as the
reference node, relative to which other nodal heads are determined. When
the total specified inflow to a network is not equal to the total specified
outflow, the specified nodal flows may become inconsistent. The unknown
nodal flow prevents this possibility. Because one node flow continuity
equation is dependent, even with consistent specific nodal flows, the total
number of available independent equations would be less than the total
number of nodes in the network by one. This condition would not give a
unique solution for the network.
The assignment rule and the path rule together guarantee that the un-
known parameters are properly distributed over the entire network. It
should be noted that when the assignment rule is satisfied, the unknown-
number rule is automatically satisfied. However, the unknown-number
rule is a fundamental rule and is therefore retained.
The verification of the unknown-number and minimum-number rules is
easy. To verify the assignment rule, it must be noted that an unknown
nodal flow can be assigned only to the concerned node; an unknown pipe
resistance assigned only to either of the two end nodes; while an unknown
nodal head assigned only to either the concerned node or to any one of the
connected nodes. To satisfy the path rule, the number of nodes with known
heads must exceed the number of pipes with unknown resistance by at least
one.
It must be noted that these rules are also developed heuristically and no
rigorous mathematical proof is provided for their validity and sufficiency
[92]. Therefore, even though these rules are observed to be sufficient, they
might also be improved upon in future.

6.6.4 COMPARISON OF RULES

The unknown number rule is the same as rule 1 of Shamir and Howard.
The minimum-number rule is similar to rule 2 of Shamir and Howard but
is more comprehensive because it also avoids the possibility of the
specified nodal flows being inconsistent. The assignment rule is similar to
rule 5 of Shamir and Howard but is more comprehensive and in combina-
tion with the path rule guarantees that all the available equations are inde-
pendent. Rules 3 and 4 of Shamir and Howard are not essential as shown
in Example 6.1. The path rule has no similar rule in the set of rules pro-
posed by Shamir and Howard, but is similar to the rules proposed by Gof-
man and Rodeh.
132 TYPES AND PARAMETERS

Example 6.1

Check whether the network shown in Figure: Example 6.1 is uniquely

solvable if the unknown parameters are as follows:
(1) //,, q,, H2, R2, H3, q3, R7, q7, H8, q9, and q10
(2) //,, qu H2, R2, H3, q3, R7, q7, Hs and q9
(3) //,, q2, q5, R6, q6, H7, R7, H8, qg, and /? I 2
(4) q,, H2, 7?2, H3, q3, R7, q?, Hs, q9, and </,<,

Node,1
Pipe, 2

5 8
10

13 9 l4
10
8
Solution

1. The network has 10 nodes, while the number of unknowns is 11. As

the unknown-number rule is violated, the network is unsolvable.
2. The 10-node network has 10 unknowns and therefore the unknown-
number rule is satisfied. The head at node 4 (and also at some other nodes)
is known, and the flow at node 1 (and also at some other nodes) is un-
known, and therefore, the minimum-number rule is also satisfied. The
assignment rule is checked as shown in Table: Example 6.1. As seen from
the table, R7 has remained unassigned while node 10 has no parameter
assigned to it. As the assignment rule is not satisfied, the network is not
solvable.
3. The unknown-number and the minimum-number rules are satisfied.
The assignment rule is also satisfied as shown in Table: Example 6.1 [The
assignments of R6, H7, and Rl2 are shown in the table as: R6 to node 3, H7
to 7, and Rl2 to 10. However, two other assignments are also possible: (1)
R6 to 3, H7 to 10, and R{2 to 7 and (2) R6 to 7, H7 to 3, and R12 to 10.J For
the path rule, the unknown pipe resistances are R6, R7, and /?, 2 , i.e., three
in number, while the known nodal heads are: H2, H3, H4, Hs, H6, H9, and
Hio, i.e., seven in number—much more than four, the minimum required
number. Further, R6 lies on path 3-7-10 (H3 and //,<> known), R7 on path 4-5
(H4 and Hs known), and Ri2 on path 3-7-10 (H3 and Hw known). However,
R6 and Rl2 lie on the same path (path 3-7-10) and therefore the path rule is
violated. Thus, the network is not uniquely solvable.
 TABLE: EXAMPLE 6.1.

Checking Assignment Rule for Network of Figure: Example 6.1

Unknown Available Available Available

parameter Unknown nodes for Assigned nodes for Assigned nodes for
number parameter assignment to node assignment to node assignment
(1) (2) (3) (4) (5) (6) (7)

Case 2
1 Hi 1,2,4 2,4 4
2 Qi 1 1 — — —
3 H2 1,2,3,5 — 2,5 — 5
4 Ft? 2,3 — 2 2
5 H3 2,3,6,7 — 2,6. — 6
6 93 3 3 — — —
7 R7 4.5 — 4,5 — 4,5
8 9/ 7 7 — — —
9 H8 4,8,9 _. 4,8 4.8
10 99 9 9 — — -
Case 3
1 H, 1,2,4 1,4 1
2 a? 2 2 — — —
3 9s 5 5 — — —
4 Re 3,7 — 3,7 — 3,7
5 96 6 6 — — —
6 H, 3,7,10 — 3,7,10 — 3,7,10
7 fi7 4,5 — 4 4 —
8 H8 4,8,9 4.8 8
133

(continued!
 TABLE: EXAMPLE 6.1. (continued).

Checking Assignment Rule for Network of Figure: Example 6.1

Unknown Available Available Available

parameter Unknown nodes for Assigned nodes for Assigned nodes for
number parameter assignment to node assignment to node assignment
(1) (2) (3) (4) (5) (6) (7)

9 9 9 _ _ _
10 7,10 — 7,10 — 7,10
Case 4
1 Qi 1 1 — — —
2 H2 1,2,3,5 — 2,5 — 5
3 R2 2,3 — 2 2 —
4 H3 2,3,6,7 — 2,6 — 6
' 5 Q3 3 3 — — —
6 fly 4,5 — 4,5 — 4,5
7 9/ 7 7 — — —
8 H8 4,8,9 — 4,8 — 4,8
9 Q9 9 9 — — —
10 Q10 10 10 — — —
 TABLE: EXAMPLE 6.1. (continued).

Checking Assignment Rule for Network of Figure: Example 6.1

Available Available Final

Assigned nodes for Assigned nodes for Assigned assignment
to node assignment to node assignment to node to node
(8) 0) (10) (11) (12) (13)

! I I I I I I I

I I I I I I I I
Case 2

4 4
1
5 5
2
6 6
3
nil nil nil
7

Case 3

1 - - - 1
2
_ _ - - - 5
- 3,7 3 - - 3
- — — — — 6
- 3,7,10 7 — — 7
_ _ — — - 4
8 8
c - - - -
in (continued)
 TABLE: EXAMPLE 6.1. (continued).

Checking Assignment Rule for Network of Figure: Example 6.1

Available Available Final

Assigned nodes for Assigned nodes for Assigned assignment
to node assignment to node assignment to node to node
(8) (9) (10) (11) (12) (13)

' 9
— 7,10 10 — — 10

Case 4

1
5 I 5
— — — — 2
— — —
—
6
— — —
— 6
3
• —

4 4 —
— 4
—

— — —
_
—

4,8 — 8
_8 7
8
— 9
— — — — — 10
 Problems 137

4. The unknown-number rule and the minimum-number rule are

satisfied. The assignment rule is also uniquely satisfied as shown in Table:
Example 6.1. Pipe 7 with unknown resistance R7 lies on path 4-5 (H4 and
Hs known), and pipe 2 with unknown resistance R2 lies on path 1-2-3-6 (//,
and H6 known). Thus the path rule is satisfied. [Note that alternate path 5-
2-3-6 (Hs and H6 known) is also available for R2.] As all the rules are
satisfied, the network is uniquely solvable.
Note that node 7 having unknown flow qn is connected to nodes 3 and
10, also having unknown flows q3 and qw, respectively. Thus, node 7 with
unknown flow is not connected to any node of known flow, thereby
violating rule 3 of Shamir and Howard. Further, consider the subsystem of
pipe 2 with unknown resistance R2 and the two end nodes 2 and 3. For this
subsystem, H2, H3, and q3 are also unknown in addition to R2. According
to rule 4 of Shamir and Howard, only one from q2, q3, H2, and H3 is per-
missible as unknown in addition to R2. Thus, rule 4 of Shamir and Howard
is also violated. However, the network is uniquely solvable [92J as can be
seen later.
As demonstrated by this example, the assignment and path rules may be
satisfied in more than one way (assignment rule for Case 3 and path rule
for Case 4). However, when all the rules are satisfied, either uniquely or
in more than one way, the network is uniquely solvable.

6.7 PROBLEMS

In the following problems verify whether the network-solvability rules

are satisfied for the given combinations of unknowns.
Problem 6.1: In the network of Figure: Problem 6.1, the unknowns are
as follows:
(1) qu Hu R3, R4, Hs, H6, q6
(2) (ji, Hi, R2, R3, Hs, H6
(3) qu Hi, R3, qA, qs, H6
(4) qu H3, q4, H6, R7, R6
(5) qt, Hi, R3, R4, Hs, Hb
(6) Hi, q2, Ru Hs, H6, q6
2
138 TYPES AND PARAMETERS

Problem 6.2: In the network of Figure: Problem 6.2, the unknowns are
as follows:
(1) qu Hu R u R2, H2, q3, R12
(2) Hi, H2, R2, q4, Hs, H7, R9
(3) q3, H3, R3, q6, H6, q7, H7
(4) q2, q4, H4, q7, H7, R9, Ri2
(5) q2, H2, q4, H4, q7, H7, R n

Problem 6.3: In the network of Figure: Example 6.1, the unknowns are
as follows:
(1) <7i> <?2, H2, R4, Hs, R6, R7, qs, H9, q10, Ri3
(2) qu q2, H2, R3, R4, Hs, H6, q7, qs, H9
(3) qu Hu q3, qs, Hs, q6, H7, q9, q10, Hl0
(4) qu Hi, q3, qs, H5, q6, H7, Hs, q9, qi0
(5) q2, H2, H4, R4, R6, H7, qs, H9, q10, Rl2
(6) Hi, qu R2, q4, Hs, q6, Hs, R9, qw, R u
(7) Hi, qx, H4, q4, H7, qs, H9, H10, Ri3, Ri4
(8) Hi, q2, q4, R s , Hs, Hs, H9, qi0, Hw, Ri4
(9) q2, H2, H4, R s , q6, H6, q9, qi0, R u , Ri3
 Formulation of Equations

7.1 INTRODUCTION

A pipe network is treated as analyzed when all the unknown parameters

-Tvare determined with the help of the known parameters and the
available interrelating equations. The pipe discharges are mostly un-
known and obtained from the analysis. (When a pipe discharge is known,
this pipe is removed from the network during analysis, and the inflow and
outflow equal to this known pipe discharge are provided at the down-
stream and upstream nodes, respectively, while the flows at the end nodes
of the pipe are modified accordingly.) Of the remaining three parameters,
i.e., nodal heads H, nodal flows q, and the pipe resistance constants R,
some may be known and some unknown. However, the number and dis-
tribution of these unknown parameters should satisfy the network solv-
ability rules described in Chapter 6.
A node with a known head is usually a source node such as an elevated
service reservoir or a pump with a fixed head. Similarly, a node with a
known flow is a demand node. Further, at the source node, the flow is
generally unknown while at the demand node the head is generally un-
known. However, occasionally, the flow also may be known in addition
to the known head at a source node, while the flow may also be unknown
in addition to the unknown head at a demand node. Therefore, a node with
a known head (flow unknown or known) will be treated herein as a source
node; while a node with an unknown head (flow known or unknown) will
be treated as a demand node. In the figures, a source node is shown by a
square n while a demand node by a bullet • .
Different combinations of known and unknown parameters may occur
in network analysis. Considering the network solvability rules, the fol-
lowing seven combinations can occur:

(1) All R, all (except one) q, and one H values known

(2) All R, some q and some H values known
139
140 FORMULATION OF EQUATIONS

(3) Some R, some q and some H values known

(4) All H, all (except one) q values known
(5) All H, and all R values known
(6) All H, some q and some R values known
(7) Some R, some H and all (except one) q values known
The first combination in which all /?, all (except one) q, and one H
values are known is common in practice. It is obtained for a single-source
network, in which the resistance constants of all pipes, the demands at all
demand nodes, and the head at the source node are known. This combina-
tion therefore will be referred to as single-source networks with known
pipe resistances.
The second combination in which all R, some q and some H values are
known occurs in multisource networks. The resistance constants of all
pipes, the demands at the demand nodes, and the heads at the source nodes
are known. This combination, therefore, will be referred to as multisource
networks with known pipe resistances.
The third combination in which some R, some q and some H values are
known is a general combination and occurs for single- and multiple-source
networks having all three types of parameters as unknowns. This combina-
tion occurs when some pipes are considered having unknown resistances,
and their sizes are determined to achieve a desired flow distribution pattern
in the network. This combination will be referred to as networks with un-
known pipe resistances.
The fourth combination in which all H and all (except one) q values are
known has all R values as unknowns. Such a combination does not occur
in the analysis but occurs in the design of a new network. However, the
nodal heads are treated in the form of inequalities, i.e., the available heads
should not be less than some specified values. A large number of possible
solutions would be available and attempts are made in practice to obtain
minimum-cost solutions. Naturally, this combination falls outside the
scope of this text.
The fifth combination in which all H and all ^? values are known has little
practical significance, as rarely, if ever, the heads at all nodes are known.
The nodes with the known heads are the source nodes, mostly reservoirs
in practice, and therefore this combination would mean a network having
all reservoirs directly connected without any intermediate junction nodes.
However, this combination is simple to analyze as each pipe can be con-
sidered separately, and pipe flows and subsequently nodal flows can be de-
termined. Therefore, this combination will not be considered.
The sixth combination in which all H, some q and some R values are
known would be partly similar to combination 5 and partly similar to com-
bination 6, and has no practical significance and therefore will not be con-
sidered.
 States of Parameters 141

The seventh combination in which some R, some H and all (except one)
q values are known can be looked upon as a particular case of the general
combination 3 and therefore would not be considered separately.
Thus, combinations 1, 2, and 3 are important combinations in the analy-
sis of pipe networks. Even though combinations 1 and 2 are particular
cases of the general combination 3, they are considered separately because
of their common occurrence in practice. This also makes it easier to de-
velop the subject and understand the formulation of the equations and the
methods of their solution. The formulation of equations, therefore, is con-
sidered for combinations 1, 2, and 3, in that order, and is also illustrated
through typical networks. Initially branching networks and subsequently
looped networks are considered for each combination.

7.2 STATES OF PARAMETERS

Four states of parameters are used in the formulation of equations.

These states are as follows:
(1) A parameter that is known and remains fixed throughout the analy-
sis. Naturally, the numerical values of these parameters will be used
in practice, but in general formulation they are indicated by a suffix-
ing subscript o. For example, Rox indicates the known resistance con-
stant of pipe x.
(2) A parameter that is unknown but the value of which is assumed for
the first iteration and updated for subsequent iterations. The assumed
value will be used in numerical problems but in general formulation
this parameter is indicated by a prefixing subscript t in which t = 1,
2, . . . etc., indicates the iteration number. For example, ,QX in-
dicates the assumed discharge in pipe x for the fh iteration; SH6 in-
dicates the assumed head at node 6 for iteration 5.
(3) A parameter that is modified to include part of a nonlinear term for
linearizing a nonlinear equation as in Equation (6.4) and Equation
(6.8). This state is indicated by a prime. For example, Rx indicates
the modified resistance of pipe x. As this would change from itera-
tion to iteration, prefixing subscripts can be used to specify the itera-
tion number. Therefore, 5/?9' would indicate the modified resistance
of pipe 9 for iteration 5.
(4) A parameter that is unknown and is used as an unknown parameter
in formulating the equations. This state is indicated without any ad-
ditional mark. For example, Qx indicates the unknown discharge in
pipe x.
In general, the equations are formulated for the first iteration and there-
fore, parameters with assumed values will be indicated by a prefixing sub-
142 FORMULATION OF EQUATIONS

script 1. However, this will be omitted for simplicity where ambiguity

does not arise.

7.3 BASIC UNKNOWN PARAMETERS

As stated earlier, the unknown parameters may be the pipe discharges Q,

the nodal heads H, the pipe resistances R, and the nodal flows q. However,
some of these parameters are treated as the basic unknown parameters,
and their values are determined initially. Once the values of the basic un-
known parameters are known, the values of the other unknown parameters
are easily determined. Consider, for example, the second combination of
multisource networks with known pipe resistances. For this combination,
if all the pipe discharges are determined, the unknown nodal flows and
heads can be determined without any difficulty.
The equations are formulated using the basic unknown parameters, and
the formulated equations are designated by these unknown parameters.
Thus, when the unknown pipe discharges are taken as the basic unknown
parameters, the equations formed are known as Q equations. Similarly, the
unknown nodal heads, unknown nodal flows, and unknown pipe resis-
tances are the basic unknown parameters in H-q-R equations [93-98].

7.4 SINGLE-SOURCE NETWORKS WITH KNOWN PIPE RESISTANCES

7.4.1 BRANCHING NETWORKS

As the flows at all the demand nodes are known, starting from the end
nodes and successively applying the node-flow continuity relationship
[Equation (6.10)], the flows and head losses in all the pipes and the supply
at the source node are determined. As the head at the source node is
known, moving along the paths and successively subtracting the head
losses in pipes, the heads at all demand nodes are determined. Thus, it is
quite simple to analyze such a network.

7.4.2 LOOPED NETWORKS

The networks are single-source looped networks as shown in Figure 7.1.

The source node is labelled 1 and the demand nodes are labelledy',y = 2,
3, . . . , J. The pipes are labelled x, x = 1, 2, . . . , X, and the loops are
labelled c, c = I, II, . . . , C. Herein, the pipe resistances are known,
i.e., they are Rox, x = 1 . . . X; the demand-node flows are known, i.e.,
they are qoj, j = 2 . . . J; and the head at the source node is known and
 Single-Source Networks with Known Pipe Resistances 143

Sourcenode, Demand node,

3
Pipe,1 2

Loop, I II

Figure 7.1 Labelled single-source looped network.

is Hoi. The unknown parameters are Qx, x = I . . . X; Hj, j = 2 . . . J

and qx.

7.4.2.1 0 Equations

The unknown pipe flows, Qx, x = 1 . . . X are taken as the basic un-
known parameters in formulating the Q equations.
General Formulation. Applying the node-flow continuity relationship
[Equation (6.10)] for the demand nodes

= 2 . . .J (7.1)
x connected
lo j

which are J — 1 linear equations. Similarly, applying the loop-head loss

relationship [Equation (6.14)] for the basic loops,

2 Ro = 0, c = I, II, . . . C (7.2)
x6,

which are C nonlinear equations. Thus, the total number of Q equations is

J — 1 + C, equal to X, the number of basic Qx unknown parameters
[Equation (6.1)].
Illustration. To illustrate the formulation of the g-equations, consider
the network shown in Figure 7.2. The network has one source node,
labelled 1; five demand nodes, labelled 2 . . . 6; seven pipes, labelled
1 . . . 7; and two basic loops, labelled I and II. The known R values
(/?oi . . . Ro7), q values (qo2 . . . ^o6) and //value (//„,) areas shown in the
144 FORMULATION OF EQUATIONS

Demand,
HGL, 0-4
1 Pipe, 2
100-00
Q2 R=20 Q5 40 5

SI
a r
a
Loop, I II

2 3 3 6
Q3 1Q Q6 \
0-3 0-1
0-1
Figure 7.2 One source, five-demand-node looped network.

figure. The directions of the unknown pipe flows are assumed as shown in
the figure.
Considering the inflow to a node as positive and outflow from it as
negative, the node-flow continuity relationships for nodes 2 . . . 6 are, re-
spectively,

Q1-Q3- 0.3 = 0 (7.3a)

0 - 0 - Q6 + 0.1 =0 (7.3b)

Q2 + Q4- Q5- 0.2 = 0 (7.3c)

Qs + Qi - 0.4 = 0 (7.3d)

Q6 - Qi - 0.1 = 0 (7.3e)

Moving in the clockwise direction along the two basic loops I and II, the
loop-head loss relationships yield, respectively

F6 = 2002 - 300; - \0Q"3 - 40(27 = 0 (7.3f)

F7 = 4 0 0 - 5 0 0 - 2 0 0 + 3 0 0 = 0 (7.3g)

Equations (7.3a) . . . (7.3g) provide seven independent Q equations

(five linear and two nonlinear) for seven basic unknowns Ql . . . 0 . Us-
 Single-Source Networks with Known Pipe Resistances 145

ing the linearized form. Equation (6.15), the loop-head loss relationships
are

F8 = R2Q2 - RtQ* - R3Q3 - fii'C?! = 0 (7.3h)

F9 = RsQs - R7Q7 - ReQ* - RAQ4 = 0 (7.3i)

Equations (7.3a) . . . (7.3e) and (7.5h), (7.5i) provide seven indepen-

dent Q equations (five linear and two linearized) for the seven basic un-
knowns Qi . . . Q7.

7.4.2.2 H Equations

The unknown nodal heads, //,, j = 2 . . . J are taken as the basic un-
known parameters in formulating H equations.
General Formulation. By applying the node-flow continuity relation-
ship, Equation (6.11), to nodes 2 . . . J, the H equations are

2 ( ^ T ^ r +1« = O,J = 2...J

ox
(7.4a)
/connected \ I
to j through A

in which //,, //, = heads at upstream node and downstream node, respec-
tively.
The linearized version of the H equation is

C ( / / , - Hj) + qOJ, = OJ = 2 . . . J (7.4b)

/ connected
toy through x

in which C,' = modified conductance of pipe x and is given by

C = ,x = 1... X (7.4c)
( l
*»» H H " ^
Kox tli — tlj

Illustration. For the network of Figure 7.2, the //equations are

"-0.3=0 (7.5a)
146 FORMULATION OF EQUATIONS

(H4 H, , . , „ . „„ , _ Q4 = 0 (? 5d)
40 / \ 50

"' ' ' " - O . I = 0 (7.5e)

Equations (7.5a) . . . (7.5e) provide five nonlinear equations for the five
basic unknowns H2 . . . H6.
The linearized version is
F6 = C,'(100 - H2) - Q(H2 - H3) - 0.3 = 0 (7.5f)

F7 = Ci(H2 - H3) - Ci(H3 - H4) - Q(H3 - / / . ) + 0.1 = 0 (7.5g)

F8 = C2'(100 - HA) + C4'(//3 - H4) - Q(H4 ~ H5) - 0.2 = 0 (7.5h)

F9 = Q(H4 - Hs) + Ci(Ht - Hs) - 0.4 = 0 (7.5i)

Fw = Ci(H3 - H6) - C7'(//« - H5) - 0.1 = 0 (7.5j)

Equations (7.5f) . . . (7.5j) provide five linearized equations for the five
basic unknowns H2 . . . H6.

7.4.2.3 AO Equations

In formulating the AQ equations, the discharges in all pipes are initially

assumed such that they satisfy the node-flow continuity relationship
[Equation (7.1)]. These assumed pipe flows will not generally satisfy the
loop-head loss relationship [Equation (7.2)]. Therefore, the pipe flows are
corrected by applying loop-flow corrections AQC, c = I, II, . . . C.
These corrections are treated as the basic variables in formulating the AQ
equations.
General Formulation. From the loop-head loss relationship of Equation
(7.2), the AQ equations for the first iteration are
R
°* ('2, + £ .A&)" = 0, c = I, II, . . . C (7.6)
 Single-Source Networks with Known Pipe Resistances 147

in which E c represents the summation for all pipes of a loop c, while L,

represents the summation of the corrections of all loops to which pipe x
belongs.
Illustration. For the network of Figure 7.2 let the assumed pipe flows be:
,£>, = 0.4 nrVs; ,Q2 = 0.5 nrVs; , g 3 = 0.1 nrVs; ,(?4 = 0.05 m3/s;
jgs = 0.35 mVs; & = 0.15 m 3 /s; and ,<27 = 0.05 m'/s in the directions
as shown in the figure. Note that these assumed values satisfy the node-
flow continuity relationships at nodes 2, . . ., 6.
As the network has two basic loops, the basic unknown parameters are
AC?, and AQ» for loops I and n, respectively. (The prefixing subscript 1
is omitted for simplicity.) Taking the clockwise direction as positive for
these corrections, the AQ equations for loops I and II are, respectively

F, = 20(0.5 + AC?,)" - 30(0.05 - AQ, + AQ,,)"

- 10(0.1 - AQ,)" - 40(0.4 - AQ,)" = 0 (7.7a)

F 2 = 40(0.35 + AQ,,)" - 50(0.05 - AQ,,)"

- 20(0.15 - AC?,,)" + 30(0.05 - AC?, + AC?,i)" = 0 (7.7b)

Note that as pipe 4 is common to both loops, the corrections AC?, and
AC?n are applied to the assumed discharge 0.05 m3/s in pipe 4.
Equations (7.7a) and (7.7b) provide two nonlinear equations for the two
basic AC? parameters, AC?, and AC?,i-

7.4.2.4 AH Equations

In formulating the AH equations, the unknown nodal heads //,,

j = 2 . . . J are assumed. As these assumed values will not generally sat-
isfy the node-flow continuity relationship, Equation (7.4a), the assumed
iHj, j = 2 . . . J values are corrected by applying additive corrections
1AHJ,j = 2. . .J.
General Formulation. From Equation (7.4a), the AH equations are:

ox
innected toL
i connected
j through x \ I. o)

Illustration. For the network of Figure 7.2, Hoi = 100 m, a given value.
Let the assumed Mj values be: ^H?. — 97 m, ,// 3 = 95 m, {H4 = 94 m,
iHs = 90 m, and ,// 6 = 92 m. The basic unknown parameters are
148 FORMULATION OF EQUATIONS

AH2 . . . AH6. (The prefixing subscript 1 is omitted for simplicity.) The

AH equations are:

„ f 100 - (97 +AH2)]U" \(97+AH2) - ( 9 5 + AH3)V'"

r, =
40 J [ 10

-0.3=0 (7.9a)

[(97 + AH2) - (95 + A//3)l "" [(95 + AH3) - (94

[ 10 | | 30

[ ( 9 S + A f t ) 2 O ( 9 2 + A g - ) ]'" + ' ' " o a»)

[ 100 - (94 + A//4)l "" [(95 + AH3) - (94 + A// 4 )l""
[ 20 J [ 30

[(94+AH,) -(90+AHs)]""
- 0.2 = 0 (7.9c)
40

[(94 + Atf4) - (90 + AH,)] "" [(92 + A//6) - (90 + A//5)] "-
40 50
L J L J
- 0.4 = 0 (7.9d)

[(95 + AH3) - (92 + A//6)l u" [(92 + A//.) - (90 + A///55; ) l " "
50 J
- 0.1 = 0 (7.9e)

Equations (7.9a) . . . (7.9e) provide five independent nonlinear equa-

tions for the five basic unknown parameters A//2 . . . AH6.

7.5 MULTISOURCE NETWORKS WITH KNOWN PIPE RESISTANCES

7.5.1 BRANCHING NETWORKS

Consider a multisource branching network shown in Figure 7.3(a). It

has three source nodes labelled 1 . . . 3, twelve demand nodes labelled
4 . . . 15, and fourteen pipes labelled 1 . . . 14 as shown in the figure. R e -
 Multisource Networks with Known Pipe Resistances 149

. \\ ?
9 2 11
2 6
4 5
D J
1
3I
10
I
5
3 12
13
12
-I
15 +<
iou +< 1oi5
(a) (b)
Figure 7.3 Multisource branching network: (a) complete network, (b) reduced network.

sistance constants of all pipes are known and they are / ? , ! . . . Rol4 for
pipes 1 . . . 14, respectively. The heads at the source nodes are known and
are Hol . . . Ho3 for source nodes 1 . . . 3, respectively. The demands at
the demand nodes are known and are qo4 . . . qol5, for nodes 4 . . . 15, re-
spectively.
The discharges in pipes 2, 3, 5, 7, 8, 9, 12, 13, and 14 can be easily deter-
mined. These pipes of known discharges are removed from the network,
and the outflows at the intermediate nodes 2, 4, 7, and 12 are correspond-
ingly modified as shown in Figure 7.3b. As q2 and q3 are unknown, the dis-
charges in the pipes of the reduced network are unknown. However, these
discharges must be such that they satisfy the node-flow continuity relation-
ship and completely utilize the available head loss along the paths. Thus,
the head loss along the path 1-4-7-2 must equal Hol — HB2 and that along
path 1-4-7-12-3 must equal HoX — Ho3. Therefore, pseudo pipes connecting
nodes 1 and 2, and 1 and 3 are introduced as shown by dashed lines in
Figure 7.3(b). Thus, two pseudo loops, I and II, are formed as shown in
the figure.
As shown herein, a multisource branching network having M source
nodes can be converted to a looped network having M — \ pseudo loops.
Therefore, the formulation of equations would be similar to that of looped
networks.

7.5.2 LOOPED NETWORKS

Consider a multisource looped network shown in Figure 7.4. The net-

work has M source nodes labelled 1 . . . M, N demand nodes labelled
M + 1 . . . M + N(= J), Xpipes labelled 1 . . . X, and Cbasic loops
150 FORMULATION OF EQUATIONS

C+1
M-1
M+1
C+M-1

M+N
N. 1

Figure 7.4 Labelled multisource looped network.

or circuits labelled / . . . C. As the network has M source nodes, intro-

duce M — 1 pseudo loops labelled C + 1 . . . C + M — 1, as shown in
the figure.

7.5.2.1 0 Equations

General Formulation. From the node-flow continuity relationships at the

demand nodes,

E Q* + q* = 0,j = M+l...M + N (7.10)

x connected
10 j

which are N independent linear equations. Applying the loop-head loss

relationship to the real as wellas the pseudo loops,

=0, c = l...C, C + 1 . . . C + M - 1 (7.11)

which are C + M — 1 independent nonlinear equations. Thus, Equa-

tions (7.10) and (7.11) together give N + C + M - 1, i.e., 7 + C - 1
independent equations for the X basic unknowns, Qx, x = 1 . . . X. As
X = J + C — 1 as per Equation (6.1), X independent equations are
available for X basic unknowns Qx, x = 1 . . . X.
Illustration. Consider the network shown in Figure 7.5. The network has
two source nodes labelled 1 and 2 ; four demand nodes labelled 3 . . . 6;
seven pipes labelled 1 . . . 7; two basic loops labelled I and II; and one
pseudo loop labelled HI. The known R values (Rol . . . Ro7), q values
 Multisource Networks with Known Pipe Resistances 151

(<Z<>3 • • • (lob) and //values (Hot, Ho2) are as shown in the figure. The direc-
tions of the unknown pipe flows are assumed as shown in the figure.
From the node-flow continuity relationship for nodes 3 . . . 6, respec-
tively,
F, = 0 - 0 - 0.3 = 0 (7.12a)

Fi = 0 + 0 - 0 - 0.2 = 0 (7.12b)

F3 = 0 + 0 - 0.4 = 0 (7.12c)

FA = 0 - 0 - 0.1 = 0 (7.12d)

From the two basic loops I and II, respectively,

F5 = 20(22 - 300; - IOQ"3 - 40Ql = 0 (7.12e)

F6 = 40(2? - 5 0 0 - 2 0 0 + 3 0 0 = 0 (7.12f)

and from the pseudo loop III,

F7 = 4 0 0 + 100 + (95 - 100) = 0 (7.12g)

Equations (7.12a) . . . (7.12g) provide seven independent Q equations

(four linear and three nonlinear) for seven basic unknowns 0 . . . 0 .
Note that Equations (7.12a) . . . (7.12f) are the same as Equations

Demand,0-2 m3/s
q. Head,
\1OO-OO

Q3 ]Q/'i*o6 Q
in q i

Figure 7.5 Two-source, four-demand-node looped network.

152 FORMULATION OF EQUATIONS

(7.5a), and (7.5c) . . . (7.5g), respectively. As one nodal head is known in-
stead of one nodal flow, Equation (7.5b) is replaced by Equation (7.12g).
In general, therefore, for each introduction of a known nodal head in place
of a known nodal flow, one nonlinear loop-head loss equation replaces a
linear node-flow continuity equation.
Using the linearized form, Equations (7.12e) . . . (7.12g) become re-
spectively,

Fs = RiQ2 - R:Q4 - RiQi - R[Q, = 0 (7.12h)

F9 = R^Qs - R;Q7 - RIQ6 + RZQ4 = 0 (7.12i)

F10 = K.'G, + RZQ, + (95 - 100) = 0 (7.12J)

Equations (7.12a) . . . (7.12d), and (7.12h) . . . (7.12J) provide seven in-

dependent Q equations (four linear and three linearized) for the seven
basic unknowns Qi • • • Q7.

7.5.2.2 H and H-q Equations

The unknown nodal heads are taken as the basic unknown parameters in
formulating H equations. Sometimes, the unknown nodal flows are also
taken as the basic unknown parameters to form H-q equations. The pseudo
pipes and the pseudo loops are not necessary in formulating the H or H-q
equations.
General Formulation. By applying the node-flow continuity relationship
to the demand nodes labelled M + 1 to M + N, the H equations are

,o. •= 0,7" = M + 1 . . . M + N (7.13a)

/ connected to *
j through x

The linearized version of the H equation is

C; (Hi - Hj) + qOJ ^' 0J = M + 1. . .M + W (7.13b)

i connected to
j through x

in which Cx' = modified conductance of pipe x and is given by

C = ,x = 1... X (7.13c)
 Multisource Networks with Known Pipe Resistances 153

In Equations (7.13a) . . . (7.13c), //, = Hoi, and //, = Hoj for /,

j = 1 . . . M. Equations (7.13a) and (7.13b) are JV independent, nonlinear
and linearized equations, respectively, for the N basic unknown parame-
ters, Hj,j = M + 1 . . . M + N.
Sometimes, to complete the formulation, the node-flow continuity rela-
tionships are also applied to the source nodes to obtain the H-q equations.
Thus, we have

(7.14a)
/ connected I
j through x

the linearized version of which is

C {H> - Hj) + qj = 0,j = 1 • • . M (7.14b)

/ umnected \o
j through A

In Equations (7.14a) a n d (7.14b), //, = Ho, a n d Hj = Hoj for i,j =

1 . . . M.
Illustration. For the network shown in Figure 7.5, applying the n o d e -
flow continuity relationship to nodes 3 . . . 6, respectively,

'•" ( " I "( T V - 0-3 =0 «,,«

F. = \ - * - ^ \ + F ^ M -0.4 = 0 (7.15c)

-0.1=0 <7.15d)

Equations (7.15a) . . . (7.15d) provide four independent nonlinear equa-

tions for the four basic unknowns H3 . . . H6. Note that these equations
are the same as Equations (7.5a) and (7.5c) . . . (7.5e), except the change
in node labels. Since the head, instead of the flow, at one additional node
is known, one H equation is absent. For each additional //-known and
^-unknown value, the number of H equations decreases by one.
154 FORMULATION OF EQUATIONS

For the H-q equations, applying the node-flow continuity relationship to

nodes 1 and 2,

(
fj QC\ I In /QC t / \ 1/n / QC _ ZJ \ II n

n3 — yj\ I yj — n4\ lyj — ti6\

+
-io-J -[-nr) -(-»-) *"0(7.1S0
in which ^ and g2 are taken as inflows at 1 and 2, respectively. Equations
(7.15a) . . . (7.15f) provide the complete//-# formulation for the network.
The linearized H-q equations are
F, = CY(1OO - H3) - C3(H3 - 95) - 0.3 = 0 (7.16a)

F2 = C 2 '(100 - H4) + C4'(95 - H4) - C 5 '(// 4 - // 5 ) - 0.2 = 0 (7.16b)

F3 = C 5 '(// 4 - // 5 ) + C 7 '(// 6 - // 5 ) - 0.4 = 0 (7.16c)

F 4 = C6'(95 - H6) - Ci(Ht - Hs) - 0.1 = 0 (7.16d)

F5 = - C i ' ( 1 0 0 - H3) - C2'(100 - H4) + qt = 0 (7.16e)

F6 = C3'(H3 - 95) - C4'(95 - H4) - C6'(95 - H6) + q3 = 0 (7.16f)

7.5.2.3 AQ Equations

In formulating the AQ equations, the pseudo pipes and pseudo loops are
introduced as is done in formulating the Q equations. It is also necessary
to initialize the pipe flow, i.e., assume the pipe flows for the first iteration.
The basic unknown parameters are AQC, c = I, II, . . . C, C +
1 . . . C + M - 1.
General Formulation. From the loop-head loss relationship for all the
real and pseudo loops

S /? o ,(,a + £ IAGC)- = 0, c = I, I I , . . . C, C + 1, . . . C + M - 1
(7.17)

in which the summations are taken as explained earlier for Equation (7.6).
Equation (7.17) is the same as Equation (7.6) except that it is now applied
to pseudo loops also. Because a known q value is replaced by a known H
 Multisource Networks with Known Pipe Resistances 155

value, a pseudo loop is necessary. Thus, the number of AQ equations in-

creases by one for each such replacement.
Illustration. Taking the initial pipe flows as the same as the values
adopted in Subsection 7.4.2.3, the AQ equations for loops I . . . Ill are:

F, = 20(0.5 + AQ,)" - 30(0.05 - AQ, + Aft,)"

- 10(0.1 - Aft + Aft,,)- - 40(0.4 - Aft + Aft,,)" = 0 (7.18a)

F2 = 40(0.35 + Aft,)- - 50(0.05 - Aft,)"

- 20(0.15 - Aft,)" + 30(0.05 - Aft + Aft,)" = 0 (7.18b)

F3 = 40(0.4 - Aft + Aft,,)" + 10(0.1 - Aft + Aft,,)"

+ (95 - 100) = 0 (7.18c)

Equation (7.18b) is the same as Equation (7.7b). However, Equation

(7.18a) is different from Equation (7.7a) as loop I contains links 1 and 3
which are common to the pseudo loop also. Equation (7.18c) is newly in-
troduced for the pseudo loop.

7.5.2.4 AH Equations

The unknown nodal heads, //,, j = M + 1 . . . M + N are initially

assumed, and the corrections to them, i.e., AHj, j = M + 1 . . .
M + N, are taken as the basic unknown parameters in formulating the AH
equations.
General Formulation. The AH equations, one each for demand nodes
labelled M + 1 . . . Af + TV, are

i connected to
j through x

+ qoj = 0,j = M + 1 . . . M + N (7.19)

in which ,//, = //„„ ,//, = Hoj, ,A//, = 0, and ,A//y = 0 for /, j =

1 . . . M. Equation (7.19) is the same as Equation (7.8), however, Equation
(7.19) now contains a lesser number of equations. Naturally the number of
AH equations reduces by one for each additionally known H value.
Illustration. In the network of Figure 7.5, Hol = 100 m and Ho2 = 95
156 FORMULATION OF EQUATIONS

m. Let the assumed nodal heads be: tH3 = 97 m, tf4 = 94 m, ,// 5 = 90

m, and ,// 6 = 92 m. The AH equations (omitting the prefixing subscript
1 for simplicity) are
,100 - (97 + AH,)]1'" [(97 + AH3) - 9 5 ] " " n , _ n
1= 1 40 10 =
J " I J (7.2 0a)
1/ U
, 100 - (94 + A// 4 )l " [95 - (94 + AH4)] "
ri — I ^
20 JI + [ 30

(94 + AJ74) - (90

- 0.2 = 0 (7.20b)

, x, . + Aft) - (90
^3 —
40

(92 + Aft) - (90 + Aft)

(?20c)
50

95 - (92 + Aft)I"" [(92 + Aft) - (90 + Aft)

20 J [ 50 J

- 0.1 = 0 (7.20d)
Equations (7.20a) . . . (7.20d) provide four independent nonlinear
equations for the four basic unknowns Aft . . . AH6. Note that these
equations are the same as Equations (7.9a) and (7.9c) . . . (7.9e) except
the change in node labels. As in the case of H equations, for each addi-
tional known H value, the number of AH equations decreases by one.

7.6 NETWORKS WITH UNKNOWN PIPE RESISTANCES [96-98]

This is a general combination in which the network contains unknown

parameters of all three types, i.e., the nodal heads, nodal flows, and pipe
resistances. Further, in the earlier two combinations the flow was unknown
at a source node while the head was unknown at a demand node. A general
situation in which the flow is also known at a source node and the demand
is also unknown at a demandnode, i.e., both head and flow are known or
unknown at a node, is considered herein.
7.6.1 BRANCHING NETWORKS

In multisource branching networks, pseudo loops would be necessary.

Therefore, the formulation of equations for a branching network would be
similar to that for a looped network.
 Networks with Unknown Pipe Resistances 157

TABLE 7.1. Network Details.

Node Details Pipe Details

Head, Flow, Resistance

3
Node m m /s Pipe Constant

1 100.00 Qi 1 12.30
2 95.32 0.15 2 R2
3 94.78 0.31 3 48.60
4 94.30 0.30 4 24.60
5 H5 0.13 5 18.60
6 He 0.07 6 Re
7 7 49.60
— — — 8 58.10
— — — 9 14.30

7.6.2 LOOPED NETWORKS

Let a looped network consist of J nodes, X pipes, and C basic circuits

or loops. Then, according to Equation (6.1)

X =J + C- 1 (7.21)

Further, let 7UH denote the number of nodes with unknown head, JVq
the number of nodes with unknown flow (inflow or outflow) and XVR
denote the number of pipes with unknown resistance. According to the
unknown-number rule for network solvability,

•AJH + JVq + Xm = J (7.22)

Note that when both, the head and theflow,are unknown at a node, the
node is included in both, Jvli and JLq.
The formulation of the different equations is explained and illustrated
through the network of Figure 7.6. The unknown parameters are qu R2,
Hs, H6, R6, q7 and H7. The values of the known parameters are given in
Table 7.1.
Note that the network contains two unknown resistances: R2 and R6.
Further, the heads and flows are known at nodes 2, 3 and 4, while they are
unknown at node 7.
The network has seven nodes and seven unknowns, and therefore the
unknown-number rule is satisfied. //, . . . H4 are known and q, and q7 are
unknown, and therefore the minimum-number rule is also satisfied. The
assignment rule is satisfied as the assignment of unknowns to nodes can be
done as follows: qi to node 1, q7 to 7, H7 to 2, R6 to 3, H6 to 4, H5 to 5,
and R2 to 6. Three other ways of assignment are also possible:
158 FORMULATION OF EQUATIONS

(1) g, to 1, q7 to 7, H7 to 2, R6 to 3, H6 to 4, R2 to 5, and // 5 to 6
(2) gi to 1, q7 to 7, /?6 to 2, // 5 to 3, H7 to 4, // 6 to 5, and R2 to 6
(3) qx to 1, ^7 to 7, fl6 to 2, Hs to 3, W7 to 4, /?2 to 5, and Hb to 6

The unknown resistance Rb lies on a path connecting nodes 2 and 3 of

known heads. Similarly, unknown resistance R2 lies on a path connecting
nodes 1, 5, 6, and 4 for which the end nodal heads //, and H4 are known.
Further, these two paths form a branching configuration, thereby satisfy-
ing the path rule. Thus, all the network-solvability rules are satisfied.
7.6.2.1 Q-R Equations, Reducible to 0 Equations

General Formulation. All the pipe flows Qx, x = 1 . . . X and the un-
known pipe resistances, Ru, u = 1 . . . XVR are treated as the basic un-
known parameters in formulating the Q-R equations. Thus, the total
number of the unknowns is X + XUR. Applying the node-flow continuity
relationship to nodes with known nodal flows, J — JUq linear equations
can be obtained. Because the network has Cbasic loops, Cnonlinear loop-
head loss equations can be obtained. Further, as the network has J — 7UH
nodes with known heads, J — Jul — 1 head-loss equations for the
pseudo loops can be obtained. Thus, the total number of the available
equations is J — JUq + C + J — Ji:H — 1 which, using Equations
(7.21) and (7.22) is equal to X + XUR, the number of basic unknowns.
Thus, an adequate number of Q-R equations can be formulated to evaluate
the basic Q-R unknowns.
As is shown in the illustration that follows, it is possible to reduce the
X + Xy,R number of Q-R equations to X number of Q equations.
Illustration. In the network of Figure 7.6, X — 9 and XUR = 2. There-
fore, eleven Q-R equations must be available.

Node.l Pipe,5 2

Figure 7.6 Looped network with unknown pipe resistances.

 Networks with Unknown Pipe Resistances 159

IV

(b)

Figure 7.7 Loops and pseudo loops for network of Figure 7.6: (a) basic loops, (b) pseudo loops,
(c) overlapping loop, (d) replaced overlapping loop.

The node-flow continuity linear relationships for nodes 2, 3, 4, 5 and 6

(nodes with known q values) are, respectively

Ft = Qs - & - ft - 0.15 = 0 (7.23a)

F2 = 0 4 + 0 6 - 0 7 - 0.31 = 0 (7.23b)

^ = 03 + 07 - 0« - 0.30 = 0 (7.23c)

F< = 0, - 0 2 - 04 - 0.13 = 0 (7.23d)

^5 = 02 - 03 - 0.07 = 0 (7.23e)

The three basic loops I . . . Ill shown in Figure 7.7(a) yield, respec-
tively

Fb = 18.600? + R6Ql - 24.6004 - 12.3007 = 0 (7.23f)

F-, = 24.6005 + 49.600? - 48.6005 - R2Q"2 = 0 (7.23g)

F8 = 14.3003 - 58.1O0S - 4 9 . 6 0 0 - R6Q"6 = 0 (7.23h)

160 FORMULATION OF EQUATIONS

Because the network has four nodes with known heads, three pseudo
loops can be selected. It is preferable to select them such that they do not
contain pipes 2 and 6 with unknown resistances. Selecting pseudo loops
IV . . . VI as shown in Figure 7.7(b), the loop-head loss relationship
yields, respectively

F9 = 18.600 + 95.32 - 100.0 = 0 (7.23i)

F 10 = 12.30(2? + 24.6002 + 94.78 - 100.0 = 0 (7.23J)

Fu = 4 9 . 6 0 0 + 94.30 - 94.78 = 0 (7.23k)

Thus, we get eleven independent Q-R equations, Equations (7.23a) . . .

(7.23k) for eleven unknowns 0 . . . 0 , /?2 and R6.
For the formulation of Q equations, the number of Q basic unknowns is
nine. Equations (7.23a) . . . (7.23e) and (7.23i) . . . (7.23k) have only the
pipe discharges Q as the basic unknowns, and the number of these equa-
tions is eight. However, the ninth independent Q equation can be obtained
by adding Equation (7.230 and Equation (7.23h), i.e., by considering an
overlapping loop obtained by adding basic loops I and III as shown in
Figure 7.7(c). This eliminates pipe 6 with unknown resistance Rb. There-
fore, from this overlapping loop

Fl2 = 18.600 + 14.300 - . 5 8 . 1 0 0 - 4 9 . 6 0 0

- 2 4 . 6 0 0 - 12.300 = 0 (7.231)

Further, as H2 and H4 are known, this overlapping loop can be replaced

by a pseudo loop VII shown in Figure 7.7(d). For this pseudo loop

F,3 - 14.3003 - 5 8 . 1 0 0 + 94.30 - 95.32 = 0 (7.23m)

Thus, Equations (7.23a) . . . (7.23e), Equations (7.23i) . . . (7.23k),

and Equation (7.231) or Equation (7.23m) provide nine independent Q
equations for the nine Q unknowns—0 . . . Q9. After obtaining the
values of 0 . . . 0 , the values of R2 and R6 can be determined from
Equations (7.23g) and (7.23h), respectively.
As is shown here, it is therefore possible to obtain X number of indepen-
dent Q equations to evaluate basic unknowns 0 , x = 1 . . . X. Thus,
only the Q equations instead of the Q-R equations can be formulated. This
reduces the number of basic unknowns from X + Xm to X. However, it
is necessary to consider overlapping or pseudo loops or both.
Nonlinear equations, Equations (7.23f) . . . (7.23m) can be linearized
 Networks with Unknown Pipe Resistances 161

so that Equation (7.23a) . . . (7.23e) and the appropriate equations from

the linearized equations provide a set of linear equations.
7.6.2.2 H-q-R Equations, Reducible to H-R, and H Equations

General Formulation. The unknown nodal heads H, unknown nodal

flows q, and unknown pipe resistances R are taken as the basic unknown
parameters in formulating the H-q-R equations. Pseudo loops are not nec-
essary in formulating these equations. The H-q-R equations are formu-
lated by writing the node-flow continuity relationship at each node. The
total number of the available H-q-R equations is equal to the total number
of the nodes which is equal to the total number of the unknowns. The
equations are formulated as shown in the illustration.
Illustration. Writing the node-flow continuity relationship for nodes
1 . . . 7 of the network of Figure 7.6, we get, respectively,

/ 100.00 - 95.32\"« / 100.00 - HA

(7.24a)

100.00 - 95.32 \ 1 / n ( 9 5 . 3 2 - 94.78 \ ' / n

Fl
' ' 18.60 J ~\ R,

95 32 - H-,'
71
14.30 -0.15=0 (7.24b)

95.32 - 94.78 \ 1 / " / Hs - 94.78 \ " "

r3 =
Rb ) \ 24.60

94.78 - 94.30 \ 1 / n
- 0.31 = 0 (7.24c)
49.60

94.78 - 94.30 \ 1 / n I H6 - 94.30 \1/n

49.60 / \ 48.60

94.30 - HAUn
- 0.30 = 0 (7.24d)
58.10

100.00 - Hs\u" I Hs - 94.78 \ 1 / n

12.30 24.60

R
162 FORMULATION OF EQUATIONS

'Hs - H(

F
* = iT^ + ^"^ ~ q7 =0 (7.24g)

Equations (7.24a) . . . (7.24g) provide seven independent H-q-R equa-

tions for the seven basic unknowns—qx, R2, // 5 , // 6 , R6, q7, and H7. In
these equations, the unknown nodal flows qx and q7 are present only in
Equations (7.24a) and (7.24g), respectively. Therefore, if these equations
are kept aside, the remaining five equations, i.e., Equations (7.24b) . . .
(7.24f) form H-R equations for the five basic H-R unknowns—R2, Hs, H6,
R6, and H7. Thus, the H-R equations can be formulated by writing the
node-flow continuity equations at nodes of known flow.
Pipe 2 with unknown resistance R2 is connected to nodes 5 and 6 and
therefore, unknown resistance R2 is present in the node-flow continuity
equations for nodes 5 and 6, i.e., in Equations (7.24e) and (7.24f). How-
ever, if these two equations are merged, we get a new independent equa-
tion:
, 100.00 - HA1'" IH5 - 94.78
rg —
12.30 / \ 24.60

H6 - 94.30 \ 1/n
- 0.20 = 0 (7.24h)
48.60
which now does not contain the unknown resistance R2 and contains only
the basic //unknowns. Equation (7.24h) can be obtained by merging nodes
5 and 6 and then writing the node-flow continuity equation for this merged
node 5-6. This merged node 5-6 has pipe 1 with inflow and pipes 4 and 3
with outflows at node 5-6. The total external flow at this merged node is
0.13 + 0.07 = 0.20 m3/s. This technique of merging nodes at the ends of
a pipe of unknown resistance to obtain a merged node is similar to the
technique of merging basic loops to obtain an overlapping loop. Both these
techniques eliminate the pipe of unknown resistance from the correspond-
ing equations.
Similarly, by combining Equations (7.24b) and (7.24c), i.e., by merging
nodes 2 and 3, we get

_ / 100.00 - 95.32 \u" _ / 95.32 - HA1'" / H5 - 94.78 * ""

18.60 / \ 14.30 7 \ 24.60

94.78 - 94.30\ u "

- 0.46 = 0 (7.24i)
49.60
 Networks with Unknown Pipe Resistances 163

Equations (7.24d), (7.24h), and (7.24i) form a set of three independent

H equations for the three basic H unknowns, Hs . . . H7.
In this illustrative network, as the external flows at nodes 5 and 6 are
known, when these nodes are merged, the total external flow at merged-
node 5-6 is also known. However, assume that ^s instead of q7 is unknown
so that the seven basic unknowns are qu R2, q$, Hs, H6, R6, and H7. (Note
that all the network-solvability rules are satisfied.) In this revised network
when nodes 5 and 6 are merged, unknown nodal flow qs is present in the
node-flow continuity equation for this merged node 5-6. Thus, we do not
get an equation containing only H parameters as basic unknowns.
Now consider the assignment rule for this revised network. The un-
known parameters and their assigned nodes are as follows: qY to node 1, qs
to 5, R2 to 6, Hs to 3, H6 to 4, /^ to 2, and H7 to 7. From the fact that Hs,
Hf,, and H7 could be assigned to nodes 3, 4, and 7, respectively, it is evi-
dent that the external flows at these nodes must be known. Therefore, the
node-flow continuity equations are written for these nodes so that these
equations contain terms having basic H unknowns Hs, H6, and H7, and at
the same time, do not contain q parameters as unknowns. If, however, a
pipe with unknown resistance is connected at any of these nodes, the two
end nodes of such a pipe are merged. (Note that the external flow at the
other end of such a pipe must be known because the unknown pipe resis-
tance must be assigned to this node to satisfy the assignment rule.) There-
fore, for the revised network, the node-flow continuity equations are writ-
ten for merged node 2-3, and nodes 4 and 7 to obtain three independent H
equations for the three basic H unknowns Hs, H6, and H7.
The general procedure for formulating H equations can now be stated as
follows:
(1) Assign the basic q, H, and R unknowns to the nodes of the network,
satisfying the assignment rule.
(2) Identify the nodes to which the basic H unknowns are assigned.
(3) Identify the pipes with unknown resistances connected to any of
these nodes.
(4) Merge the nodes at the end of such pipes with unknown resistances.
(5) Write the node-flow continuity equations at all the identified nodes
including merged nodes, if any, to obtain the H equations.

To illustrate this procedure, consider the original network of Figure 7.6.

As stated earlier, the assignment rule is satisfied in four ways. Consider the
original assignment: qx to node I, q7 to 7, H7 to 2, R6 to 3, H6 to 4, Hs to
5, R2 to 6 (step 1). Herein the unknown //assignments are H7 to 2, H6 to
4, and Hs to 5 (step 2). At node 2, pipe 6 with unknown resistance is con-
nected (step 3). Therefore, the node at the other end of pipe 2, i.e., node
3, is merged with node 2 to obtain merged node 2-3 (step 4). Similarly, at
164 FORMULATION OF EQUATIONS

node 5, as pipe 2 with unknown resistance is connected, node 6 is merged

with node 5 to obtain merged node 5-6 (steps 3 and 4). The required H
equations are now written for merged node 2-3, node 4, and merged node
5-6 (step 5). These equations are Equations (7.24i), (7.24d), and (7.24h),
respectively, as obtained earlier.
Next consider the alternative assignment (1) in which the H assignments
are H1 to 2, H6 to 4, and Hs to 6. Following the aforementioned procedure,
this assignment gives nodes 2-3, 4, and 5-6. These nodes are the same
nodes obtained earlier. Similarly, if the other two alternative assignments
are considered, the same nodes 2-3, 4, and 5-6 are obtained, giving the
same three H equations. This therefore shows that even if several assign-
ments are possible to satisfy the assignment rule, the same H equations are
always obtained to get a unique solution for the network.
If the node-flow continuity equations are written for nodes 1, 2-3, 4, 5-6,
and 7, five H-q equations are obtained for the five basic H-q unknowns: qx,
H5, H6, H7, and q7.
The nonlinear H-q and H equations can be linearized as explained ear-
lier in Subsection 7.5.2.2 to obtain a set of linear equations.

7.6.2.3 AQ-Af? Equations, Reducible to AO Equations

General Formulation. In formulating the AQ-AR equations, pseudo

loops are formed so that the total number of real and pseudo loops is
C + J - JUI - 1, i.e., X - Jm. The pipe flows Qx, x = 1 . . . X,
and unknown pipe resistances Ru, u = 1 . . . XUR are initially assumed.
The Q values are initially assumed such that they satisfy the node-flow
continuity equations at nodes of known flow, this number being J — JUq.
In order to satisfy the loop-head loss equations for the real and pseudo
loops, the assumed Q and R values must be corrected by applying AQ and
AR corrections. The AQ-AR equations are formed for the real and pseudo
loops. Out of these X — 7UH equations, XLR equations are used to adjust
the assumed R values and the remaining X — Jm — Xm equations to
adjust the assumed Q values.
Illustration. For the network of Figure 7.6, X — JUH = 9 — 3 = 6,
and therefore six AQ-AR equations must be formulated. As XVR = 2,
two of these equations will be used to adjust the assumed R values and the
remaining four equations to adjust the -assumed Q values. The six loops
shown in Figures (7.7a) and (7.7b) can provide AQ-AR equations. Out of
these six loops, loops I and II will be used to adjust the assumed Re and
R2 values, respectively, while loops III . . . VI will be used to adjust the
assumed pipe flows. Let the corrections for loops I . . . VI be AR6, A/?2
(AR additive) and AQm . . . AQV, (clockwise positive), respectively.
Using the node-flow continuity equations for nodes 2 . . . 6, let the
 Networks with Unknown Pipe Resistances 165

assumed Q and R values be: ,<2i = 0.7 rrrVs; iQ2 = 0.3 m 3 /s;
, g 3 = 0.23 m 3 /s; ,Q4 = 0.27 nWs; ,QS = 0.6 m°7s; , g 6 = 0.3 mVs;
,(?7 = 0.26 m 3 /s; ,(?„ = 0.19 m 3 /s; ,(?, = 0.15 m 3 /s; ^ = 25; and
i/?6 = 30. The AQ-AR equations for loops I . . . VI will be, respectively

Ft = 18.60(0.6 + Aflv)" + (30 + AR6) (0.3 - AQm)n

- 24.60(0.27 + AQwY - 12.30(0.7 + Agv)" = 0 (7.25a)

F2 = 24.60(0.27 + AQy)" + 49.60(0.26 - AQIU + AQvl)"

- 48.60(0.23)" - (25 + Afl2) (0.3)" = 0 (7.25b)

F3 = 14.30(0.15 + Aflii)" - 58.10(0.19 - Ag,,,)"

- 49.60(0.26 - Afiii + AfiVI)" - (30 + AR6)

X (0.3 - AQm)n = 0 (7.25c)

F4 = 18.60(0.6 + A£?IV)" + 95.32 - 100.00 = 0 (7.25d)

Fs = 12.30(0.7 + AQvy + 24.60(0.27 + AQV)" + 94.78

- 100.00 = 0 (7.25e)

F6 = 49.60(0.26 - Aft,, + A£>VI)" + 94.30 - 94.78 = 0 (7.250

Thus, six independent AQ-AR equations are obtained for six unknowns:
A/?2, A/?6, AQm . . . AGvi.
To obtain AQ equations, since there are four AQ unknowns, four equa-
tions containing only AQ terms and not AR terms must be formed. Equa-
tions (7.25d), (7.25e), and (7.250 do not contain AR terms. Equations
(7.25a) and (7.25b) contain AR terms but they are used to adjust the
assumed R6 and R2 values, respectively. As Equation (7.25c) contains a AR
term, it is replaced by an equation for a loop that does not contain either
pipe 2 or pipe 6. Therefore, consider the overlapping loop as shown in
Figure (7.7c), or simply the pseudo loop shown in Figure (7.7d). Thus,
Equations (7.25a), (7.25b), (7.25c), and (7.25f) would now modify to, re-
spectively,

F7 = 18.60(0.6 + Afiv)" + (30 + AR6) 0.3"

- 24.60(0.27 + A0V)" - 12.30(0.7 + AQV)" = 0 (7.25g)

166 FORMULATION OF EQUATIONS

Fs = 24.60(0.27 + AQV)" + 49.60(0.26 +

- 48.60(0.23)" - (25 + AR2) (0.3)" = 0 (7.25h)

F9 = 14.30(0.15 + AGvn)" - 58.10(0.19 - Afivu)"

+ 94.30 - 95.32 = 0 (7.25i)

F,o = 49.60(0.26 + AQVi)n + 94.30 - 94.78 = 0 (7.25J)

Thus, Equations (7.25g), (7.25h), (7.25d), (7.25e), (7.25j), and (7.25i) is

a set of six modified AQ-AR equations for the six basic unknowns. In these
equations A/?6 and AR2 are present in Equations (7.25g) and (7.25h) only
and are absent in the other four equations. Therefore, Equations (7.25d),
(7.25e), (7.25J) and (7.25i) provide four AQ equations for the four basic
AQ unknowns—A<2,v . . . AQVU.

7.6.2.4 AH-Aq-AR Equations Reducible to AH-AR, and AH Equations

General Formulation. AH the unknown H, q, and R parameters are ini-

tially assumed and corrections AH, Aq, and AR to these assumed values,
respectively, are treated as the basic unknowns in formulating the AH-Aq-
AR equations. The equations are formulated by writing the node-flow con-
tinuity equations at all the J. nodes of the network. Pseudo loops are not
necessary.
Illustration. For the network of Figure 7.6, let the assumed values be:
,(?, = 1.00 m3/s, tR2 = 25, , # 5 = 98.00 m, ,/f6 = 96.00 m, ,/?6 = 30,
3
xqn = 0.08 m /s, and JI-, = 94.00 m. The node-flow continuity equations
for nodes 1 . . . 7 yield, respectively,
u
«r nnnxA , / 100.00 - 95 .32 \ "
F, = (1.00 + A ) | 18.60 j
100.00 - (98.00 + AZ/s)1""
= 0 (7.26a)
12.30

100.00 - 95.32\ 1/n - Z95.32 - 94.78V'

F2 =
18.60 / \ 30 + A/?6

95.32 - (94.00 + AH7)]U

- 0.15 = 0 (7.26b)
14.30
 Networks with Unknown Pipe Resistances 167

95.32 - 9 4 . 7 8 \ U n [ ( 9 8 . 0 0 + A// 5 ) - 9 4 . 7 8 l ' / n

3_
+
~ 30 + AR6 ) +
[
[ 24.60 J

94.78 - 94.30\ u " + [(96.00 + A//6) - 94.30

49.60 / ) [I 48.60

» - M - £ r "»]"•-0.3.-.
100.00 - (98.00 + AH5)V/n [(98.00 + AH5) - 94.781""
_
12.30 | I 24.60 J

- 0.13 = 0 (7.26e)

_ [(98.00 + Art5) - (96.00 + AH6)]l/"

25 + AR2
^ 0 1 \ln
- 0.07 = 0 (7.26f)

95.32 - (94.00 + AH,)]1'" [94.30 - (94.00 + AH7)

- (0.08 + Aq7) = 0 (7.26g)

Equations (7.26a) . . . (7.26g) are seven independent AH-Aq-AR equa-
tions for the seven corrections: A ^ , AR2, AH5, AH6, ARf,, Aq7, and AH7.
As the corrections Aqt and Aq7 are present in Equations (7.26a) and
(7.26g) only, if these equations are kept aside, the remaining five equations
give AH-AR equations. Further, as is done in formulating H-q-R equa-
tions, by merging nodes, we get merged nodes 5-6 and 2-3 for which, re-
spectively,
_ [100.00 - (98.00 + Affs)]""
] [[(98.00 + A//5) - 94.78 V'
12.30
J
J
[
[ 24.60
JJ
- 0-20^0 (7.26h)
48.60
168 FORMULATION OF EQUATIONS

_ 100.00 - 95.32\"" _ r95.32 - (94.00 + A//7)

9
~ ' 18.60 ) ~ 14.30

,(98.00 + A//5) - 94.781""

'" _ /94.78
/ 94.78 -- 94.30
24.60 J \ 49.(
49.60

- 0.46 = 0 (7.26i)

Equations (7.26d), (7.25h), and (7.26i) provide three independent AH

equations for the three corrections AH5, AH6, and A// 7 .

7.7 INCLUSION OF PUMPS

As stated in Chapter 5 section 2, pumps are included in distribution net-

works as supply pumps to provide water to the network from external
sources or as booster pumps on the pipes to boost up pressure within the
networks. Two types of head-discharge relationships are used for the
pumps: (1) A relationship giving the head as a function of discharge
[Equations (5.4) . . . (5.7)], and (2) a relationship giving discharge as a
function of the head [Equation (5.8) . . . (5.12)].
The inclusion of pumps in the different types of equations is explained
herein through a network having a supply pump. For simplicity, the resis-
tance constants of all the pipes are treated as known. However, by combin-
ing the principles explained herein with those in earlier sections, inclusion
of pumps in the formulation of equations for networks with unknown pipe
resistance can be done.

7.7.1 SUPPLY PUMPS

Consider the network shown in Figure 7.8 with a pump provided in pipe
1 to lift water from the sump and supply it to the network. The network
also contains an elevated reservoir at node 2, connected to the network
through pipe 7.
Consider for the time being that reservoir 2 and pipe 7 are absent so that
the network is a single-source pumped network. As all the nodal demands
are known, the discharge supplied by the pump is known, and using the
head-discharge relationship, the head supplied by the pump is known. This
is then similar to the single-source networks described earlier.
The network of Figure 7.8 has two source nodes—a pump and a reser-
voir—and therefore pseudo loop, loop III, is introduced.
 Inclusion of Pumps 169

2
Elevated
Reservoir,
H02

Figure 7.8 Network with supply pump.

7.7.1.1 0 Equations

The network has seven pipes and therefore seven independent Q equa-
tions are necessary. Node-flow continuity equations at nodes 3 . . . 6, and
loop-head loss equations for real loops I and II provide six equations. The
seventh equation is available from the pseudo loop as follows:

Ro2Q"2 - - Ro7Q"7 + Ho2 - Hol = 0 (7.27)

in which hp is the head supplied by the pump. The parameter hp is ex-

pressed in terms of <2, through any one of the head-discharge relationship,
Equations (5.4) . . . (5.7). Thus, Equation (7.27) provides the seventh Q
equation. Note that if the relationship of Equation (5.7) is used, all terms
in Equation (7.27) would be similar. This would help in solving the equa-
tions.

7.7.1.2 H Equations

There will be five unknowns, H3 . . . H6, and hp in the formulation of

H equations. The node-flow-continuity relationships at nodes 4, 5, and 6
yield three H equations. Now for pipe 1

hp — H3 = (7.28)

from which

(7.29)
170 FORMULATION OF EQUATIONS

Therefore, the node-flow continuity relationship at node 3 would yield

IHO1 +hp- H3y _ IH3 - H4y- _ IH3 - H5y- _ qi = 0

l o2 3
\ ^° I \ ^ / \ ^° / (7 30)

Further, as Qx = Qp, Equation (7.29) and one equation from Equations

(5.9) . . . (5.12) would yield the fifth equation. For example, Equations
(7.29) and (5.11) would yield

(Hol + hp - HAi/n _ \ K - {Ho - B2/4A) 1° 5 _ B_

\ Rol J ~[ A \ 2A
(7.31)

Thus, node-flow continuity equations at nodes 4, 5, and 6 and Equations

(7.30) and (7.31) provide five independent H equations for five unknowns,
H3 . . . H6, and hp.
Now, if the head-discharge relationship as given by Equation (5.7) is
used for the pump, substituting the value of hp from Equation (5.7) in
Equation (7.28), we get

Hol + Hop - ROPQ{ - H3 = RoiQ{ (7.32)

or

1. + H - H3
(7.33)
K
op

which modifies Equation (7.30) to

Hol + Hop - H3V" (H3 - HA1'"

+ Roi I \ R<

/"
i<a - qo3 = 0 (7.34)

Thus, hp need not be taken as the fifth unknown parameter. The un-
known parameters are H3 . . . H6 for which the three node-flow continuity
equations at 4, 5, and 6 and Equation (7.34) provide the required four H
equations.
 Inclusion of Pumps 171

7.7.1.3 AQ Equations

The pipe flows are initially assumed so that they satisfy the node-flow
continuity relationships at nodes 3 . . . 6. Thus, the AQ equations for
loops I . . . Ill are, respectively,

RoidQi + AG, - Afin)" + /U,<2 4 + A<2. ~ AQ,,)"

- RAiQs - A<2,)" = 0 (7.35a)

- RosdQs ~ Ag n + AGm)" + RoeUQe + Afl,)"

-Ro4dQ* + AQ, - Afl,)- = 0 (7.35b)

/ U t S r + Aflii)" + RosdQs - AQn + AQ,,,)"

- RAiQi + AG, - AGm)" - RoidQi - AGm)"

+ hp + HBl - Ho2 = 0 (7.35c)

In Equation (7.35c) hp is expressed in terms of Qp which is now

iQi — AGIH- Equations (7.35a) . . . (7.35c) provide three AQ equations
for the three basic unknowns AQ{ . . . AQm.

7.7.1.4 AH Equations

The four H unknowns, H3 . . . H6, and the pump head hp are initially
assumed, and corrections AH3 . . . AH6 and Ahp are applied so that the
corrected values are iH3 + AH3 . . . !# 6 + AH6, and ihp + Ahp. These
values are then used in node-flow continuity equations at nodes 4, 5, and
6 in Equations (7.30) and (7.31). If Equation (5.7) is used as a head-
discharge relationship for the pump, hp need not be assumed as explained
earlier. The node-flow continuity equations at nodes 4 . . . 6 and 3 [Equa-
tion (7.34)] provide four AH equations for the four basic AH unknowns,
AH3 . . . AH6.

7.7.2 BOOSTER PUMPS

Booster pumps are provided within the networks and therefore they lie
on pipes forming real or pseudo loops. The formulation of different equa-
tions, however, is similar to that for supply pumps and therefore it is not
described separately.
172 FORMULATION OF EQUATIONS

7.8 INCLUSION OF CHECK VALVES

We have seen in Chapter 5 that when check valves are provided in pipes
they permit flow in one direction. Different situations regarding the provi-
sion of check valves are (1) check valve in an external pipe, (2) check valve
in an internal pipe, and (3) check valves in several pipes. Formulation of
different types of equations for these situations is explained herein.

7.8.1 CHECK VALVE IN AN EXTERNAL PIPE

Assume that a check valve is provided in an external pipe, pipe 7 as

shown in Figure 7.9(a). (This network is the same as that given in Figure
7.2 but a check valve is included in pipe 7.) This check valve permits flow
from node 6 to node 5 but prevents flow in the reverse direction. Pipe 7 is
external, and therefore the check valve is present in only one basic loop,
loop II.

7.8.1.1 0 Equations

The Q equations are the same as Equations (7.3a) . . . (7.3g). However,

during the solution of these equations if it is observed that Q7 becomes

Figure 7.9 Network with check valve: (a) check valve in external pipe, (b) check valve in internal
pipe.
 Inclusion of Check Kilves 173

negative, Q7 is made zero. This case would be as if pipe 7 does not exist,
and the network has six pipes. Therefore, Q7 is made zero in the node-flow
continuity equations for node 5 and 6, i.e., Equations (7.3d) and (7.3e), re-
spectively. Further, as pipe 7 does not exist, loop II also does not exist and
therefore the head loss equation for this loop, i.e., Equation (7.3g), is
dropped. Thus, for the six unknowns Qx . . . Q6, the available six Q equa-
tions are Equations (7.3a) . . . (7.3c), modified Equations (7.3d) and
(7.3e), and Equation (7.3f).

7.8.1.2 H Equations

The H equations arc the same as Equations (7.5a) . . . (7.5e) in which

the flow direction in pipe 7 is from node 6 to node 5, i.e., H6 > Hs. How-
ever, during the solution of the equations if it is observed that H6 < Hs,
i.e., H6 — Hs < 0 giving flow from node 5 to node 6, (H6 - Hs) is made
zero. This condition will apply to the node-flow continuity equations for
nodes 5 and 6, i.e., Equations (7.5d) and (7.5e), respectively.

7.8.1.3 AO Equations

The AQ equations are the same as Equations (7.7a) and (7.7b). However,
during any stage of the solution of the AQ equations if the application of
the correction AQn to the then assumed discharge in pipe 7 reverses the
flow direction, i.e., for the fh iteration if ,Q7 + ,AQn < 0, the correc-
tion tAQu is restricted so that ,Q7 + tAQu is zero. Thus, for the next
iteration, t+1£?7 = 0. For example, for the first iteration, as iQ7 = 0.05
m3/s from node 6 to node 5, if iAQn > 0.05 nrVs, in the clockwise direc-
tion, then ,AQ,i is restricted to 0.05 m3/s in the clockwise direction.

7.8.1.4 AW Equations

The AH equations are the same as Equations (7.9a) . . . (7.9e) in which

the flow direction in pipe 7 is from node 6 to node 5. However, during the
solution of AH equations if it is observed that for the t'h iteration,
(,HS + ,AH6) < (,HS + ,A// 5 ), the discharge in pipe 7 is made zero by
making [(,H6 + ,AH6) - (,HS + ,AHS)] equal to zero.

7.8.2 CHECK VALVE IN AN INTERNAL PIPE

Assume that a check valve is provided in pipe 4 as shown in Figure

7.9(b). The permissible flow direction in pipe 4 is from node 3 to node 4.
Pipe 4 is internal and thus is common to both the basic loops, I and II.
174 FORMULATION OF EQUATIONS

7.8.2.1 Q Equations

If the procedure described earlier for the check valve in an external pipe
is followed, there is no problem for the node-flow continuity equations.
However, as pipe 4 is common to both loops, the procedure would elimi-
nate both the loops, and the number of available equations would be five
for the six Q unknowns. This can be avoided by externalizing pipe 4, i.e.,
keeping it in only one loop. This can be achieved by considering two
loops: (1) loop I, an overlapping loop comprising pipes 1, 2, 5, 7, 6, and
3 and (2) loop II comprising pipes 1, 2, 4 and 3 or pipes 4 , 5 , 7, and 6. The
Q equations are then formulated as explained earlier for a check valve in
an external pipe.

7.8.2.2 H Equations

There is no problem herein and the procedure described earlier for a

check valve in an external pipe can be followed.

7.8.2.3 AO Equations

As the pipe is an internal pipe, the discharge in it is affected by both the

AQ corrections. Depending upon the magnitude of Q4, and the magnitudes
and directions of AQU and A(Q,,, the following situations may arise:
1. Q4 = 0.05 nrVs, AQ, =. -0.10 m7s, and AQn = 0.10 m3/s. There
is no problem in this situation and both the corrections can be applied as
they increase Q4.
2. Q4 = 0.05 nvVs, A£>, = -0.10 m3/s, and AQn = - 0 . 2 0 m3/s.
Herein AQ, can be fully applied while AQU is restricted to —0.15 m3/s
so that Q4 = 0.
3. Q4 = 0.05 m3/s, AQ, = 0.20 m3/s, and AQU = 0.10 m3/s. Herein
A<2n can be fully applied while AQi is restricted to 0.15 mVs so that
Q4 = 0.
4. Q4 = 0.05 m3/s, AQ, = 0.10 nvVs, and AQU = -0.10 m3/s. Herein
neither correction can be applied fully and AQ, and AQ] should together
be restricted such that Q4 = 0. However, infinite possibilities exist,
for example, AQ, = 0, and AQu = -0.05 m 3 /s, AQ, = 0.01 m3/s, and
AQ], = —0.04 m3/s, and so on.
Consideration of such possibilities can be avoided if pipe 4 is external-
ized as explained in formulating Q equations. The discharge Q4 is affected
by only one AQ correction, and the procedure explained for formulating
AQ equations for a check valve in an external pipe can be followed.
 Inclusion of Pressure Reducing Valves 175

7.8.2.4 AH Equations

No problem exists herein and the procedure described earlier for a

check valve in an external pipe can be adopted.

7.8.3 CHECK VALVES IN SEVERAL PIPES

If check valves exist in two pipes, say pipes 7 and 4, or in pipes 5 and
4, the check valves can be externalized by considering the overlapping
loop and the basic loop I, and the procedure described earlier can be fol-
lowed.

7.9 INCLUSION OF PRESSURE REDUCING VALVES

We have seen in Chapter 5 that a pressure reducing valve, PRV, can

function in three different modes:
(1) Mode 1 in which it becomes inoperative when the pressure is less
than the pressure setting
(2) Mode 2 in which it becomes operative, i.e., it maintains a constant
pressure on the downstream regardless of the pressure on the up-
stream
(3) Mode 3 in which it behaves like a check valve if it is bypassed, or the
flow direction is reversed

Mode 1 of operation in which a PRV is inoperative can be easily in-

cluded in the formulation of equations by considering the additional resis-
tance offered by the PRV. Mode 3 of operation in which a PRV behaves
like a check valve can be included in the formulation of equations as
described in the last section. However, the formulation of equations with
PRVs for their mode 2 of operation requires special treatment. Herein also
three situations exist: (1) PRV in an external pipe, (2) PRV in an internal
pipe, and (3) PRVs in several pipes.

7.9.1 PRV IN AN EXTERNAL PIPE

Consider again the network of Figure 7.2 in which a PRV is provided in

an external pipe, pipe 7, as shown in Figure 7.10(a). Let the preset HGL
value at the downstream end d of the PRV be // sel , a known value. Let the
known resistance constant of pipe segment d5 of pipe 7 be RO.M. Two con-
ditions prevail for this PRV:
(1) The HGL at the downstream end d of the PRV remains Hxt, a
176 FORMULATION OF EQUATIONS

2 3 3
(d)

d,

(f)

Figure 7.10 Networks with pressure-reducing valves: (a) PRV in external pipe, (b) considered
loops, (c) PRV in internal pipe, (d) considered loops, (e) PRVs in several pipes without isolating
a portion, (f) considered loops, (g) PRVs in several pipes isolating a portion, (h) considered
loops.

known and fixed value. Therefore, d can be treated as a source node

with HGL equal to H^. The single-source network has become a
two-source network with 1 and das source nodes. Note that the HGL
at the upstream end M of the PRV is unknown. Similarly, the resis-
tance constant of the PRV is variable and therefore unknown.
(2) Flow continuity must be maintained at the PRV. Therefore, dis-
charges in pipe segments 6-M and d-5 must be the same and equal to

From this it is clear that a PRV can be replaced by (1) a source node d
 Inclusion of Pressure Reducing Valves Yll

with known HGL equal //se, and inflow Q7 and (2) a sink node u with
unknown head and outflow Q7.
7.9.1.1 0 Equations

The node-flow continuity equations for nodes 2 . . . 6 and the head loss
equation for loop I are given by Equations (7.3a) . . . (7.3f), respectively.
However, the loop-head loss equation, Equation (7.3g), is not applicable
because of condition 1. The PRV is replaced by source and sink nodes, and
a pseudo loop is formed as shown in Figure 7.10(b). Therefore, for this
pseudo loop, loop II

F7 = 2Og2 + 40{£ - & - „ Q" + //sc - 100 = 0 (7.36)

Thus, Equations (7.3a) . . . (7.3f) and (7.36) provide seven Q equations

for 0 . . . . Q7.
7.9.1.2 H Equations

The H equations for the flow continuity at nodes 2 . . . 4 remain unaf-

fected and are given by Equations (7.5a) . . . (7.5c), respectively. How-
ever, the H equations for nodes 5 and 6 modify to

and

F, - ( « ^ * ) - ( « - = ; * ) - 0..-0 (7.37b,

Note that as Hu is unknown, discharge in pipe segment 6-u cannot be ex-

pressed in terms of//„. Similarly, although the HGL at d is known, the re-
sistance constant of the PRV is variable and therefore unknown. Therefore,
the discharge in pipe segment 6-d cannot be expressed in terms of R6d.
Thus, the discharge in pipe segment 6-u which is equal to the discharge in
pipe segment d-5 (condition 2) is expressed in terms of Hxt, H5, and

Equations (7.5a) . . . (7.5c), (7.37a), and (7.37b) provide five H equa-

tions for Hi . . . H6.
7.9.1.3 AQ Equations

The AQ equations can be written for basic loop I and pseudo loop II
shown in Figure 7.10(b). When the correction AQ, is obtained for loop I
178 FORMULATION OF EQUATIONS

and applied to the pipes of loop I, node-flow continuity is maintained at all

nodes. Because only two pipes are considered at a node in loop I, the effect
of the correction on the discharge in one pipe is neutralized by that in the
other for all nodes in the loop.
The pseudo loops considered in the earlier sections were formed by
source nodes with unknown flows; therefore, corrections can be applied to
the pipes of the pseudo loops without violating the node-flow continuity
equations at all nodes including the source nodes. The change in the flow
at one source node would be balanced by the corresponding change in the
other.
The pseudo loop is formed herein by splitting a PRV into source and
sink nodes. If the correction AQn obtained from pseudo loop II is applied
to the pipes of this pseudo loop, the node-flow continuity relationship is
violated at the PRV and also at node 1. Therefore, to maintain the node-
flow continuity at the PRV, i.e., at nodes d and u, the correction applied
to pipe segment d-5 must continue to pipe segment 6-u, and therefore, to
pipes 6 and 4. Therefore, the loop correction AQn obtained from the
pseudo loop should be applied to pipes of real loop II comprising pipes 4,
5, 6 and 7.
The A<2 equations can now be written for real and pseudo loops. The
A<2 equation for the real loop, loop I, is the same as Equation (7.7a). The
A<2 equation for the pseudo loop is

F2 = 20(0.5 + AQi)n + 40(0.35 + AQn)n - Ro.d5(0.05 - AQn)n

+ Hx{ - 100 = 0 (7.38)

Equations (7.7a) and (7.38) provide two AQ equations for two correc-
tions, AQi and AQn. As stated earlier AQ will be applied to pipes
1 . . . 4 of loop I and AQt to" pipes 4 . . . 7 of real loop II.

7.9.1.4 AH Equations

The AH equations for the flow continuity at nodes 2 to 4 remain unaf-

fected and are given by Equations (7.9a) . . . (7.9c). The AH equations for
nodes 5 and 6 now modify tp

(94 + AH,) - (90 + AH5)]U" f//seI - (90 + AHS)'

40 J L. *-* J
- 0.4 = 0 (7.39a)
 Inclusion of Pressure Reducing Valves 179

and

(95 + AH3) - (92 + AH,)]1"1 \Hxt - (90 + AH,)]1'"

20

- 0.1 = 0 (7.39b)

Equations (7.9a) . . . (7.9c), (7.39a), and (7.39b) provide five H equa-

tions for five unknowns AH2 . . . AH6.

7.9.2 PRV IN AN INTERNAL PIPE

When a PRV is provided in an internal pipe, say pipe 4 as shown in

Figure 7.10(c), pipe 4 can be externalized by considering the overlapping
and basic loops as shown in Figure 7.10(d). Equations can now be formu-
lated as follows.

7.9.2.1 0 Equations

The node-flow continuity equations for nodes 2 . . . 6 give Q equations

which are the same as Equations (7.3a) . . . (7.3e). The head loss equa-
tions for the overlapping and the pseudo loops are, respectively,

F6 = 2 0 0 + 4 0 0 - 5 0 0 - 2 0 0 - 1 0 0 - 4 0 0 = 0 (7.40a)

and

F 7 = 2 0 0 - /?o-d40 + Hxt - 100 = 0 (7.40b)

7.9.2.2 H Equations

The H equations for the flow continuity at nodes 2, 5, and 6 are the same
as Equations (7.5a), (7.5d), and (7.5e), respectively. The H equations for
nodes 3 and 4 modify to, respectively,

1
~\ 10 / \ Ro.d4 ) [ 20 j
+ 0.1 = 0 (7.41a)
180 FORMULATION OF EQUATIONS

and

100 - HAV" + / HM - H4
3
~ ' 20 / • \ R^4 ) \ 40

- 0.2 = 0 (7.41b)

7.9.2.3 AQ Equations

The AQ equations for the overlapping and pseudo loops are, respec-
tively,

20(0.5 + A<2, + AQ,,)" + 40(0.35 + AQ,)" - 50(0.05 - A£i)"

- 20(0.15 - AQ,)" - 10(0.1 - AQi)n

- 40(0.4 - AQ,)" = 0 (7.42a)

and

20(0.5 + A<2, + AQnT - #,.,,4(0.05 - Afi,,)" + Hsa - 100 = 0

(7.42b)

The correction AQ, will be applied to pipes of the overlapping loop,

i.e., pipes 2, 5, 7, 6, 3 and 1. The correction AQ:, will be applied to real
loop comprising pipes 1 . . . 4, to maintain node-flow continuity.

7.9.2.4. AH Equations

The AH equations for flow continuity at nodes 2, 5, and 6 are given by

Equations (7.9a), (7.9d) and (7.9e). The AH equations for nodes 3 and 4
are, respectively,

(97 + AH2) — (95 + A//3) Hxt - (94 - AH4)V>n

10 Ro-dA J

(95 + AH3) — (92 +'A// 6 )

+ 0.1 = 0 (7.43a)
20

and
 Problems 181

100 - (94 + AH4)V'" \HXX - (94 +

20 J [ Ro-d4

(94 + AH4) - (90 + A//5) 1 "

- 0.2 = 0 (7.43b)
40

7.9.3 PRVs IN SEVERAL PIPES

When PRVs are provided in several pipes of a network, two situations

arise: (1) when the PRVs do not isolate a portion of a network and (2) when
the PRVs isolate a portion of the network.
The first situation arises if no portion of the network is isolated on split-
ting the PRVs into source and sink nodes. For example, if PRVs are pro-
vided in pipes 4 and 7, as shown in Figure 7.10(e), no portion of the net-
work is isolated. The PRVs are externalized and the two pseudo loops
shown in Figure 7.10(f) are considered. Different equations can be written
following the procedure described earlier.
If the PRVs are provided in pipes 5 and 6, or in pipes 6 and 7, or in pipes
3,4, and 7, nodes 5 and 6, or node 6, or nodes 3 and 6 are isolated, respec-
tively. Consider, for example, the last situation as shown in Figure 7.10(g).
The easier method to analyze such a situation would be to treat the two
parts as separate networks. The downstream ends of the PRVs are source
nodes, and loops are formed as shown in Figure 7.10(h). The isolated net-
work is analyzed, and the values of Q3, Q4, and Q7 are found and are taken
as demands at nodes ut, U2, and u3, respectively. The analysis of the main
portion is then carried out.
Instead of considering two portions separately, the entire network can be
considered at the same time and the equations can be formulated. In AQ
equations, however, the corrections AQj and AQn are applied to the real
loops I and II shown in Figure 7.10(g). On application of the corrections,
if the flow direction reverses through some PRVs, such PRVs would func-
tion as check valves and therefore the concerned AQ corrections should be
restricted as explained earlier.

7.10 PROBLEMS

Formulate all possible types of equations in the following problems.

Problem 7.1: The single-source network shown in Figure: Problem 7.1
has node 1 as a source node with H G L = 100.00 m, and nodes 2 . . . 5
182 FORMULATION OF EQUATIONS

2 0-08m% - 0-06

as demand nodes with demands as shown in the figure. The length, diam-
eter, and HW coefficient of the pipes are as follows: Pipe 1—300 m, 250
mm, 120; pipe 2—250 m, 300 mm, 110; pipe 3—300 m, 200 mm, 120;
pipe 4—400 m, 100 mm, 100; pipe 5—200 m, 200 mm, 90; and pipe
6—300 m, 150 mm, 100; respectively.
Problem 7.2: The two-source network shown in Figure: Problem 7.2
has nodes 1 and 2 as source nodes with HGL values 110.00 m and 100.00

m, respectively. Nodes 3 . . . 6 are demand nodes with demands as

shown in the figure. The resistance constant (R) values in the head loss
equation h = RQ2 (h in meters, Q in cubic meters per minute) are shown
underlined in the figure.
Problem 7.3: In the network of Figure: Problem 7.1, R4 is unknown and
H5 = 85.00 m.
Problem 7 4 : In the network of Figure: Problem 7.2, R3 and /?6 are un-
known and H3 = 107.00 m, and H6 = 95.2 m.
Problem 7.5: The source at node 2 in Figure: Problem 7.2 is removed
and node 2 is connected to a sump (node 7, W.L. = 90.00 m) by pipe 9
(R = 0.1). A pump with head-discharge characteristic h = 27.4 — 0.06
Q2 (h in meters, Q in cubic meters per minute) is fitted in pipe 9.
Problem 7.6: In the network of Problem 7.5, ^?6 is unknown and H3 =
109.00 m.
Problem 7.7: In the network of Figure: Problem 7.2, a check valve
(K — 2.5) is fitted in pipe 3 (diameter = 150 mm) with permissible direc-
tion of flow form (1) node 3 to node 5, and (2) node 5 to node 3.
Problem 7.8: The two-source network shown in Figure: Problem
 Problems 183

7.8 has nodes 1 and 2 as source nodes with HGL values of 105.00 m and
100.00 m, respectively. Nodes 3 . . . 6 are demand nodes with demands as
shown in the figure. The R values for the pipes are shown underlined in the
figure. The head loss in pipes is given by h = RQl8&2 Qi in meters, Q in
cubic meters per second). A PRV (K = 10) is fitted in pipe 6 (diame-
ter = 400 mm) just downstream of node 5 as shown in the figure, with a
set pressure head of 90.00 m.
 Hardy Cross Method

8.1 INTRODUCTION

r has been shown in Chapter 7 that except for a single-source branch-

I ing network some or all formulated equations are nonlinear. Because the
number of the available independent equations is equal to the number of
the unknown parameters, a solution would be feasible. However, no
method is available that can directly solve nonlinear equations, and there-
fore an iterative procedure is necessary for their solution [99]. Three
methods are commonly used in practice for the iterative solution of these
equations and thereby for the analysis of water distribution networks.
These methods, in chronological order of their application to network
analysis are (1) the Hardy Cross method, (2) the Newton-Raphson method,
and (3) the linear theory method. The Hardy Cross method is explained in
this chapter and the other two in subsequent chapters.
Hardy Cross [100] was perhaps the first to suggest in 1936 a systematic
iterative procedure for network analysis [1,14,94]. His approach as origi-
nally proposed by him, is based on loop-flow correction equations, i.e.,
AQ equations. This approach, as explained later, is also known as the
method of balancing heads. Later, Cornish [101] also applied the principles
to nodal-head correction equations, i.e., AH equations. This approach is
known as the method of balancing flows. Both these approaches, i.e., the
method of balancing heads and the method of balancing flows, are pres-
ently included under one name, the Hardy Cross method.

8.2 METHOD OF BALANCING HEADS

8.2.1 BASIC CONCEPT

The method of balancing heads as proposed by Hardy Cross iteratively

solves the AQ equations on the following assumptions:
185
186 HARDY CROSS METHOD

A iQi
i Roi
1
oc o
E en
q
e —,
k Rok
,Qk B

Figure 8.1 Single loop of distribution network.

(1) Only one equation is considered at a time from the available set of
AQ equations.
(2) The effect of adjacent loops is neglected and therefore each AQ equa-
tion contains only one AQ term as unknown. For example, in Equa-
tion (7.7a) of Chapter 7, the correction AQ, is retained and AQ,, is
discarded. Similarly, in Equation (7.7b) the correction AQn is re-
tained and AQ, is discarded. Thus, except the correction for the
loop under consideration, corrections for other adjoining loops are
assumed to be zero.
(3) Each term of the modified AQ equation is expanded in a Taylor's
series with only the first-power AQ terms retained, and higher-power
AQ terms are neglected.
Now consider a single loop of a distribution network as shown in Figure
8.1 (assumption 1). Let it consist of four pipes—i,j, k, and m, with known
resistances Rol, Roj, Rok, and Rom, respectively. Let the assumed pipe dis-
charges satisfying the node-flow continuity relationships be ,£?., iQj, i2*>
and iQm in pipes i, j , k, and m, respectively, in the direction as shown in
the figure. Let IAQ be the clockwise correction along the loop as shown in
the figure. The AQ equation for this loop, after disregarding the effect of
adjacent loops (assumption 2) is

oAtQ, + iAQ)" + + - iA0"

- iA0- = 0 (8.1)
Expanding in a Taylor's series and neglecting higher-power terms in
(assumption 3), we get

l + n-.G7--.AG) + ROAIQJ + «-.G7"'-.AG)

- JW.Gz - i-.Gr'-iAG) - RoMQl. - n- = 0
(8.2)
 Method of Balancing Heads 187

Rearranging terms, we get

1 U
~ Rorn.Q;-1 + ^ - n ' , 0 - ' + i^-n-ifiT 1 + iC-n-ifiE" 1
(8.3)

Note that in Equation (8.3) the numerator gives the algebraic sum of the
head losses in pipes along the loop, the head loss being positive for pipes
/ and j and negative for pipes k and m. The denominator, however, is the
sum of all positive terms because when the direction of flow in a pipe is
negative, the correction is also negative. Therefore, Equation (8.3) can be
written as

E R- • i<2"
AQ = - ^ r 7 (8-4)
n • R^ • iQ"'1

Generalizing Equation (8.4) for all loops of a network, the correction

after the t'h iteration is given by

• ,Ql
, c = I, H . . . C (8.5)
» • * « • 'Q"'1

As each term of the numerator represents the head loss in a pipe, Equa-
tion (8.5) can also be written in terms of the head losses in pipes as

EA
, c = I, II . . . C (8.6)
n • ,hx

In the procedure originally proposed by Hardy Cross, after calculating

the correction for a loop, the discharges in the pipes of the loop are imme-
diately adjusted and later another loop is considered. Thus, the loop cor-
rections are calculated and applied sequentially until all loops are covered.
However, in current practice, the correction terms for all loops are
calculated initially and later applied simultaneously to complete one itera-
tion. Liu [102] has termed this simultaneous approach as the Newton-Cross
method because it considers all loops simultaneously as in Newton's
method (explained in next chapter) for the application of the corrections,
188 HARDY CROSS METHOD

but considers each loop separately for the calculation of the corrections as
in the Hardy Cross method. However, the simultaneous approach which is
more popular than the sequential approach continues to be known as the
Hardy Cross method and is also so explained in textbooks of fluid
mechanics, hydraulics, and water supply engineering, for example, refer-
ences [9,82]. Herein also the simultaneous approach is considered and
termed the Hardy Cross method. Liu has also shown that convergence is
faster, and therefore, fewer iterations are required in the simultaneous ap-
proach than in the sequential approach.
The approach based on AQ equations tries to make the algebraic sum of
the head losses along loops equal to zero. In other words, it tries to balance
the nodal heads obtained from different paths. For example, when the loop
of Figure 8.1 is balanced, the head obtained at node B from that at node
A by considering the path consisting of pipes / and j must be the same as
that obtained by considering the path consisting of pipes m and k. There-
fore, this approach based on AQ equations is known as the method of bal-
ancing heads.
The Hardy Cross method of balancing heads is based on the three
assumptions cited earlier. The third assumption that neglects the higher-
power AQ terms in Taylor's expansion is tested in Example 8.1. The net-
work consists of a single loop so that the effect of the first two assumptions
is absent.

Example 8.1

In the one-source, three-demand node network of Figure: Example 8.1,

node, \0-4

Hol = 100.00 m, Rol = 8, Ro2 = 10, Ro3 = 20, Ro4 = 15, qo2 = 0.2
m3/s, qo3 = 0.4 m3/s, and qo4 = 0.3 nrVs. Determine the pipe discharges
and the heads at demand nodes if the head loss in a pipe is given by
h = RQ2.
 Method of Balancing Heads 189

Solution

Assume the first-iteration pipe discharges as the following: XQ\ = 0,

iQ2 = 0.9 m3/s, ,03 = 0.7 nvVs, ,£>4 = 0.3 m3/s. (These unlikely values
are deliberately chosen to study the effect of assumption 3.) Note that the
assumed values satisfy the node-flow continuity equations at nodes 2, 3,
and 4. The solution is obtained in a tabular form as shown in Table: Exam-
ple 8.1. At the end of iteration 4 the error is negligible and obtained
answers are g , = 0.4933 rnVs, Q2 = 0.4067 nvVs, Q3 = 0.2067 m 3 /s,
and Q4 = —0.1933 m3/s along the directions shown in the figure. Since
Q4 is negative, the direction of flow in pipe 4 is from node 4 to node 3 and
not from node 3 to node 4 as initially assumed. Further, qx = 0.9 m 3 /s;
H2 = 100.00 - 1.654 = 98.346 m; H3 = 98.346 - 0.855 = 97.491
m; and H4 = 100.00 - 1.947 = 98.053 m.
From this single-loop example it is observed that:
(1) The initially-balanced, node-flow continuity equations remain bal-
anced, even after applying the AQ corrections, throughout the itera-
tive procedure.
(2) The assumption of neglecting higher-power terms of AQ is
justifiable. Even for the unlikely assumed discharges in pipes, only
four iterations are necessary for the convergence.
(3) The AQ corrections at the end of iterations 1 . . . 4 are - 0 . 3 5 ,
-0.1389, -0.0044, and -0.00004 nrVs, respectively. Thus, the rate
of convergence accelerates.

8.2.2 SINGLE-SOURCE NETWORKS WITH KNOWN

PIPE RESISTANCES

The loop corrections are initially calculated using Equation (8.4) for all
loops of a network and are then simultaneously applied to complete one
iteration. When a pipe is common to more than one loop, the corrections
of all these loops are applied to the discharge in the pipe to obtain revised
discharge for the next iteration. The iterative procedure is continued until
satisfactory convergence is reached.

Example 8.2

Analyze the single-source, three-demand node, two-loop network

shown in Figure: Example 8.2. The head loss in a pipe is given by h =
RQ1852.
 TABLE: EXAMPLE 8.1.
(Clockwise Positive, Anticlockwise Negative.)

Pipe,
X ,0, ,AO

Iteration 1

1 8 0.0 -0.00 0.00

19.25
2 10 0.9 8.10 18.00
55.00
3 20 0.7 9.80 28.00
= -0.35
4 15 0.3 1.35 9.00
L 19.25 55.00

Iteration 2

1 8 0.35 - 0.9800 5.60 2 AQ

4.4575
2 10 0.55 3.0250 11.00
32.10
3 20 0.35 2.4500 14.00
= -0.1389
4 15 -0.05 a -0.0375 1.50
L 4.4575 32.10

Iteration 3

1 8 0.4889 -1.9122 7.8224 3AQ

0.1339
2 10 0.4111 1.6900 8.2220
30.1554
3 20 0.2111 0.8913 8.4440
= -0.0044
4 15 -0.1889 a -0.5352 5.6670
L 0.1339 30.1554

Iteration 4

1 8 0.4933 -1.9468 7.8928 4AO

0.0012
2 10 0.4067 1.6540 8.1340
30.0938
3 20 0.2067 0.8545 8.2680
= -0.00004
4 15 -0.1933 8 -0.5605 5.7990 - Q
L 0.0012 30.0938

"Direction of flow is opposite to the assumed one.

 Method of Balancing Heads 191

q o2 =0-8nrr 3 /s

Hoi=100-00m

Solution

Let the assumed direction and the magnitude of discharges in the pipes
be as shown in the figure. The analysis is carried out as shown in Table:
Example 8.2 in which the first two iterations are given. Convergence is
reached in five iterations and the solution is as follows: Qt = 1.6745 m3/s;
Q2 = 1.3255 m 3 /s; Q3 = -0.2216 nrVs; Q4 = 0.6529 nrVs; Qs =
0.6471 nrVs; q, = 3.0000 m 3 /s; H2 = 87.010 m; H3 = 84.285 m; and
H4 = 86.518 m.
Note that as pipe 3 is common to both loops, loops I and II, both the cor-
rections AQi and AQU are applied to the discharge in it. Convergence is
slightly delayed because the effect of loop II on loop I and that of loop I
on loop II are neglected. In general, the number of iterations increases as
the size of the network increases because of the interaction among the
loops.

8.2.3 MULTISOURCE NETWORKS WITH KNOWN PIPE RESISTANCES

Because the network has multiple sources, pseudo loops, one less than
the number of sources, are introduced. The loop corrections are calculated
for the real loops using Equation (8.5). Applying the procedure used in
deriving Equation (8.5) to the pseudo loop consisting of starting node S,
end node E, and real and pseudo pipes as shown in Figure 8.2, the correc-
tion term for the fh iteration of this pseudo loop is given by

-Q" + HF - Hs
xeP
(8.7)
,Qr
 TABLE: EXAMPLE 8.2.
(Clockwise Positive, Anticlockwise Negative.)

Loop I Loop II

Pipe 0 floO1852 1.852fl 0 Q° 852

AO, Pipe 0 RoQ1852 1.852fl0Q°852 AQ,,
Ro

Iteration 1

1 5 1.8 14.850 15.279 3 8 0.1 0.112 2.083

2 8 1.2 -11.213 17.305 -0.102 4 6 1.1 7.158 12.052 -0.426
3 8 0.1 -0.112 2.083 5 5 -0.2 -0.254 2.350
£ 3.525 34.667 E 7.016 16.485

Iteration 2

1 5 1.698 13.330 14.538 3 8 -0.224 -0.501 4.141

2 8 1.302 -13.042 18.551 -0.021 4 6 0.674 2.890 7.940 -0.016
3 8 - 0.224 0.501 4.141 5 5 - 0.626 -2.100 6.213
£ 0.789 37.230 £ 0.289 18.294
 Method of Balancing Heads 193

Eo- 1

pseudo loop, p

Figure 8.2 Pseudo loop in distribution network.

in which HE, Hs = heads at end node and starting node, respectively, and
subscript p denotes a pseudo loop. The term HE — Hs can be looked upon
as the head loss in the pseudo pipe and therefore Equation (8.5) can be
used in place of Equation (8.7). The head loss in the pseudo pipe remains
fixed throughout the iterative procedure. Note that the denominator con-
tains the terms for the real pipes only and no term for the pseudo pipe.

Example 8.3

Analyze the two-source, four-demand node network shown in Figure 7.5

of Chapter 7 shown here again as Figure: Example 8.3. The head loss is
given by h = RQ1*5.
demand,0-2mys

Solution

Assume the first-trial pipe discharges as ,(?, = 0.35 m 3 /s, tQ2 = 0.45
m /s, ,03 = 0.05 m 3 /s, ,Q 4 = 0.05 m 3 /s, XQS = 0.30 m 3 /s, tQ6 = 0.20
3

m 3 /s, and tQ7 = 0.10 nvVs along the directions shown in the figure. The
first two iterations are shown in Table: Example 8.3. Note that pipe 4 is
 TABLE: EXAMPLE 8.3.
(Clockwise Positive, Anticlockwise Negative.)

Loop I Loop II Loop III

Pipe Ro Q R00185 1.85R0Q°<»s Pipe Ro R0Qi*s 1.85 R O Q° 85 Pipe R .85 R D Q° 85

Iteration 1

1 40 0.35 -5.736 30.32 4 30 0.05 0.118 4.35 1 40 0.35 5.736 30.32

2 20 0.45 4.565 18.77 5 40 0.30 4.313 26.59 3 10 0.05 0.039 1.45
3 10 0.05 -0.039 1.45 6 20 0.20 -1.018 9.42 Pseudo _. — -5.000 3 —
4 30 0.05 -0.118 4.35 7 50 0.10 -0.706 13.07 pipe
L - 1.328 54.89 2.707 53.43 S 0.775 31.77
iAQ, = 0.024 iAQ M = -0.051 ,AQm = -0.024

Iteration 2

1 40 0.302 -4.366 26.74 4 30 -0.025 -0.033 2.41 1 40 0.302 4.366 26.74

2 20 0.474 5.026 19.62 5 40 0.249 3.055 22.70 3 10 0.002 0.000 0.09
3 10 0.002 - 0.000 0.09 6 20 0.251 -1.550 11.43 Pseudo — — -5.000 a —
4 30 -0.025 0.033 2.41 7 50 0.151 -1.514 18.55 pipe
L 0.693 48.86 E -0.042 55.09 r -0.634 26.83
aAQ, = -0.014 2AOM = 0.001 2 AQ M I = 0.024
°H2-H,.
 Method of Balancing Heads 195

common to loops I and II, while pipes 1 and 3 are common to loop I and
pseudo loop III. HE - Hs = 95.00 - 100.00 = -5.00 m. After several
iterations the solution is <2i = 0.3246 m3/s, Q2 = 0.4715 m3/s, Q3 =
0.0246 m3/s, Q4 = -0.0219 m3/s, Q5 = 0.2496 m'/s, Q6 = 0.2503 nVVs,
Q7 = 0.1504 m 3 /s, H3 = 95.010 m, HA = 95.025 m, H5 = 91.956 m,
H6 = 93.459 m, q, = 0.7961 m3/s, and q2 = 0.2039 m3/s.

8.2.4 NETWORKS WITH UNKNOWN PIPE RESISTANCES [96-98]

This method of balancing heads for the analysis of networks with un-
known pipe resistances solves either the AQ-AR equations or the reduced
AQ equations. (The formulation of both these types of equations is ex-
plained in Chapter 7.) When the approach is based on AQ-AR equations,
the assumed unknown pipe resistances are also simultaneously adjusted
along with the assumed pipe discharges, and the method is implicit. When
the method is based on the reduced AQ equations, only the pipe discharges
are assumed and adjusted. After obtaining satisfactory convergence, the
unknown pipe resistances are calculated in the end. This approach is ex-
plicit.

8.2.4.1 Implicit Approach

Consider a single-looped network shown in Figure 8.3. Let A be the

source node with known head H^ and the other nodes as demand nodes
with known nodal flows. Let pipe u have unknown resistance /?„. To com-
pensate for the additional unknown Ru and thereby satisfy the unknown-
number rule for network solvability, the head at one of the demand nodes
must be known. Assume that the head at node B is known and it is HoB.
(As per the definition given in Chapter 6, node B is now a source node and

r-,-4
H OA
'0,

RL
IQU
pseudo loop

Figure 8.3 Loop having pipe with unknown resistance.

196 HARDY CROSS METHOD

the network becomes a two-source network as shown in Figure 8.3.) Let

iQx,x = i, j , k, m, and u be the assumed pipe discharges satisfying the
node-flow continuity equations and tRu be the assumed resistance for pipe
u. Let iAQ (clockwise positive) and iARu (additive) be the corrections to
the assumed pipe discharges ^QX and assumed pipe resistance tft,, respec-
tively. Following the procedure outlined earlier in this chapter, we get

AQ - - ^ x = i,j, k, m
L, I n- Rox- .QT 1 | + | n • ,ft, • .QT 1 |
(8.8)

If it is assumed for simplicity that the suffixing subscript o also includes

the assumed value of the resistance for pipe u, Equation (8.8) can be gener-
alized for loops, each having a pipe of unknown resistance, and therefore,
for the fh iteration the loop correction ,AQC for loop c is given by Equation
(8.5) in which Rox = ,/?, for pipes with unknown resistances.
Next consider the pseudo loop formed by pipes k, u, m, and the pseudo
pipe shown by broken line in Figure 8.3. The head loss equation for this
loop is

+ RoMQrr, + ,AG)"

+ HoB - HoA = 0 (8.9)

Equation (8.9) is used for adjusting the assumed value of pipe resis-
tance, i.e., iRu. Therefore correction ,A/?U is retained and correction ,A(?
is disregarded. Therefore, Equation (8.9) becomes

It* • iQl + Ron, • ,(£, + (,/?„ + tARH) • XQL + HoB - HoA = 0 (8.10)

Replacing ,/?„ by Rou for simplicity as before, and remembering that

HoB — HoA represents the head loss in the pseudo pipe, Equation (8.10)
can be rearranged, and the correction term can be expressed as

.Aft, - -^—^ (8-1D

in which the subscript x also includes the pseudo pipe, the head loss in
which is HoB — HoA and remains fixed throughout. Thus, the numerator
represents the algebraic sum of the head losses in all pipes (including the
 Method of Balancing Heads 197

pseudo pipe) of the pseudo loop. Note that the numerator may be positive
or negative and so also the denominator depending upon the direction of
flow in pipe u.
Generalizing Equation (8.11) for all pseudo loops used to adjust un-
known pipe resistances, Ru, u = 1 . . . U, the correction for the t'h itera-
tion is given by

,ARU = - -zr ,u = l...U (8.12)

Thus, the adjustment of the pipe discharges is made through Equation

(8.5) for loops used to adjust pipe discharges, and the adjustment of the
pipe resistances is made through Equation (8.12) for loops used to adjust
pipe resistances.

8.2.4.2 Explicit Approach

As explained in Chapter 7, it is possible to reduce X — JVM number of

AQ-AR equations to X — Jm — XVR number of AQ equations. There-
fore, in the explicit approach X — Jm — XVR independent real and
pseudo, basic and overlapping loops are formed such that all pipes of un-
known resistances are excluded. The pipe discharges are then assumed and
iteratively corrected using Equation (8.5) until satisfactory convergence is
reached. The unknown pipe resistances are then calculated using loops,
each containing a pipe with unknown resistance.

Example 8.4

The head loss in a pipe of the network shown in Figure: Example 8.4(a)
is given by h = RQ2. Analyze the network by head-balancing procedure
using (1) implicit and (2) explicit approaches.

Solution

The unknown parameters are qx, H3, H4, R4 and Hs. All the network
solvability rules are checked and found to be satisfied.
1. Implicit Approach. For this network, X = 6 and / UH = 3. Thus,
% ~ -AJH = 6 - 3, i.e., three independent loops are available. Out of
these three loops, X lR , i.e., one loop will be used to adjust the assumed
R4 value and the remaining two loops to adjust the pipe discharges.
Herein, the two basic loops, loops I and II in Figure: Example 8.4(b) are
198 HARDY CROSS METHOD

0-5 q

node,
100-00m'

Loop, I 3 3 II 2
(b)

Loop,I

used to adjust the pipe discharges and the third loop, loop III is used to ad-
just the unknown resistance R4. Let the assumed pipe discharges be ,Q, =
0.7 m3/s, ,Q 2 = 0.5 m3/s, , & = 0.2 nrVs, ,Q4 = 0.4 nrVs, tQ5 = 0.3
m3/s, and ,Q6 = 0.3 m3/s. Let the assumed pipe resistance be ,V?4 = 20.
The first two iterations of the iterative procedure are given in Table 1:
Example 8.4. Continuing the procedure for several iterations satisfactory
convergence is reached. The solution is: Qt = 0.6851 m 3 /s; Q2 = 0.5149
m 3 /s; Q3 = 0.2003 m 3 /s; Q4 = 0.3854 m 3 /s; Q5 = 0.3146 m 3 /s; Q6 =
0.2854 m 3 /s; R4 = 14.85; q, = 1.2 mVs; H3 = 97.349 m 3 /s; H4 =
96.245 m; and // 5 = 94.040 m.
2. Explicit Approach. In the explicit approach, two loops, loops I and II,
shown in Figure: Example 8.4(c) are considered for AQ corrections, and
the calculations are shown in Table 2: Example 8.4. As nodal flow at node
2 is known and fixed, when the correction AQ]U is applied to pipes 2 and
6 of loop II, the node-flow continuity equation at node 2 is violated.
Therefore, to maintain the node-flow continuity relationship at node 2 and
also at nodes 5, 4, and 1, the correction AQn is extended to pipes 5, 4,
and 1. Therefore, for iteration 2, 2Qi = 0.7 - 0.010 (correction for loop
 TABLE 1: EXAMPLE 8.4.
(Implicit Approach.)
Loop ! Loop II Loop
2 2
Pipe 0 RO 2RQ Pipe RO 2RQ Pipe Q RQ2 Q2

Iteration 1

8 0.7 3.92 11.2 27.5 0.2 1.10 11.0 1 8 0.7 3.92

10 0.5 2.50 10.0 20 0.4 3.20 16.0 4 20 0.4 3.20 0.16
27.5 0.2 1.10 11.0 40 0.3 -3.60 24.0 5 40 0.3 -3.60
8 0.3 0.72 4.8 Pseudo -2.00
pipe
0.32 32.2 1.42 55.8 1.52
0.32 1.42 1.52
iAO, = - - -0.010 ,AO, =
32.2 55.8 = -0.025 0.16 - -9.5
Iteration 2

1 8 0.690 3.809 11.04 3 27.5 0.185 0.941 10.18 1 8 0.690 3.809 —

2 10 0.510 -2.601 10.20 4 10.5 0.375 1.477 7.88 4 10.5 0.375 1.477 0.1406
3 27.5 0.185 -0.941 10.18 5 40 0.325 4.225 26.00 5 40 0.325 -4.225
6 8 0.275 0.605 4.40 Pseudo — — - 2.000
pipe
X. 0.267 31.42 D 1.202 48.46 r -0.939

0.267 -1.202 -0.939

31.42 48.46 0.1406
200 HARDY CROSS METHOD

TABLE 2: EXAMPLE 8.4.

(Explicit Approach.)

Loop I Loop II

Pipe Ro Q 2R0Q Pipe 2R0Q

Iteration 1

1 8 0.7 3.92 11.2 6 8 0.3 0.72 4.8

2 10 0.5 -2.50 10.0 2 10 0.5 -2.50 10.0
3 27.5 0.2 -1.10 11.0 Pseudo — — 2.00 —
pipe
L 0.32 32.2 £ 0.22 14.8

0.32 0.22
= -0.010 1A = -0.015
32.2 14.8

Iteration 2

8 0.675 3.645 10.80 6 8 0.285 0.650 4.56

10 0.525 -2.756 10.50 2 10 0.525 -2.756 10.50
27.5 0.210 -1.213 11.55 Pseudo — 2.000 -
pipe
L -0.324 32.85 -0.106 15.06

- 0.324 -0.106
2 AQ. = - 0.010 2AQ,, = - = 0.007
32.85 15.06

I) — 0.015 (correction for loop II) = 0.675 m3/s. Thus, loop II compris-
ing real pipes 2 and 6 is used to calculate AQn but the correction is ap-
plied to the overlapping loop comprising pipes 2, 6, 5, 4 and 1. The itera-
tive procedure is continued until satisfactory convergence is reached. In
the end, loop III is used to calculate the value of ft, and the final solution
as given earlier is obtained.

Example 8.5

In the network of Example 8.4, R4 is given as 14.85 while R3 is un-

known. Using other data as in Example 8.4, analyze the network by im-
plicit and explicit approaches-
Solution

1. Implicit Approach. The selected three loops are shown in Figure: Ex-
ample 8.5(a). Loops I and II are the same as loops I and II of Figure: Ex-
ample 8.4(b), respectively. Further, as in Example 8.4, these two loops are
 Method of Balancing Heads 201

used to adjust pipe discharges. Because the resistance of pipe 3 is un-

known, loop III is chosen, as shown in the figure, to adjust the assumed
R3 value. Herein, as all the three loops contain pipe 3 of unknown resis-
tance, any two of them can be used to adjust pipe discharges and the third
to adjust the value of R3. Following the iterative procedure, the solution
given in Example 8.4, with R3 = 27.5 is obtained.
2. Explicit Approach. In the explicit approach pipe 3 with unknown re-
sistance should be excluded from the loops. Both the real loops contain
pipe 3. However, two pseudo loops are available: loop I comprising pipes
1, 4 and 5 and loop II comprising pipes 2 and 6, as shown in Figure: Ex-
ample 8.5(b). If these loops are considered and corrections are applied, it
is observed that the discharge in pipe 3 remains uncorrected and at the
same time node-flow continuity is violated at nodes 1 and 2. Therefore, the
loops shown in Figure: Example 8.5(c) are selected for the following net-
work analysis:
(1) Loop I with correction AQ,. The correction is obtained from the

Loop I
202 HARDY CROSS METHOD

pseudo loop I but it is applied to pipes 1, 4, 5, 6, and 2 to satisfy the

node-flow continuity relationship at nodes 1 and 2. (Alternatively,
overlapping loop comprising pipes 1, 4, 5, 6, and 2 can be consid-
ered for evaluating and applying the Ag, correction.)
(2) Loop II with correction AQn. The correction will be applied to all
the pipes of this loop. This will satisfy the node-flow continuity rela-
tionship. However AQU cannot be obtained from this loop as it con-
tains pipe 3 with unknown resistance.
(3) Loop HI to determine correction AQU

According to this procedure the AQ equations for the given and assumed
values are

8(0.7 + A0, + AQuf + 14.95(0.4 + A0,) 2

- 40(0.3 - AQd2 + 98.00 - 100.00 = 0 (1)

and

8(0.3 + AQ,)2 - 10(0.5 - A<2, - AQn)2 + 100.00 - 98.00 = 0 (2)

Neglecting AQn in Equation (1) and AQ, in Equation (2), we get

8(0.7 + Afi) 2 + 14.95(0.4 + A&) 2 - 40(0.3 - A0,) 2 - 2 = 0 (3)

8(0.3) 2 - 10(0.5 - A0,,)2 + 2 = 0 (4)

Equations (3) and (4) are solved iteratively, and the first two iterations
are given in Table: Example 8.5. Note that the correction AQ, is applied
to pipes 1, 4, 5, 6, and 2, while AQn is applied to pipes 1, 2, and 3. Fur-
ther, AQn is obtained from loop III in which the RQ2 values are taken for
pipes 6 and 2 while the IRQ value is taken for pipe 2 only as it is the pipe
common to loops II and in.
8.2.5 NETWORKS WITH PUMPS

The AQ equations for the real or pseudo loops contain hp and therefore
an expression of Qp as explained in Chapter 7. The balancing of heads for
such loops is carried out as shown in the following example.
Example 8.6

Analyze the network shown in Figure: Example 8.6 using the Darcy-
Weisbach formula for the head loss in pipes and the Swamee-Jain explicit
 Method of Balancing Heads 203

TABLE: EXAMPLE 8.5.

(Explicit Approach.)

Loop I Loop III

2
Pipe Flo 0 R0Q2 2ROO Pipe Ro O f ?oO 2R0Q

Iteration 1

1 8 0.7 3.920 11.20 6 8 0.3 0.72

4 14.85 0.4 2.376 11.88 2 10 0.5 2.50 10.00
5 40 0.3 -3.600 24.00 Pseudo — 2.00
Pseudo -2.000 pipe
pipe
E 0.696 47.08 E 0.22 10.00

0.696 0.22
iAQ, = - = -0.015 iAON = -0.022
47.08 10

Iteration 2

1 8 0.663 3.517 10.61 6 8 0.285 0.650

4 14.85 0.385 2.201 11.43 2 10 0.537 - 2.884 10.74
5 40 0.315 -3.969 25.20 Pseudo — 2.000
Pseudo -2.000 pipe
pipe
E -0.251 47.24 E - 0.234 10.74

-0.251 -0.234
2AQ,, = - 0.022
47.24 10.74

relationship for the friction coefficient. The pipe details are as given
in Table 1: Example 8.6. Three points on the head-discharge curve for
the pump have the following discharge to head characteristics: 0.1
nvVs - 11.89 m; 0.2 nWs - 11.01 m; and 0.3 m3/s - 9.37 m, respec-
tively.

Sump E.S.R.
W.L? 80-00 m W.LrlOO-OOm
204 HARDY CROSS METHOD

TABLE 1: EXAMPLE 8.6.

-
Roughness
Length, Diameter, Height,
Pipe m mm mm
1 300 300 0.20
2 400 200 0.25
3 400 150 0.25
4 150 150 0.25
5 150 200 0.25
6 150 150 0.30
7 500 350 0.20

Solution

Using the pump data, the head-discharge equation for the pump is ob-
tained as

hp = 12 + 2.7 Qp - 38.2 Q\ (1)

Transforming the pump equation according to Equations (5.6) and (5.5),

we get, respectively,

hp = 12.05 - 38.2 (Pp (2)

and

Gp = Qp - 0.0353 (3)

The first-trial discharges in pipes satisfying the node-flow continuity

equations at nodes 3 . . . 6 are assumed as follows: ,£), = 0.22 m3/s,
,Q2 = 0.08 m3/s, . f t = 0.06 m3/s, , £ 4 = 0.02 nWs, ,<25 = 0.06 m3/s,
\Qb = 0.04 m3/s, and ig 7 = 0.22 nvVs. The first iteration is shown in
Table 2: Example 8.6.
Because the Darcy-Weisbach head loss formula is used, the friction
coefficient / and therefore the resistance constant R of the pipes change
from iteration to iteration. This increases the number of iterations for
convergence. However, t h e / a n d R values stabilize in the initial iterations
and practically remain constant in later iterations. The final solution is
as follows: / , = 0.01852, /?, = 189.0, 0 , = 0.1619 m 3 /s; f2 = 0.02167,
R2 = 2238, Q2 = 0.0508 m 3 /s; f3 = 0.02325, R3 = 10123, Q3 = 0.0311
m 3 /s;/ 4 = 0.02322, R4 = 3791, Q4 = 0.0324 m 3 /s;/ 5 = 0.02125, R5 =
823.3, 0 5 = 0.1016 m 3 /s;/ 6 = 0.02392, R6 = 3905, Q6 = 0.0565 nWs;
f7 = 0.01776, R7 = 139.7, Q-, = 0.2781 m 3 /s; and hp = 11.44 m.
 Method of Balancing Heads 205

If the head-discharge relationship given by Equation (5.7) is selected to

represent the pump data, and taking n = 2 since the DW formula is used,
we get

hp = 12.18 - 2 9 . 3 0 for 0.1 < Qp < 0.2 (4)

and

h, = 12.32 - 3 2 . 8 0 for 0.2 < Qp < 0.3 (5)

For iteration I, as !<2i = 1 0 = 0.22 m3/s, we use Equation (5). The

pump can be made an integral part of pipe 1 with modified resistance con-
stant = 187 + 32.8 = 219.8. The head at the sump becomes 80 +
12.32 = 92.32 m. Subsequently, when <2i becomes less than 0.2 m3/s, we
use Equation (4). To include the pump effect, the pipe resistance constant
is increased by 29.3, and the sump head becomes 92.18 m. This gives prac-
tically the same final solution.
The same procedure can be followed for the HW-formula, but the value
of n is 1.852 in Equation (5.7).

TABLE 2: EXAMPLE 8.6.

Loop
c Pipe x iAO

2 0.08 0.0214 2207 14.13 353.2

1 4 0.02 0.0237 3865 1.55 154.6 0.0118
3 0.06 0.0229 9947 -35.81 1193.7
L -20.13 1701.5

4 - 1.55 154.6
II 5 0.06 0.0215 835 -3.00 100.2 -0.0031
6 0.04 0.0241 3933 6.29 314.7
r 1.74 569.5

7 0.22 0.0179 141 6.81 61.9

5 — — ... 3.00 100.2
2 _ _ - -14.13 353.2
III 1 0.22 0.0184 187 -9.07 82.46 00370
Pump 0.18473 — 38.2° 10.75' 14.11°
Pseudo — — -20.00 -
pipe
-22.64 611.9

"Coefficient of ,Gf., i.e.. , f l p .

206 HARDY CROSS METHOD

8.2.6 NETWORKS WITH VALVES

The formulation of the AQ equations for networks with check valves and
pressure reducing valves has been explained in Chapter 7. The loops are
selected such that these valves are externalized.
For check valves the correction is restricted so that the discharge in the
pipe containing a check valve does not change direction.
When a PRV is present in a network, it is initially presumed that it is in
operative condition. An appropriate pseudo loop is formed for calculating
the AQ correction and is applied to the loop such that flow continuity is
maintained. The procedure is explained in Example 8.7.

Example 8.7

Analyze the network shown in Figure: Example 8.7(a). The R values for

5 0-4

100-00 m,
a -1

Loop I

the pipes are as given in the figure. Pipe 7 having diameter of 350 mm con-
tains a PRV, just downstream of node 3 with a set pressure head of 80.00
m. The head loss in pipes is given by h = RQ2.

Solution

It is presumed that the PRV is in an operative condition and is, therefore,

replaced by source and sink nodes. (The sink node is not shown separately
and is merged with node 3.) Thus, the pressure head at the downstream
end d of the PRV is 80.00 m. Real loop I and pseudo loop II are selected
as shown in Figure: Example 8.7(b).
 Method of Balancing Heads 207

As stated in Chapter 5, the resistance constant of the PRV, i.e., Rfm =

(8£pRv)/7T2gD?. Taking KPW = 10, and as Dx = 0.35 m, /?PRV = 55.06.
The pipe discharges are assumed as shown in Table: Example 8.7. Con-
sidering the path from the source node 1 to the downstream end d of the
PRV, i.e., path 1-2-3-d, the head at d, i.e., Hd = 100.00 - 2 x 2 2 -
4 X 0.52 - 55.06 X O.I2 = 90.45 m > 80.00 m. Therefore, the
assumption that the PRV is operative is correct for the assumed pipe dis-
charges.
The iA(2 values are obtained as shown in the table. The correction
tAQi is applied to loop I, while the correction iAQ,, is applied to loop III
to maintain flow continuity at the nodes. For the corrected pipe discharges
Hd is calculated to verify that the PRV remains in the operative condition.
The iterative procedure is continued and the final solution is Qt = 2.00
nVVs, Q2 = 0.9959 m3/s, Q3 = 0.7041 m3/s, Q4 = 0.3149 m 3 /s, Q5 =
0.8036 m3/s, Qb = 0.4108 m7s, and Q7 = 0.1892 m 3 /s.
The head at the downstream of the PRV = 88.05 m > 80 m, and there-
fore the PRV is operative and reduces the downstream pressure from 88.05
to 80 m.
If the pressure setting of the PRV had been 89 m, even though the PRV
is initially operative, during the iterative procedure Hd becomes less than
89 m and therefore the PRV becomes inoperative. At this stage loop III re-
places loop II and therefore the corrections are calculated and applied to
loops I and III. Because of the PRV, the resistance constant of pipe 7
becomes 50 + 55.06 = 105.06. The final solution is 0 , = 2.00 mVs,
0 2 = 0.9393 mVs, Q3 = 0.7607 nWs, Q4 = 0.2728 m3/s, Qs = 0.7121
nv»/s, Q6 = 0.3121 m3/s, Q7 = 0.2879 m3/s, and Hd = 85.12 m < 89 m,
the setting of the PRV.

TABLE- EXAMPLE 8.7.

Loop I Loop I

Pipe Ro iO 2R 0 -,C 3 Pipe V1Q2 2fV,Q

2 6 1.2 8.64 14.4 1a 2 2.0 8.00 8.0
3 4 0.5 -1.00 4.0 2 6 1.2 8.64 14.4
4 40 0.2 -1.60 16.0 5 9 0.9 7.29 16.2
y
6.04 34.4 6 12 0.5 3.00 12.0
7 50 0.1 -0.50 10.0
Pseudo -20.00
6.43 60.6

,AQ = -0.1756 ,AQ .1061

a f
As discharge in pipe 1 remains ixoa. the head loss in it .s also fixed. This gives H, - 92.0 m. There-
fore, pipe 1 could be omitted I'om oseudo loop II so that the pseudo loop cons^sts of pipes 2. 5, 6, 7.
and the pseudo o pe
208 HARDY CROSS METHOD

Assume that the nodal demands are 0.2, 1.5, 0.2, 0.2, and 0.1 m3/s at
nodes 2, 3, 4, 5, and 6, respectively. Even though initially it is assumed
that the PRV is operative and the flow direction in pipe 7 is from node 3
to node 6, during the iterative procedure, the correction changes the flow
direction in pipe 7. Herein the PRV acts as a check valve and prevents flow
from node 6 to node 3. The discharge in pipe 7 is zero, and the subsequent
correction in loop II (or loop III) is clockwise, and therefore it is restricted
to zero. Thus, the correction is applicable to loop I only. The final solution
is Ql = 2.2 nrVs, Q2 = 0.7595 m3/s, Q3 = 1.2405 m3/s, Q4 = -0.2595
m3/s, Q5 = 0.3 m3/s, Q6 = 0.1 nrVs, and Q7 = 0. H, = 84.16 m and
H6 = 85.93 m.

Example 8.8

Assume that in the network of Figure: Example 8.7, a PRV is also pro-
vided in pipe 5 (500 mm diameter) just downstream of node 4, as shown
in Figure: Example 8.8(a). The pressure-head settings of PRV-1 and PRV-2
are 60.00 m and 65.00 m, respectively. Analyze the network.

100-OOm 1
r

° ' 5 ^ je £°°9* U0-4

Solution

It is presumed that both the PRVs are operative and are replaced by
source and sink nodes as shown in Figure: Example 8.8(b). The PRVs have
separated nodes 5 and 6 from the main network. This separated network,
loop II, is analyzed and the pipe discharges are Q<, = 0.7847 m3/s,
 Method of Balancing Flows 209

Q6 = 0.3847 mVs, and Q7 = 0.2153 m°7s. Therefore, the flows through

PRV-1 and PRV-2 are 0.2153 m3/s and 0.7847 m'/s, respectively. Consider-
ing them as additional outflows at nodes 3 and 4, respectively, Loop I is
analyzed and the pipe discharges are Q2 = 0.9806 m'/s, Q3 = 0.7194
m3/s, Q4 = 0.3041 m3/s, and 0 , = 2 mVs.
The Hd-i and H^2 values are 87.38 m and 78.09 m, respectively. Since
these values are more than 60.00 m and 65.00 m, respectively, both the
PRVs are operative.

8.3 METHOD OF BALANCING FLOWS

8.3.1 BASIC CONCEPT

The method of balancing flows |101| solves the AH equations on the fol-
lowing assumptions:
(1) Only one equation is considered at a time from the available set of
AH equations.
(2) The effect of the adjacent nodes is neglected and therefore each AH
equation contains only one AH term as unknown. Thus, except the
correction for the node under consideration, corrections for other
connected nodes are assumed to be zero.
(3) Each term of the modified AH equation is expanded in a Taylor's
series and only the first-power AH terms are retained and the higher-
power AH terms are neglected.

Consider a part of a distribution network connected to node^, as shown

in Figure 8.4. The nodey, having known flow qoj, is connected to nodes a,
b, k, and m through pipes a, b, k, and m of known resistance constants Roa,
Rob, Rok and Rom, respectively. Let the assumed heads at nodes a, b, j ,
k, and m be ,//„, ,//,,, ,//,, iHk, and ,// m , respectively. Let ,//„ and ,//,, be
more than ,//,, and let xHk and ,//„ be less than ,//, so that the flow direc-
tions in the pipes are as shown in the figure. Let , A//, be the additive cor-

Figure 8.4 Part of distribution network connected to node./.

210 HARDY CROSS METHOD

rection to ,//,. Neglecting the corrections at connected nodes, the AH

equation is

[ Roc,

UHj + AHj) -
_ _
ft* J I ft- J (8.13)
tjoj —

or

[ 1HJ-1Hk lAHJV" [ , ]

(8.14)

Expanding Equation (8.14) in a Taylor's series, neglecting higher-power

terms in I A / / , , rearranging terms, and simplifying using Equation (6.6),
we get

AU l(2» + iQb — iQk — iQm — (joj ,„ ,

1 1 = {
TTn n n TT^
I / \ja \\lb \\lk [J
h h h
in which iQa, ,Qb . . . = discharges in pipes a, b . . . for the first itera-
tion and iha, ihb . . . = head losses in pipes a, b . . . for the first itera-
tion. The numerator in Equation (8.15) gives the excess of total inflow over
total outflow at nodey, for the first iteration, i.e., xFj. Generalizing Equa-
tion (8.15) for all the demand nodes, M + 1, . . . , M + N of the net-
work, the correction after the t'h iteration is given by

, ; = M + I,. . . ,M + N (8.16)

During the iterative procedure, when the head //, and that at a connected
node are equal, Qlnh becomes infinite for the pipe connecting them. In
this case, such a pipe is omitted for this iteration.
 Method of Balancing Flows 211

As this approach tries to make the algebraic sum of the flows at a node
equal to zero, i.e., it tries to balance the total inflow and total outflow, it
is known as the method of balancing flows.

8.3.2 NETWORKS WITH KNOWN PIPE RESISTANCES

The method of balancing flows does not require the introduction of

pseudo loops and therefore it is immaterial whether the network has one
or more source nodes (nodes with known heads). However, for a given
network, as the number of source nodes increases, the number of unknown
nodal heads and therefore the number of unknown basic parameters de-
creases.

Example 8.9

Solve Example 8.3 by the method of balancing flows.

Solution

Let the first-trial nodal heads be A = 96.00 m, , # 4 = 94.00 m,

iHs = 92.00 m, and ,// 6 = 93.00 m. The first iteration of the iterative
procedure is shown in Table: Example 8.9. Using the corrections, the
second-trial values are 2 // 3 = 94.46 m, 2 // 4 = 95.52 m, 2HS = 91.32 m,
and 2H6 = 93.47 m. The iterative procedure is continued and the solution
given in Example 8.3 is obtained.

8.3.3 NETWORKS WITH UNKNOWN PIPE RESISTANCES

The method of balancing flows for analysis of networks with unknown

pipe resistances solves either the AH-AR equations or the reduced AH
equations. The approach based on AH-AR equations is implicit and the
unknown pipe resistances are assumed and adjusted along with the un-
known nodal heads. The approach based on reduced AH equations is ex-
plicit and the unknown nodal heads are evaluated initially, and after attain-
ing convergence the unknown pipe resistances are evaluated in the end.

8.3.3.1 Implicit Approach

Considering the assignment rule for solvability of pipe networks, each

node with a known flow must have a basic H or R unknown assigned to it.
In the AH-AR equations formed at these nodes, the AH or AR term corre-
sponding to the assigned unknown is retained and the other AH and AR
terms are ignored. These reduced equations are then used to determine the
212 HARDY CROSS METHOD

TABLE: EXAMPLE 8.9.

(Inflow Positive; Outflow Negative.)

Assumed
Head, h,, Q, Correction,
Of
Node m Pipe m m3/s n • hij m

3 96.00 1-3 40 4 0.2880 0.0389 1&H3 = - 1 . 5 4

3-2 10 1 -0.2880 0.1557
-0.3a —
E -0.3 0.1946
4 94.00 1-4 20 6 0.5216 0.0470 , A H 4 = 1.52
2-4 30 1 0.1591 0,0860
4-5 40 2 -0.1980 0.0536
-0.2a —
E 0.2827 0.1866
5 92.00 4-5 40 2 0.1980 0.0536 iAH5 = - 0 . 6 8
6-5 50 1 0.1207 0.0653
-0.4a —
E -0.0813 0.1189
6 93.00 2-6 20 2 0.2880 0.0779 i A H 6 = 0.47
6-5 50 1 -0.1207 0.0653
-0.1a —
E 0.0673 0.1432

"External outflow.

retained AH or AR corrections. To understand this, consider Equations

(7.26b) . . . (7.26f) of Chapter 7. These are the AH-AR equations for
nodes 2 . . . 6 of known flow, respectively, of the network of Figure 7.6.
The unknowns assigned to nodes 2, 3, 4, 5, and 6 are as follows: H7 to
node 2, /?6 to node 3, H6 to node 4, H5 to node 5, and R2 to node 6. There-
fore, the basic unknown AH7 is retained in Equation (7.26b), A/?s in Equa-
tion (7.26c), AH6 in Equation (7.26d), AHS in Equation (7.26e) and AR2 in
Equation (7.26f). Other AH, AR unknowns in these equations are ignored.
Thus, each equation now has only one AH or AR as the basic unknown and
its value can then be determined.
Consider a part of a distribution network around node j of known flow
qoj as shown in Figure 8.5. As explained earlier, the unknown parameter
that may be assigned to this node7 is (1) unknown head at this node itself,
i.e., Hj\ (2) unknown head at a connected node, say /, i.e., //,; or (3) un-
known resistance of a connected pipe, Ru. Therefore, the basic unknown
retained in the AH-AR equation for node j is AHj, A//,, or A/?tf. Writing
the AH-AR equation for node j , and retaining only the A//y, A//,, or A/?tf
 Method of Balancing Flows 213

term corresponding to the unknown assigned to nodey, we can get, for the
general t'h iteration,

p.
,AHj = zz when unknown //, is assigned to node j
(8.17)
i connected
toy

,A//, = - when unknown //, is assigned to node j (8.18)

tQijin • Jitj

TJ TJ

,{R + AR)ij = i when unknown Ru is assigned to nodej

J lUij l
' (8.19)

in which ,Fj = residue of inflow over outflow at node j , for the t'h itera-
tion.
Equation (8.17) is the same as Equation (8.16) except that it includes a
term for the flow in a pipe of unknown resistance, assumed as f/?,v for the
f" iteration. Equation (8.18) gives the additive correction to the assumed
Hi value, for the fh iteration, while Equation (8.19) directly gives the Rtj
value for the (t + 1)" iteration. Equations (8.17) . . . (8.19) are used in the
iterative procedure to determine the unknown H and R parameters. Care
should be taken, during the iterative procedure, to use proper signs for the
various terms.

8.3.3.2 Explicit Approach

As explained in Chapter 7, it is possible to obtain JVH number of AH

equations for the assigned, single or merged nodes so that all pipes with
unknown resistances are eliminated. Therefore, the AH corrections are ob-
tained with the help of Equation (8.17) or Equation (8.18). For the illus-
trative network of Figure 7.6 of Chapter 7, Equations (7.26d), (7.26h), and

Figure 8.5 Part of distribution network having pipe with unknown resistance.
214 HARDY CROSS METHOD

(7.26i) provide three independent AH equations for the three basic un-
knowns, AH6, A// 5 , and AH7, respectively. Therefore, AHe is obtained at
node 4 using Equation (8.18); A// 5 is obtained at merged node 5-6 using
Equation (8.17); and AH7 is obtained at merged node 2-3 using Equation
(8.18).

Example 8.10

Solve Example 8.4 by the flow-balancing procedure using (1) implicit

and (2) explicit approaches.

Solution

1. Implicit Approach. The //and R unknowns in the network are H3, H4,
R4, and Hs. To satisfy the assignment rule, H3 is assigned to node 3, H4 to
node 4, R4 to node 5, and H5 to node 2. Therefore, AH3 will be obtained
at node 3, A// 4 at node 4, AR4 at node 5, and AHS at node 2. Considering
the assumed flow directions in pipes, shown in Figure: Example 8.4, let
the first-trial values be ,// 3 = 97.00 m, ,// 4 = 96.00 m, ,// 5 = 95.00 m,
and ,/?4 = 20. The first iteration of the iterative procedure is given in
Table 1: Example 8.10. The second-trial values obtained at the end of first
iteration are 2 // 3 = 97.30 m, 2H4 - 96.59 m, 2HS = 95.60 m, and
2R4 = 5.51. Continuing the procedure for several iterations, the solution
obtained in Example 8.4 is reached. The same solution is obtained if the
alternate assignment of H3 to node 2, H4 to node 3, R4 to node 4, and Hs
to node 5 is made.
2. Explicit Approach. Pipe 4 with unknown resistance R4 has nodes 4
and 5 as end nodes. Therefore these two nodes are merged and the total
outflow is 1.2 mVs. For the flow directions assumed in the figure, Pipes 1,
3, and 5 are the supply pipes to this merged node 4-5. The unknown
assignments are H3 to node 3, H4 to merged node 4-5, and Hs to node 2.
The first iteration of the iterative procedure is given in Table 2: Example
8.10. The first-trial values are the same as those assumed for the implicit
approach. The second-trial values obtained at the end of first iteration are
2H3 = 97.30 m, 2 // 4 = 95.88 m, and 2 // 5 = 95.60 m. After several itera-
tions, the solution given earlier is obtained.

8.3.4 NETWORKS WITH PUMPS

The following procedure is followed when a pump is present in a net-

work:
(1) Assume the first-trial values for the unknown nodal heads, and also
for the head delivered by the pump
 TABLE 1: EXAMPLE 8.10.

Assumed
Value of
Assigned Assigned Connected O
Node Unknown Unknown Pipe m m3/s n-hH Parameter

3 H3 97.00 m 1-3 10 3.00 0.5477 0.0913 AH§ = 0.304 m

3-4 27.5 1.00 -0.1907 0.0953
2-3 8 1.00 0.3536 0.1768
-0.6000 a
r 0.1106 0.3634
4 HA 96.00 m 1-4 8 4.00 0.7071 0.0884 AHS = 0.590 m
3-4 27.5 1.00 0.1907 0.0953
4-5 20 c 1.00 -0.2236. 0.1118
-0.5000 3
i; 0.1742 0.2955
5 RA 20 4-5 20<: 1.00 0.2236 0.1118 RA + ARa4 = 5.508
2-5 40 3.00 0.2739
-0.7000 a
£ -0.2025

CM CM
CO
2 H5 95.00 m 8 1.00 -0.3536 AHi - 0.602 m
3.00 -0.2739 0.0457
LO
40
0.6000e
E -0.0275
a
External outflow.
B
Using Equation (8.17).
c
Firsttrial assumed value.
d
Using Equat.on (8.19).
e
External inflow.
'Using Equation (8.18).
 TABLE 2: EXAMPLE 8.10.

Assumed
Value of
Assigned Connected
Node Assigned Unknown, Pipe QH AW,,
i Unknown m ij m m3/s n-h,. m

3 H3 97.00 1-3 10 3.00 0.5477 0.0913 AH3 = 0.30

3-4 27.5 1.00 -0.1907 0.0953
2-3 8 1.00 0.3536 0.1768
-0.6000 s
E 0.1106 0.3634
4-5 H4 96.00 1-4 8 4.00 0.7071 0.0884 AH 4 = -0.12
3-4 27.5 1.00 0.1907 0.0953
2-5 40 3.00 0.2739 0.0457
- 1.2000a
E - 0.0283 0.2294
2 95.00 2-3 8 1,00 -0.3536 AH5 = 0.60
2-5 40 3.00 -0.2739 0.0457
0.6000a
E -0.0275
a
External nodal flows—inflow positive, outflow negative.
 Method (/[Balancing Flows 217

(2) Carry out one iteration and obtain the pipe discharges and the AH
corrections
(3) Taking the discharge in the pipe in which the pump is situated as the
discharge delivered by the pump, find the head delivered by the
pump
(4) Update the unknown nodal heads
(5) Using the updated nodal heads (step 4) and the head delivered by the
pump (step 3), carry out the next iteration (step 2)
(6) Continue the iterative procedure until satisfactory convergence is
reached

Example 8.11

Solve Example 8.6 by the method of balancing flows.

Solution

Let the first-trial value for the head delivered by the pump, ,/?p be
10.75 m, a value corresponding to XQP — §.Tl rrvVs, assumed in Example
8.6. Considering the assumed flow directions in pipes (Figure: Example
8.6), let ^ 3 = 85.00 m; ,// 4 = 83.00 m; ,// 5 = 80.00 m; and ,// 6 =
90.00 m. The coefficient of friction is taken as 0.02 for all pipes, and the
corresponding R values are obtained.
The first iteration is shown in Table: Example 8.11: hi-3 = tHi +
lhp - tH3 = 80.00 + 10.75 - 85.00 = 5.75 m. Since <2,-3 obtained
after the first iteration is 0.1679 rrvVs, 2hp = 12 + 2.7 X 0.1679 -
38.2 X 0.16792 = 11.38 m. Using the obtained pipe discharges, the/and
R values of the pipes are updated, the heads for nodes 3 . . . 6 are cor-
rected, and second iteration is carried out.
After several iterations, the solution given in Example 8.6, with
H3 = 86.49 m; HA = 80.71 m; H, = 76.73 m; and H6 = 89.20 m is ob-
tained.
As stated in Example 8.6, if the head-discharge relationship given by
Equation (5.7) is used for the pump, the pump can be made an integral part
of the pipe in which it is situated. The sump level is increased by Hp and
the pipe resistance constant by Rp.

8.3.5 NETWORKS WITH VALVES

In the case of a check valve, as long as the flow in the pipe is in the
permissible direction, additional resistance offered by the check valve is
considered for the pipe in which it is present, and the normal iterative pro-
218 HARDY CROSS METHOD

TABLE: EXAMPLE 8.11.

Assumed
Node Value of Pipe On AH,,
/ Hr m ij m m3/s n-hH m
3 85.00 1-3 204 5.75 0.1679 0.0146 1.32
3-4 2066 2.00 -0.0311 0.0078
3-5 8707 5.00 -0.0240 0.0024
-0.0800 a -
E 0.0328 0.0248
4 83.00 3-4 _ — 0.0311 0.0078 -1.23
4-5 3265 3.00 -0.0303 0.0051
6-4 775 7.00 0.0950 0.0068
-0.1200 3 —
-0.0242 0.0197
5 80.00 3-5 — 0.0240 0.0024 -1.01
4-5 _ _ 0.0303 0.0051
6-5 3265 10.00 0.0553 0.0028
-0.1200 3 —
s -0.0104 0.0103
6 90.00 2-6 157 10.00 0.2524 0.0126 -0.81
6-4 — — - 0.0950 0.0068
6-5 - _ - 0.0553 0.0028
-0.1200a —
-0.0179 0.0222
a
External nodal flows—inflow positive, outf ow negative.

cedure is followed. However, if the flow direction reverses, the pipe con-
taining the check valve is excluded from the next iteration.
In the case of a pressure reducing valve, it is initially assumed that it is
in the operative mode and is therefore replaced by source and sink nodes.
The discharge in the pipe downstream of the PRV is determined, and this
discharge is taken as additional outflow at the upstream node of the pipe.
The iterative procedure is carried out and after adjusting the nodal heads,
a check is made at the end of each iteration to verify that the PRV has re-
mained in the operative mode. If the PRV becomes inoperative, the origi-
nal pipe is introduced in the iterative procedure but its resistance is in-
creased to include the additional resistance offered by the PRV. On the
other hand, after adjusting the nodal heads if it is observed that the head
at the downstream end of the pipe is more than that at the upstream end,
the PRV functions as a check valve and therefore the pipe is excluded from
the next iteration.
Thus, for check valves and PRVs it is necessary to ascertain their mode
of operation before the next iteration is carried out.
 Method of Balancing Flows 219

Example 8.12

Solve Example 8.7 by the method of balancing flows.

Solution

Considering the flow directions as shown in Figure: Example 8.7(a), let

t 2 = 97.00 m; ,//, = 95.00 m; ,// 4 = 90.00 m; ,// 5 = 85.00 m; and
H
,H6 - 75.00 m. As ,// 3 > //,., for the PRV, the PRV is assumed to be in
the operative mode.
The PRV is replaced by sink node u and source node d. The discharge
in pipe d-6 is [(80 - 75)/5O]05 = 0.3162 mJ/s. For this discharge, the
head downstream of the PRV is 95.00 - 55.06 X 0.31622 = 89.49
m > 80.00 m and therefore the assumption that the PRV is operative is
correct. Node 3 now has supply pipe 2-3, distribution pipe 3-4, and an
outflow of 0.2 + 0.3162, i.e., 0.5162 nvVs. The iterative procedure is car-
ried out and the final solution is H2 = 92.00 m, H3 = 90.02 m,
H4 = 86.05 m, Hs = 80.24 m, and Hb = 78.21 m. The discharge in pipe
3-6 is 0.1892 m3/s. The head downstream of the PRV, i.e., Hd is 90.02 -
55.06 X 0.18922 = 88.05 m > 80.00 m and the PRV has remained
operative.
If the pressure setting of the PRV had been 89 m, for the assumed H
values it is initially in the operative mode as 89.49 m > 89 m. However,
in the iterative procedure, the downstream pressure becomes less than 89
m and therefore the PRV is removed and the resistance of pipe 3-6 is in-
creased by 55.06. The normal iterative procedure is followed and the final
solution is H2 = 92.00 m, H3 = 89.69 m, H4 = 86.71 m, H5 = 82.14
m, and H6 = 80.98 m. Q3-6 is 0.2879 m3/s, so that Hd = 89.69 -
55.06 X 0.28792 = 85.13 m < 89 m, the setting of the PRV.
For the changed nodal demands given in Example 8.7, the PRV operates
as a check valve and the final solution is H2 — 90.32 m, H3 = 84.16 m,
H4 = 86.86 m, // 5 = 86.05 m, and Hb = 85.93 m.

Example 8.13

Solve Example 8.8 by the method of balancing flows.

Solution

As explained in Example 8.8, the nodes 5 and 6 are separated from the
main network. The analysis of the separated network gives Hs = 59.46
m, H6 = 57.68 m, Q4_5 = 0.7847 mVs, and ft-6 = 0.2153 m3/s. The
main network is now analyzed with outflows of 0.2 + 0.2153, i.e., 0.4153
nvVs at node 3 and 0.5 + 0.7847, i.e., 1.2847 rrrVs at node 4. The solution
is H2 = 92.00 m, H3 = 89.93 m, and H4 = 86.23 m.
220 HARDY CROSS METHOD

8.4 MODIFIED HARDY CROSS METHOD [103,104]

The usual Hardy Cross method, described so far, is based on three as-
sumptions: (1) considering one loop (node) correction equation at a time,
(2) neglecting the effect of adjacent loops (nodes), and (3) neglecting the
higher-power correction terms in the Taylor's series. As seen in Example
8.1, the third assumption doesn't have much effect on convergence. The
method is now modified herein so that the second assumption of neglect-
ing the effect of adjacent loops (nodes) is not made, i.e., the method takes
into consideration the effect of the adjacent loops (nodes) in deriving the
correction terms.

8.4.1 METHOD OF BALANCING HEADS

The network of Figure 8.1 having four pipes i, j , k, and m is shown in

Figure 8.6 with adjacent loops. Let the loop under consideration be c and
the adjacent loops be ai, aj, ak, and am so that these adjacent loops are
connected to loop c through pipes i,j, k, and m, respectively. As the effect
of adjacent loops is considered for all loops, all AQ corrections will be dif-
ferent from those obtained by neglecting the effect of adjacent loops. Let
these modified corrections be denoted by the additional subscript m. Thus,
A<2a*m denotes the modified correction for the adjacent loop ak. Taking
clockwise AQ corrections as positive, Equation (8.1) modifies to

- ,A0. l w )" H - 1AQaJmT

- RokdQk - xAQcm + iAQakm)"

- RoMQ™ - iAQcm + iA0.mm)- = 0 (8.20)

Expanding in the Taylor's series, neglecting higher-power terms of AQ,

rearranging terms and simplifying, the correction term for the fk iteration
is given by

E *~ • >Q"
x
n • Rox • tQ7

n - R o x - ,QTX
,AQaxm (8.21)
n • Rox • ,QT
 Modified Hardy Cross Method 221

iAQ 0 i m

i iQi
E Roi 'o1
iAQ arnrT1 ,-j iAQ Q jm

E iQk
'k ' Rok
iAQ o km

Figure 8.6 Four-piped loop with adjacent loops.

The first term on the right-hand side of Equation (8.21) represents the
usual correction tAQc given by Equation (8.5). The second term is the
summation of the influences of the adjacent-loop corrections. The term
within parentheses is termed the adjacent loop influence factor and is
denoted by Fax. Thus, Fai is a factor representing the influence of an adja-
cent loop connected through pipe x. Equation (8.21) can, therefore, be
written as

2 Fax • ,AQa (8.22)

To evaluate the value of ,AQcm, the values of ,AQaxm are necessary. How-
ever, these values are also unknown and depend upon other loops. There-
fore, it would be necessary to solve simultaneously all such loop-flow cor-
rection equations. However, because only one loop is considered at a time
(assumption 1), ,AQaxm is replaced by ,AQax, the corrections obtained by
the usual method, and only half the effect is considered to compensate for
the replacement of tAQaxm by tAQax. Thus, Equation (8.22) modifies to

(8.23)

Equation (8.23) is used in evaluating the AQ corrections. As the effect

of adjacent loops is considered, the modified method converges in less
number of iterations. However, the computational work per iteration has
increased.
222 HARDY CROSS METHOD

Example 8.14

Solve Example 8.3 by the modified Hardy Cross method of balancing

heads.

Solution

Using the values of iteration 1 from Table: Example 8.3, we get

= 0.024 + ^ | T T g 9 (-0.051)

(-0.024) = 0.015 m3/

iAQ,Im = - 0.051 + M | ^ | ( 0 - 0 2 4 ) ] = -0.050 m3/s

.Aflllm = - 0.024 + ^ f f 33 00 ' 33 ^ / " 44 55 ) (0-024)

(0-024) 11 = -0.012 m3/s

These corrections are applied, the pipe-discharges are corrected, and

the next iteration is carried out. The iterative procedure is continued and
the solution given earlier is obtained.

8.4.2 METHOD OF BALANCING FLOWS

In the modified method, the effect of the adjacent nodes is considered.

Therefore, for the network of Figure 8.4, Equation (8.13) modifies to
 Modified Hardy Cross Method 223

or

r.ff.-.ff,-
R,.,,

-+

,ff, - ,Aff,m - ,Affm

=0 (8.25)

Expanding in the Taylor's series, neglecting higher power terms of AH,

rearranging terms and simplifying, the correction term for the t'h iteration
is given by

,AHim = ,AHin
Q_

•on neci ed
t" / (8.26)

The first term on the right-hand side of Equation (8.26) represents the
usual correction ,AHj given by Equation (8.16). The second term is the
summation of the influences of the adjacent-node corrections. The term
within parentheses is termed the adjacent node influence factor and is
denoted by Fai. Thus, Fai is a factor representing the influence of the adja-
cent node /. Equation (8.26) can now be written as

,AHjm = , FBi • ,AHi (8.27)

Herein also, to evaluate the value of ,AHjm, the values of ,A//,m are nec-
essary. However, these values are also unknown and depend upon other
nodes. Therefore, it would be necessary to solve simultaneously all such
nodal-head correction equations. However, since only one node is con-
sidered at a time, ,AH,m is replaced by ,Aff,-, and only half the effect is con-
sidered. Equation (8.27) now modifies to

,A#,m = ,AH, + \ Fa, • ,A//, (8.28)

224 HARDY CROSS METHOD

Equation (8.28) gives the modified AH correction term, and because the
effect of adjacent nodes is considered, it converges faster than the usual
method. However, the computational work per iteration has increased.

Example 8.15

Solve Example 8.9 by the modified Hardy Cross method of balancing

flows.

Solution

Using the values of iteration 1 from Table: Example 8.9, we get

• * " < - = - ' • * •

= -1.54 m

= 1.42 m

1 [0.0536 .. „ . 0.0653
= -°-68 + T 0J189 ( 1 - 5 2 ) + aTT89 ( 0 - 4 7 )
= - 0.21 m

AH
,A// 6m
- n 47 , [ [° 0 7 7 9 (0)
- 0.47 + 2 [ 0 J 4 3 2
,m+, °- 0653
= 0.31 m

Applying these corrections, the second-iteration values become 2H3 =

94.46 m, 2H4 = 95.42 m, 2 // 5 = 91.79 m, and 2H6 = 93.31 m. By contin-
uing the iterative procedure, the solution given earlier is obtained.

8.5 SELECTION OF INITIAL VALUES

In the Hardy Cross method, it is necessary to initialise the pipe flows

(method of balancing heads) or the nodal heads (method of balancing
 Selection of Initial Values 225

flows) and also the pipe resistances if they are unknown. The computa-
tional work involved depends upon the initial values of these parameters.
If the initial values are close to the final values, the computational work
and therefore the number of iterations is less. Therefore, a good selection
of the initial values is desirable.

8.5.1 SELECTION OF PIPE FLOWS

For better initial guess of pipe flows, Conklin [105] suggested the use of
a distributional factor for each pipe as a function of its diameter and
length. Williams [ 106] used distributional factors for initializing pipe flows
in his studies on convergence problems in network analysis. Two other ap-
proaches based on (1) a minimum spanning tree and (2) a minimum path
length tree can also be used. Using pipe resistance constants, these ap-
proaches convert the looped network into a branching configuration, i.e.,
a tree. (If the DW head ioss formula is used, the coefficient of friction
equal to 0.02 can be used for all pipes, and pipe resistance constants are
evaluated.) Pipes not on the tree are assigned zero flow and therefore the
discharges in pipes on the tree are obtained as described in Section 7.4.1
of Chapter 7.

8.5. Vlinimum-Spanning-Tree Approach

A minimum spanning tree of a network is a tree in which the total pipe

length is minimum. Epp and Fowler [107] proposed the minimum span-
ning tree approach, but instead of the pipe lengths, the pipe resistance con-
stants are used. Epp and Fowler suggested the use of Prim's [108] algo-
rithm in obtaining the minimum spanning tree. This algorithm is described
herein with the help of a distribution network shown in Figure 8.7(a). The
pipe resistance constants are as shown along the pipes. The algorithm is as
follows.
Consider any node of the network (say, node 7). From amongst the pipes
connected to the node (pipes 7-6, 7-3, 7-8, and 7-11), pick the one with the
smallest resistance constant (pipe 7-3 with R — 80). Call the two nodes
and the connecting pipe a partial spanning tree (nodes 3, and 7, and pipe
3-7). The remaining nodes in the network are the isolated nodes. Next con-
sider all pipes connecting a node on the partial spanning tree with an iso-
lated node (pipes 3-2, 3-4, 7-6, 7-8, and 7-11). Select the pipe with the least
R value (pipe 3-2 with R = 120), and add it to the partial spanning tree to
obtain a new partial spanning tree and continue the procedure. When all
isolated nodes are connected to the partial spanning tree, the resulting net-
work is the minimum spanning tree. The minimum spanning tree for the
network of Figure 8.7(a) is shown in Figure 8.7(b). The numbers near the
226 HARDY CROSS METHOD

250 j
a—
10 120 ' 200
o o CD
00
C\J

40 6 160 7 250

200 ,
140 '
s CD
CD

80 160 ( 200
10 11 12
(a)
10 130 33U
o- 1 5* '
2 9
60 460
'6 7
5 3*
oo 6
7 JOj 1 8, . 1 \t —•
10 11 12 80 150 310 510
(b) (c)

Figure 8.7 Selection of initial pipe flows by conversion of looped network to Iree: (a) using mini-
mum spanning tree approach; (b) using minimum path length tree approach.

pipes show the order in which the pipes are added. The same minimum
spanning tree is obtained if any other node is selected as a starting node.

8.5.1.2 Minimum Path Length Tree Approach [1,85]

Considering the pipe resistances instead of the pipe lengths, the lengths
of the different paths connecting a demand node to a source node are ob-
tained, and the path with the minimum path length is selected. This con-
cept is used in obtaining a tree. The algorithm, explained with the help of
the network in Figure 8.7(a) is as follows:
Consider all nodes directly connected to the source node (nodes 2 and
5), obtain their path lengths (paths 1-2 and 1-5 with path lengths of 10, and
20, respectively), and select the node with the least path length (node 2
with path length 10). The two nodes and the connecting pipe form a partial
spanning tree (nodes 1 and 2 and pipe 1-2). Next consider all nodes di-
rectly connected to the nodes on the partial spanning tree (nodes 5, 6, and
3), obtain their path lengths (20, 70, and 130 for nodes 5, 6, and 3, respec-
tively), and select the node with the least path length (node 5 with path
length 20). Add it to the partial spanning tree to obtain a new spanning tree
and continue the procedure. When all nodes are connected to the partial
spanning tree, the resulting network is the required tree. The tree obtained
for the network of Figure 8.7(a) is shown in Figure 8.7(c). The numbers
near the nodes show path lengths, and those near the pipes show the order
 Convergence Problems 227

in which the pipes are connected. The minimum path length tree approach
can also be used for multiple source networks [86] in which a separate tree
is obtained for each source node.
Both these methods have some merits and limitations [107,109] but in
general provide good starting values.

8.5.2 SELECTION OF NODAL HEADS

No specific approaches for selecting the initial values of heads at nodes

are available. However, the minimum spanning tree approach or the mini-
mum path length tree approach can be used to evaluate the initial pipe
flows from which the initial values of the nodal heads can be determined.

8.5.3 SELECTION OF UNKNOWN RESISTANCES

The initial value for the unknown resistance of a pipe can be taken as the
average value of the resistances of the pipes connected to the pipe under
consideration at its end nodes. For example, in the network of Figure
8.7(a), the initial value of the resistance constant of pipe 2-3, if unknown,
can be taken as 87.5, the average of 10, 60. 80, and 200, the resistance con-
stants of pipes 1-2, 6-2, 3-7, and 3-4, respectively.

8.6 CONVERGENCE PROBLEMS

Even though the Hardy Cross method is simple, easy to apply, and
therefore, suitable for hand calculation, its greatest drawback is the pres-
ence of the convergence problems. It is observed to converge very slowly
and therefore requires large number of iterations. The number of iterations
increases as the size of the network increases. In some situations the
method starts diverging and does not converge at all.
In addition to the size of the network, the number of iterations required
for convergence depends upon (1) the convergence criterion. (2) the pipe
resistance constants, and (3) the pipe discharges. Lower values of conver-
gence criterion, larger values of some pipe resistances (small diameters or
large lengths of pipes or both) in comparison with the R values of other
pipes, and lower values of discharges in some pipes tend to increase the
number of iterations. The convergence criterion depends upon the accu-
racy one desires, consistent with the time, effort, and money one is pre-
pared to spend. Techniques to improve convergence when the delay is due
to unusually large pipe resistances or unusually low pipe discharges have
been developed by several investigators [110-113]. Some of them are briefly
described herein.
228 HARDr' CROSS METHOD

Voyles and Wilke [110] have shown that the convergence is improved in
the method of balancing heads when the loops are selected such that the
sum of the common flow resistance factors is minimum. For example, in
Figure 7.5 of Chapter 7 if ^, 4 is large (300 instead of 30), it is preferable
to select loop 2 comprising pipes 1, 2, 5, 7, 6, and 3 instead of pipes 4 . . .
7. In other words, pipes with large R values are externalized.
For the method of balancing flows, Dillingham [111] treated a pipe with
a large R value as an elongated joint and applied the computed head cor-
rection to both the end nodes of the pipe.
When the discharge in a pipe is extremely low, the head loss through it
is negligible. This results in a very large value of Q/h for this pipe in the
denominator of the AH correction equation [Equation (8.16)]. This gives
an extremely small value for the AH correction, thus delaying conver-
gence. Dillingham suggested the use of the linear head loss relationship
h = RQ for such a pipe so that the corresponding Q/h term can never ex-
ceed the MR value of this pipe and thus convergence is improved.
Williams [106] exclusively studied the problem of convergence in the
method of balancing heads and suggested the use of a convergence acceler-
ation factor A, given by

A = 0.41966 - 7.77649 [log (pipes)] + 7.53672 [log (total pipes)]

(8.29)

in which pipes represent the number of pipes in the network, i.e., X; and
total pipes represent the total number of pipes used in the iteration. Thus,
a pipe common to two loops is considered twice. For example, for the net-
work of Figure 7.5, pipes would equal 7, while total pipes would equal 10,
since pipes 1, 3, and 4 are common to two loops. The correction obtained
at the end of each iteration, ,AQC [Equation (8.5)] is multiplied by A and
the resulting value A • ,AQC is used to modify the pipe flows.
Instead of using a common fector, different factors for individual loops
or nodes can also be used. The modified Hardy Cross method described
earlier modifies individual corrections. McCormick and Bellamy [112] and
Van den Berg as reported by Epp and Fowler [107] suggested special mea-
sures to maximize individual corrections and thereby improve conver-
gence.
For the method of balancing flows, Kootattep and Aya [113] have sug-
gested the use of buffered successive over relaxation, i.e., the use of an
accelerator and decelerator.
If the present correction ,AHj is in the same direction as that of the pre-
vious correction t -iA//,, i.e., if both the corrections are of the same sign,
the present correction is accelerated to enhance convergence. On the other
hand, if the present correction is in the reverse direction and therefore of
 Problems 229

the opposite sign, the present correction is decelerated to suppress oscilla-

tion and divergence. Thus, the corrected nodal head for the next iteration
is given by

,+,Hj = ,Hj + A • ,AHj (8.30)

in which

A = Au if ,AHj X ,-AHj > 0 (8.31a)

A = A2, if tAHj X ,-iAHj < 0 (8.31b)

The appropriate value of accelerator A i is between 1.3 and 1.4 and that
of decelerator A2 between 0.5 and 1.0. They have recommended At as 1.3
and A2 as 0.5. They have named the method as buffered successive over re-
laxation because the work of the decelerator looks like that of a buffer
against excessive over correction.
This principle can also be used for the method of balancing heads.

8.7 PROBLEMS

Problems 8.1 . . . 8.8: Analyze the networks described in Problems

7.1 . . . 7.8 by Hardy Cross method based on (a) balancing heads and (b)
balancing flows.
 Newton-Raphson Method

9.1 INTRODUCTION

S seen in Chapter 8, the Hardy Cross method attempts to solve the

A nonlinear equations involved in network analysis by making certain
assumptions. The effect of neglecting higher-power correction terms is tol-
erable, and as seen in Example 8.1, a small number of iterations are re-
quired for a single loop even when the initial guess is poor. However, the
effect of neglecting adjacent loops (connected nodes) and considering only
one correction equation at a time is considerable, and the number of itera-
tions required for convergence increases as the size of the network in-
creases. Even the partial consideration of the effect of adjacent loops (con-
nected nodes) in the modified Hardy Cross method does little to improve
the situation. Several techniques described in Chapter 8 improve conver-
gence and reduce the number of iterations. However, in practice this num-
ber can be quite large for networks. Therefore, instead of considering only
one correction equation at a time and solving it, if the effect of all adjacent
loops (connected nodes) is considered and if all the correction equations
are solved simultaneously, convergence can be achieved in a smaller num-
ber of iterations. The Newton-Raphson method described in this chapter
and the linear theory methoc described in the next chapter attempt to colve
all the concerned equations simultaneously. However, because the equa-
tions are nonlinear, an iterative procedure is necessary for their solution.

9.2 BASIC CONCEPTS [114,115]

The general expression for the finite Taylor's series is

b
F(a + b) = F(a) + bF'(a) + ^rF" y
2! ' ' (n - 1)!

X F'-\a) + ~^F"(a + Ob), 0 < 0 < 1

231
232 NEWTON-RAPHSON METHOD

in which F', F'",..., F" denote the first, second, . . . , « " " derivative of
F, respectively, and the last term denotes the remainder after n terms. Us-
ing the remainder after two terms, the finite Taylor's series becomes

F(a + b) = F(a) + bF'(a) + y F"(a + 6b), 0 < 6 < 1 (9.2)

The Newton-Raphson method (the underlying idea is due to Newton but

the method as now used is due to Raphson) is based on Equation (9.2) and
is used to find the real roots of F(x) = 0 if F'(x) is a simple expression
and easily found.

9.2.1 SINGLE-VARIABLE FUNCTION

Consider a single-variable nonlinear function F(x) = 0. Let a be one of

its roots so that F(a) = 0. Tofindthe value of a by an iterative procedure,
let it be assumed that ,x is the value of x used in t'h iteration. Let the value
of the function F(x) for the fh iteration, ,F(x) be nonzero, and therefore
let ,Ax be the additive correction for the fh iteration so that

F(x) = tF(x + Ax) = 0 (9.3)

Expanding Equation (9.3) in the form of Equation (9.2), we get

,F(x) + ,Ax • ,F'(x) + ~^~ ,F"(x + 6Ax)

= 0, 0 < 6 < 1 (9.4)

Now if tAx is relatively small in comparison with ,x, (tAx)2 will still be
smaller, and therefore we can neglect the remainder after two terms. The
resulting expression is
,F'(x)-,Ax = -,F(x) (9.5)
from which the expression for the correction is

The value of x for the next iteration is given by

[ F(x)
 Basic Concepts 233

Repeatedly using Equation (9.7), the value of x satisfying F(x) — 0,

i.e., x = a, is found.
For a geometrical interpretation of the Newton-Raphson method, con-
sider the magnified view of the graph y = F(x) in the vicinity of x = a,
as shown in Figure 9.1. Let point P represent the t'h iteration position so
that OA = ,x, and AP = ,F(x). Draw a tangent at P intersecting the x-
axis in point B. The slope of the tangent at P, ,F'(x) = tan z. JCBP =
-PA/AB = -,F(x)/AB. Therefore, AB is equal to -,[F{x)/F'(x)], and
thus it represents the correction ,Ax. Point B represents t+lx and point Q
gives the value of,+, F{x). Continuing the procedure of drawing successive
tangents, we reach point S at which F{x) = 0 and x — a, the desired
value.
In deriving the correction term of Equation (9.6), the remainder after
two terms in Equation (9.4) has been neglected. This introduces an in-
herent error in the Newton-Raphson method. It can be shown that the in-
herent error in (t + l)lh iteration, i.e., , +1 £, is given by [115]

\r+lE | < (,Ax)2 (9.8)

Thus, the error in the (t + l)'h iteration is proportional to the square of

the error in the t'h iteration. Each subsequent reduction in error in the
Newton-Raphson method is proportional to the square of the previous er-
ror, and thus the convergence property is quadratic. (This is the reason for
accelerated convergence in Example 8.1 of Chapter 8. Because only one
loop is considered therein, the Hardy Cross and Newton-Raphson meth-
ods are identical.)

9.2.2 MULTIPLE-VARIABLE FUNCTION

The Newton-Raphson method can be extended to more than one varia-

ble. Consider, for example, two variables x, and x2 for which the two equa-
tions are

F1(xltx1) = 0 (9.9a)

and

F2(x,,x2) = 0 (9.9b)

Let the t'h trial values of*! and x2, be ,x, and ,x2 and the corrections be
,Ax, and ,Ax 2 , respectively. Following the procedure outlined earlier, and
assuming ,Ax, and ,Ax2 to be small, and therefore neglecting their
234 NEWTON-RAPHSON METHOD

Figure 9.1 Geometrical interpretation of the Newton-Raphson method.

squares, products, and higher powers, we get corresponding to Equation

(9.5),

(9.10a)

and

Writing in the matrix form,

i i i i

3F, 3F, '

AJC, F,
dxi dx2
(9.11)
dF2 dl\
Ax2 F2
dx, dx2 , i i t

in which the prefixing subscript t denotes the matrix containing elements

for the t'h iteration, and F, and F2 denote FI(JCI,^2) and F2(xlrx2), respec-
tively.
Extending the Newton-Raphson method to n variables, xt,x2 xn,
interrelated through n nonlinear equations, F,(JCI,JC2 . . . . xn) = 0, F2(xu
 Head Equations 235

x2 . . . . xn) — 0 through Fn(x,,x2 . . . . xn) — 0, we get, for the /"" itera-

tion

3F,
A.x, F,
dXi dx2 ' ' • dxn

dF2 dF2 dF2

Ax2 (9.12)
dx, dx2 ' • • dxn

dx2 dxn

The first matrix is the coefficient matrix, also called the Jacobian of the
« functions F,, F2, . . ., Fn with respect to the n variables xt, x2 xn
for the t'h iteration. The second matrix is a column matrix representing the
corrections Ax,, Ax2, . . ., Axn for the /'* iteration, and the matrix on the
right-hand side is a column matrix giving the values or the residues of the
functions Fu F2, . . ., Fn for the t'h iteration.
The application of the Newton-Raphson method to the analysis of water
distribution networks was first proposed by Martin and Peters [116] and
since then it has been extensively used in research and in the field f88,91,
107,112,117-119]. The Newton-Raphson method can be applied to any type
of equations described in Chapter 7. However, it is commonly applied in
practice to the nodal-head equations, i.e., H equations, and to the loop-
flow correction equations, i.e., AQ equations.

9.3 HEAD EQUATIONS

As illustrated in Chapter 7, the head equations, i.e., H equations, are

formed taking the nodal heads as the basic unknowns. The application of
the Newton-Raphson method to the H equations is explained and illus-
trated herein for different situations.

9.3.1 NETWORKS WITH KNOWN PIPE RESISTANCES

For single- and multiple-source networks with known pipe resistances,

H equations are formed with unknown heads as basic unknowns. The
236 NEWTON-RAPHSON METHOD

number of these equations decreases as the number of source nodes in-

creases. (If the unknown nodal flows are also treated as basic variables, we
get H-q equations.) These equations are then simultaneously solved using
the Newton-Raphson method.

Example 9.1

Solve Example 8.3 (Example 8.9) by the Newton-Raphson method based

on H equations.

Solution

The network reproduced herein as Figure: Example 9.1, has two source

Head d e m a n d , 0-2 nr?/

0-1

nodes 1 and 2, with known heads, Hol = 100.00 m and Ho2 = 95.00 m.
The heads H3 . . . H6 are unknown.
Because n = 1.85, (l/n) = 0.54. Therefore, the //equations at nodes
3 . . . 6 are, respectively,

F3 =

100 - // 4 \ 0 5 4 /95 - tf,y-5 H< -

20 I + 30 40

0.2 = 0 (2)

(3)

95 - H6 - // 5
F6 = -0.1 = 0 (4)
20 50
 Head Equations 237

[These equations are the same as Equations (7.15a) . . . (7.15d) with

minor modifications.]
Therefore, the correction equations, omitting, for simplicity, the prefix-
ing subscript t are

dF3 dF3 dF3 dF3 AH3

dH3 a//4 dHs dH6

dF4 dF4 dF4 dF4

AH4 F4
dH3 dH4 dH5 dH6
(5)
dF5 dF5 dFs dFs
dH3 dH4 dH5 dH6

dF6 dF6 dF6 dF6

8H3 dH6

The elements of the coefficient matrix are

dH3 ~ (40)
°- 5 ^(ioo-ftr-«(-i)- (10) 05 - 95)-

dH4 dHs ' dHii ' dH3

dF4 0.54
^^(100-//4)-046(-l)
8H4 (20)

0.54 ( f t - Hs)-°46
(40)0.54

dF4 0.54 /If 1/ \-0

\ n * n

dH5 (40),0.54 " S>

dF4
dH6

dH4 ~ (40)°54 (
~ (40)
4 5)

ft - //s)- 0 4 6 (-D (ft -

238 NEWTON-RAPHSON METHOD

0.54
(H6 -
dH6

3F
dH3
-=O;4£ =O
m.
dF6 0.54
(H6 -
dH5 (50)°-s'

0.54 0.54
(95 - - Hsy
(5O)05'

Let the first iteration heads for the nodes be YH3 = 96.00 m, tH4 =
94.00 m, ,/Zs = 92.00 m, and ,/7 6 = 93.00 m as assumed in Example 8.9.
Substituting these first-iteration H values and evaluating the values of the
elements of the Jacobian and also the F3 . . . F6 values, Equation (5) be-
comes, for the first iteration

-0.1946 0 0 0
0 -0.1866 0.0536 0
0 0.0536 -0.1189 0.0653
0 0 0.0653 -0.1432

-0.3000
iA// 4 0.2827
X (6)
.A// 5 -0.0813
0.0673

Solving Equation (6) we get ,A// 3 = - 1 . 5 4 m, ,A// 4 = 1.63 m,

i&H5 = 0.41 m, and iAH6 = 0.66 m. Applying these corrections the H
values for the second iteration become 2H3 = 94.46 m, 2H4 = 95.63 m,
2HS = 92.41 m, and 2H6 = 93.66 m. The iterative procedure is continued
and the answers as given in Example 8.3 are obtained.
We observe the following points from this example:
(1) The coefficient matrix contains several elements having zero values.
If a node under reference (node i) is not connected to another node
(node;), then (dFJBHj) = 0 and (dFj/dHt) = 0. In the given net-
work, since node 3 is not connected to node 4, (dF3/dH4) = 0 =
(dFJdH3), and so on. Therefore, in large networks, because a node
would generally be connected to a maximum of 5 to 6 nodes, the
number of zero elements is quite large and the number of nonzero el-
ements is relatively small. Thus, the coefficient matrix is sparse.
 Head Equations 239

(2) The elements of the coefficient matrix are symmetrical about the
main diagonal, i.e., (d/\/d//,) = (dF,/3//,), i =£ j . If node / is con-
nected to node y through pipe ij, the value of (3F,/d//,) is obtained
through the term denoting discharge in pipe ij. Because the same
pipe is considered for evaluating {dFjIdF,), these two elements are
equal. Thus, only the elements on the main diagonal and in the trian-
gle above it, i.e., upper triangle, need be evaluated. The elements in
the lower triangle can be written because of symmetry.
(3) The elements on the main diagonal, dFJdHj, i = j are numerically
the same as the corresponding Y.(Quln • hu) values in Table: Exam-
ple 8.9. For example, 3FJdHA and T,{QJn • hiA) are numerically
equal (0.1866). Similarly, the nondiagonal elements, i.e., dFJdHj,
i ± j , are the same as the corresponding Qijln • hu values in Table:
Example 8.9. For example, 8FJdH5 = Q4-s/n • h4-s = 0.0536.
Further, the values in Example 8.15 (modified Hardy Cross method)
are identical with the corresponding values in this example. How-
ever, the Newton-Raphson method solves the equations exactly,
while the modified Hardy Cross method solves them approximately.

9.3.2 NETWORKS WITH UNKNOWN PIPE RESISTANCES

Shamir and Howard [88] extended the Newton-Raphson method based

on head equations to networks with unknown pipe resistances. It has been
shown in Chapter 7 that when a network contains all the three types of un-
knowns (unknown nodal heads, unknown nodal flows and unknown pipe
resistances) as long as the network-solvability rules are satisfied, it is pos-
sible to formulate an adequate number of independent H-q-R equations
which are also reducible to H-R equations and //equations. Therefore, for
networks with unknown pipe resistances, these equations are formulated
and iteratively solved using the Newton-Raphson method as illustrated in
the following example.

Example 9.2

Solve Example 8.4 (Example 8.10) by the Newton-Raphson method us-

ing (1) H-q-R equations, (2) H-R equations, and (3) H equations.

Solution

In the network of Figure: Example 8.4(a) reproduced herein as Figure:

Example 9.2, the unknowns are qx, H3, // 4 , R4, and Hs. All network solv-
ability rules are satisfied as shown in Example 8.4.
240 NEWTON-RAPHSON METHOD

1. H-q-R Equations

Writing the node-flow continuity equations for nodes 1 . . . 5, the H-q-R

equations are, respectively,

100 - HA0* /100 - //3\05

h (1)

(2)

100 -
10
(3)
inn
IUU H \o.5 Iu u \o.5
~ "4 1 , {^3 - HA - 0.5 = 0
I + I
27.5 I
(4)

98 - // 5 \ 0 5
- 0.7 = 0 (5)
40
For the five corrections, AH3, AH4, A// 5 , Aqu and A/?4, the five equa-
tions in matrix form are

dH3

0 0 A// 4

dF3
'\dH3 1 1
 Head Equations 241

11
I dF4 dF4 6t4
dH3 dH4 3R4
dFs
AR4
m4 dHs dR4
Let the first-iteration values be the same as those in Example 8.10 so that
tH3 = 97.00 m, ,// 4 = 96.00 m, ,H5 = 95.00 m, and ,/? 4 = 20. Let
,<7i = 1.00 mVs. [Actually, applying the flow continuity equation to the en-
tire network, we get q, — 1.2 rrrVs. Therefore, qx could have been treated
as a known and the unknowns could be the remaining four parameters for
which Equations (2) . . . (5) are the required equations. However, tqt is
deliberately taken different from 1.2 m3/s to illustrate the method.] There-
fore, the five equations for the first iteration are

0.0913 0.0884 () 1 0
0.1768 0 0.()456 0 0
0.3634 0.0953 () 0 0
0.0953 -0.2955 0. 118 0 0.00559
0 0.1118 -0. 1574 0 -0.00559
1
AH3 ' -o.
2548 '
AH4 -0.0274
X AHS 0.1106 (7)
Aqi 0.1742
AR* . L -0.2025 .

Solving these equations simultaneously we get ,A// 3 = 0.3676 m,

AH4 = 0.2413 m, AHS = -0.8245 m, ,A<?, = 0.1990 m3/s, and
AR4 = —8.182. Therefore, the second-iteration values are 2H3 = 97.37
m, 2H4 = 96.24 m, 2 // 5 = 94.18 m, 2ql = 1.20 m3/s, and 2R4 = 11.82.
The iterative procedure is continued and the solution given in Example 8.4
is finally obtained.
2. H-R Equations

The H-R equations are Equations (2) . (5) for which the correction
equations in matrix form are

dF2 BF2
AH3 F2
dH3

\dF3 dF^
! dH3 dH4 \\*H4\
242 NEWTON-RAPHSON METHOD

i i (8)
dF4 dF4 dF4 3F4
A// 5 F4
dH3 dH4 dHs dR4

dFs dFs dFs

0 AR4 Fs
dH4 dHs dR4

Using the first-iteration values, four correction equations similar to

Equation (7) are formulated and solved. The iterative procedure is contin-
ued to get the final solution.
3. H Equations

As explained earlier, the H equations are obtained by writing the node-

flow continuity equations at nodes 2, 3, and merged-node 4-5. The //equa-
tion for the merged-node 4-5 is

27.5 40
(9)

Thus, Equations (2), (3), and (9) provide three equations for H3, H4,
and Hs. The corresponding correction equations are

0 AH3
dH3

dF3 dF3
dH4
AH4 F3 (10)
dH3

dF9 dF9 dF9

A// 5
dH4

The iterative procedure is continued to get the final solution.

Note that when unknown pipe resistances are present in a network, the
symmetry of elements in the coefficient matrix is lost.
9.3.3 NETWORKS WITH PUMPS

It has been shown in Chapter 7 that an additional equation for the head
supplied by a pump can be formulated [for example, Equation (7.31)].
However, as explained in Example 8.11, the head delivered by a pump, hp,
 Head Equations 243

is initially assumed, and the head at the upstream node of the pipe in
which the pump is located is correspondingly increased. If the head-
discharge relationship is assumed for the pump such that it is consistent
with the head loss relationship for pipes, the pump can be made an integral
part of the pipe in which the pump is located.
Example 9.3

Solve Example 8.6 (Example 8.11) by the Newton-Raphson method

based on nodal head equations.

Solution

The node-flow continuity equations for nodes 3 . . . 6 of the network of

Sump,
W.L.=80-00 m : ESR
W.L=100-00m

Figure: Example 8.6 reproduced herein as Figure: Example 9.3 are as fol-
lows:
80.00 + hB- H3\05 (H3 - HA05 - H5\05
R3

- 0.08 = 0 (1)
<H3 - HMS

- 0.12 = 0 (2)

F t = !±=JlL

- 0.12 = 0 (3)

100.00 - HA0* (H6 - HA0-* (H6 - //5\05

F6 =

- 0.12 = 0 (4)
244 NEWTON-RAPHSON METHOD

Because the pump is situated in pipe 1, the head supplied by it, hp, is in-
cluded in pipe 1 as shown in Equation (1). As explained in Example 8.11,
the R values of the pipes change from iteration to iteration. They are ini-
tially obtained by taking / = 0.02 and are updated for each subsequent
iteration. Further, as hp is also assumed and updated, the nodal heads
H3 . . . H6 are the variables for which Equations (1) . . . (4) are available.
The equations for the AH corrections in matrix form are

3F3 dF3 dF3

0 AH3
dH3 dH4 dHs

dF4 dF4 3F4 dF4

AH4 F4
dH3 dH4 dH5 dH6
(1)
dFs dF5 dF5 dFs
AHs
dH3 dH4 dH5 dH6

n dF6 8F6 3F6

AH6
bH4

For the first iteration, XH3 = 85.00 m, ,// 4 = 83.00 m, ,// 5 = 80.00 m,
and ,// 6 = 90.00 m. Further, ,hp = 10.75 m, ,R, = 204, ,R2 = 2,066,
>R3 = 8,707, ,/? 4 = 3,265, ,/? 5 = 755, ,/? 6 = 3,265, and ,/? 7 = 157, as
in Example 8.11. Using these first-iteration values, the elements of the co-
efficient matrix in Equation (1) are calculated and the equations are solved
to get the head corrections and subsequently the nodal heads for the sec-
ond iteration. The pipe discharges obtained in the first iteration are used
in the second iteration to determine the friction coefficients, the pipe resis-
tances, and the power supplied by the pump 2hp . The iterative procedure
is continued and the solution given earlier is finally obtained.

9.3.4 NETWORKS WITH VALVES

As stated in Section 8.3.5 of Chapter 8, for a check valve, the resistance

of the pipe in which it is situated is correspondingly increased. In the for-
mulation of H equations, this, pipe is considered if the direction of dis-
charge in it is permissible or it is dropped if the direction reverses and the
discharge is not permitted.
A PRV is initially assumed in the operative mode and is, therefore, re-
placed by a sink node and a source node. The discharge in the pipe down-
stream of the PRV is obtained considering the set-pressure head of the
 Head Equations 245

PRV and the assumed head of the downstream node of the pipe. This dis-
charge is then taken as the additional outflow at the upstream node of the
pipe. If the PRV becomes inoperative or prevents flow in the reverse direc-
tion, the procedure described earlier for the check valve is followed. A
check is made at the end of each iteration to verify that the PRV has re-
mained in the assumed mode.
Example 9.4

Solve Example 8.7 (Example 8.12) by the Newton-Raphson method

based on H equations.

Solution

Using the known head for node 1 and the set-pressure head of the PRV,

0-3 rcfe
100-OOnn t
1
° R=2 2

0-6
o-2 y\ 5Q 6

the H equations for nodes 2, 3, 4, 5, and 6 of the network of Figure: Exam-

ple 8.7(a) reproduced herein as Figure: Example 9.4 are, respectively,

100.00 - // 2 \ 0 5 H2 -
F2 =

- 0.3 = 0 (1)

H2 - H3\°-s (H3 - K
F3 = - (0.2 + ft.,,) = 0 (2)
40

- 0.5 = 0
40
(3)

#4 - / / 5 \ ° 5 / t t , - // 6 \ 0 5
- 0.4 = 0 (4)
9 / \ 12

80.00 - HA05 (Hs - H6\os

+ 12 - 0.6 = 0 (5)
246 NEWTON-RAPHSON METHOD

in which Q3-u = discharge in the pipe on the upstream of the PRV and
Rd-6 = resistance of the pipe on the downstream of the PRV.
As the PRV is situated just downstream of node 3, Rd-t = R3~6 = 50.
Assuming the first-iteration H values as in Example 8.12, Q3-u = Qd-6 =
0.3162 nWs. Equations (1) . . . (5) are solved by the Newton-Raphson
method, the corrections are evaluated, and the nodal heads H2 . . . H6 are
corrected. Discharge in pipe 3-6 is determined, the pressure head at the
downstream end of the PRV is calculated, and the mode of the PRV is de-
termined. If the PRV is in the operative mode, Q3-u is determined and the
next iteration is carried out. If the PRV becomes inoperative, Q3-u in
Equation (2) and the term 1(80.00 - H6)/Rd-6]°* in Equation (5) are re-
placed by [(//, - //6)/(/?3-6 + /?PRV)J°-S. If the PRV behaves as a check
valve, these terms in Equations (2) and (5) are dropped. Thus, at the end
of each iteration, the operational mode of the PRV is determined and the
appropriate forms of Equations (2) and (5) are selected.

9.4 LOOP EQUATIONS

The Newton-Raphson method is also used extensively for loop equa-

tions in which the loop-flow corrections, AQ values, are the basic un-
knowns. The formulation of different types of loop equations is explained
in Chapter 7. The application of the Newton-Raphson method for the solu-
tion of these equations is explained and illustrated herein.

9.4.1 NETWORKS WITH KNOWN PIPE RESISTANCES

The formulation of AQ equations for real loops of single-source net-

works and real and pseudo loops for multisource networks is explained in
Chapter 7. These AQ equations are solved simultaneously by the Newton-
Raphson method as illustrated in Example 9.5.

Example 9.5

Solve Example 8.3 (Example 9.1) by the Newton-Raphson method based

on AQ equations.

Solution

Let the first-iteration pipe discharges satisfying the node-flow continuity

equations at nodes 3 . . . 6 of the network of Figure: Example 9.5 be the
same as in Example 8.3. i.e., ,(?, = 0.35 m3/s, ,(? 2 = 0.45 m 3 /s,
,G 3 = 0.05 m7s, , e 4 = 0.05 m3/s, , f c = 0.30 nWs, ,Q 6 = 0.20 nrVs,
and iQ7 = 0.10 m J /s. in the directions as shown. Taking AQ, . . . AQm
 Loop Equations 247

demand, 0-2 m/%

Head, 04
q,^OO-K m
Q2 U Q5
' pipe,2" R=20 5 40 T
i*
?!
6' Loop, I j II
a
^-

3, Q 3
, i
.2 Q* R
3 J0_ /T95-00 6 22
q 0-1
III z i
. i

as the loop flow corrections in the clockwise direction for the three loops
I . . . Ill, respectively, the AQ equations are
F, = 20(0.45 + AQ,) 185 - 30(0.05 - AQ, + AQ,,)' 85

- 10(0.05 - AQ, + AQ,,,) 185

- 40(0.35 - AQ, + AQ,,,) 185 = 0 (1)

F,, = 40(0.30 + AQ,,) 185 - 50(0.10 - AQ,,) 185

- 20(0.20 - AQ,,) 185 + 30(0.05 - AQ, + AQ,,)1 -M = 0

(2)
F,,, = 4 0 ( 0 . 3 5 - AQ, + AQ,,,) 185

+ 10(0.05 - AQ, -I- AQ,,,) 18S + (95 - 100) = 0 (3)

The basic unknowns are AQ, . . . AQ,,, which will be assumed and
successively corrected by applying corrections. Let these additive correc-
tions be z, . . . 2ni to the assumed AQ, . . . AQ,,, values, respectively.
The corrections z, . . . Zm will be successively evaluated until they
become less than some prespecified value, and the AQ, . . . AQ,,, values
stabilize. The correction equations are

dF, dF, dF,

dAQ, dAQ,, dAQ,,, z,
dF,, dFu dF,,
Zn F,, (4)
dAQ, dAQ,, SAG,,,

dF,,, dF,,,
3AG, 3AGn
248 NEWTON-RAPHSON METHOD

The nine elements of the Jacobian are

ir = 20 X 1.85(0.45 + AC?,)085
dAQ,

+ 30 X 1.85(0.05 - A0, + AC?,,)0'85

+ 10 X 1.85(0.05 - AQ, + A0,,,) 085

+ 40 X 1.85(0.35 - AQ, + AC?,,,)085 (5)

dFm
= 40 X 1.85(0.35 - AC?, + AC?,n)08
SAG,,

+ 10 X 1.85(0.05 - AQ-, + AGni)0-1 (13)

For the first iteration taking ,AC?i = iAC?>> = iAC?m = 0, Equation (4)
becomes

54.88 -4.35 -31.77 ' -1.328 '

-4.35 53.43 0 iZn = - 2.707 (14)

-31.77 0 31.77 J , 0.775 ,

The solution of Equation (14) yields t z, = 0.0146 m3/s, au = -0.0495

m /s, and iZm = —0.0098 m 3 /s, which are the corrections to iAC?>,
3

iAC?,,, and iAQm, respectively. Because the XAQ values are zero, the 2AC?
values are 2A£?i = 0.0146 m3/s, 2AC?n = -0.0495 m 3 /s, and 2AC?,n =
— 0.0098 m 3 /s. Taking these AC? values, the elements of the coefficient
matrix and the F values are reevaluated to get correction equations, similar
to Equation (14), for the next iteration. The iterative procedure is con-
tinued and the results are shown in Table: Example 9.5. The answers given
earlier in Example 8.3 are finally obtained.
9.4.2 NETWORKS WITH UNKNOWN PIPE RESISTANCES

The application of the loop equations to networks with unknown pipe re-
sistances was first suggested by Gofman and Rodeh [91]. They used graph
theory in formulating the appropriate equations.
As seen in Chapter 7, we can formulate either the AQ-AR equations or
the AC? equations for networks with unknown pipe resistances. When AC?-
 TABLE: EXAMPLE 9.5.

Loop I Loop II Loop III

AO,, F,, * i . AO,,, AO,,,, Fm,

Iteration m3/s m m3/s m3/s m m3/s m3/s m m3/s

1 0 -1.328 0.0146 0 2.707 -0.0495 0 0.775 -0.0098

2 0.0146 -0.175 0.0070 -0.0495 0.059 -0.0009 -0.0098 0.029 0.0060
3 0.0216 0.005 - 0.0504 - 0.003 -0.0038 0
250 NEWTON-RAPHSON METHOD

AR equations are formulated and solved, the unknown pipe resistances are
also assumed and corrected along with the pipe discharges and the method
is implicit. When the AQ equations are formulated and solved, the method
is explicit. The unknown pipe resistances are evaluated in the end after sat-
isfactory convergence is reached.

Example 9.6

Solve Example 8.4 by the Newton-Raphson method of loop equations

using the implicit and explicit approaches.

Solution

Implicit Approach. As explained in Example 8.4, three loop equations

are necessary. The three loops are selected as shown in Figure: Example
8.4(b). The corrections for loops I and II are AQi and AQ,,, respectively,
in the clockwise direction and the additive correction to R4 is AR4. As-
suming the first-iteration pipe discharges and R4 as in Example 8.4, the
three AQ-AR equations for loops I, II, and III are, respectively,

F, = 8(0.7 + AQ,)2 - 27.5(0.2 - AQ, + AQ,,)2

- 10(0.5 - AQ,)2 = 0 (1)

Fn = 27.5(0.2 - AQ, + AQ,,)2 + (20 + AR4)(0A + AQ,,)2

- 40(0.3 - AQ,,)2 + 8(0.3 + AQ,,)2 = 0 (2)

F,,, = 8(0.7 + AQ,)2 + (20 + AR4)(0.4 + AQ,,)2

- 40(0.3 - AQ,,)2 + (98 - 100) = 0 (3)

Let the corrections to AQ, and AQ,, be z, and z,,, respectively, and the
correction to AR4 be r4. The correction equations in matrix form are

dF, dF,
u F,
dAQ, dAQ,,

dF,, dF,, dF,,

Zn F,, (4)
dAQ, dAQ,, dAR4

dF,,, dF,,, dF,,,

dAQ, dAQ,, 8AR4
 Loop Equations 251

For the first iteration, taking iAQ, = i = iA/?4 = 0, Equation (4)

becomes

32.2 -11 0 ' 0.32 '

-11 55.8 0.16 = _ 1.42 (5)

11.2 40 0.16 , 1-52 ,

Solving Equation (5) we get xz{ = -0.01498 rrvVs, jzn = -0.01473

m 3 /s, and ,r4 = -4.767. Therefore, for the second iteration, 2AQ, =
-0.01498 m3/s, 2AQl{ = -0.01473 m3/s, and 2AR4 = -4.767. The iter-
ative procedure is continued and the results are shown in Table: Example
9.6. The answers given in Example 8.4 are finally obtained.
Explicit Approach. Considering loops I and II of Figure: Example
8.4(c), the explicit AQ equations are obtained. As explained in Example
8.4, the correction AQn is also extended to pipes 1, 4, and 5, i.e., it is ap-
plied to the pipes of the overlapping loop, although the AQ equation is for
the pseudo loop. Therefore, the two equations are

F, = 8(0.7 + A<2, + AQnf - 27.5(0.2 - AQ,)1

- 10(0.5 - AC, - AQuf = 0 (6)

F n = 8(0.3 + - 10(0.5 - AQ, - AQnf + (100 - 98) = 0

(7)

The correction equations are

(8)

For the first iteration, taking ,Ag, = AQu = 0, Equation (8) becomes

32.2 21. 2 ' \Z\ ' 0.32 '

(9)
10 14. 8 0.22

which yields ,2, = -0.00027 m3/s and ,z,, = -0.01468 nWs. This gives
2AQ, = -0.00027 m3/s and 2AQU = -0.01468 mVs.For these AQ
 TABLE: EXAMPLE 9.6.

Loop I Loop II Loop III

AO,, AO,,, Fn, zn, Fm.

Iteration m3/s m m3/s m3/s m m3/s Afl4 m

1 0 0.3200 -0.01498 0 1.4200 -0.01473 0 1.5200 -4.767

2 -0.01498 -0.0008 0.00006 -0.01473 0.0527 0.00009 -4.767 0.0529 -0.383
3 -0.01492 0.0001 -0.01464 -0.0002 -5.150 0
 Loop Equations 253

values, as the 2F, and 2F,, values are negligible, the final AQ values are
AQ, = -0.00027 mVs and Ag,, = -0.01468 nrVs. These corrections
are applied to loop I (pipes 1, 2, and 3) and loop II (pipes 6, 2, 1, 4, and
5) to obtain the final Ql . . . Q6 values that are as obtained earlier in Ex-
ample 8.4. Loop III comprising pipes 3 . . . 6, the overlapping loop com-
prising pipes 6, 2, 1,4, and 5 or the path comprising pipes 1, 4, and 5 can
then be used to evaluate the value of R4 that is as obtained earlier in Exam-
ple 8.4.

9.4.3 NETWORKS WITH PUMPS

The head delivered by a pump is expressed in terms of the discharge de-

livered by it, and therefore, in terms of the discharge of the pipe in which
the pump is located. This is then used to formulate the AQ equations as
shown in Example 9.7.
Example 9.7

Solve Example 8.6 by the Newton-Raphson method based on loop equa-

tions.
Solution

The three loops shown in Figure: Example 8.6 are considered and the
first-iteration R values are taken as shown in Table 2: Example 8.6. The
pipe discharges are also taken as shown therein. From Equations (2) and
(3) of the solution of Example 8.6, the head delivered by the pump is
given by

hp = 12.05 - 38.2(0, - 0.0353) 2 (1)

Because Qp — Q,, Equation (1) becomes

hp = 12.05 - 38.2(0, - 0.0353)2 (2)

Using Equation (2), the AQ equations for the first iteration are

,F, = 2207(0.08 + AQ, - AQ,n)2 + 3865(0.02 + AQ, - AQ,,)2

- 9947(0.06 - AQ,)2 = 0 (3)

,Fn = -3865(0.02 + AQ, - AQ,,)2 - 835(0.06 - AQ,, + AQ,,,)2

+ 3933(0.04 + AQ,,)2 = 0 (4)

254 NEWTON-RAPHSON METHOD

,F,,, = 141(0.22 + AQmf + 835(0.06 - AQn + AQmf

- 2207(0.08 + AG, - AQm)2 - 187(0.22 - AQ,,,)2

+ [12.05 - 38.2(0.22 - 0.0353 - AQ,,,)2] + (80 - 100) = 0

(5)

Note that as the Darcy-Weisbach friction formula is used, the R values

of the pipes change from iteration to iteration. The AQ equations given by
Equations (3) . . . (5) also change and therefore the prefixing subscript 1
is used for F, . . . Fm. In the earlier examples, as R values remained con-
stant, the AQ equations did not change.
The correction equations for the first iteration are

1701.5 -154.6 -353.2 ' -20.13 '

-154.6 569.5 -100.2 iZ\\ = — 1.74 (6)

-353.2 -100.2 611.9 J IZ.II . , -22.62 ,

from which ,z, = 0.02397 m3/s, ,z,, = 0.01276 m3/s, and ,?,„ = 0.05289
m3/s. Therefore, 2A£>, = 0.02397 m3/s, 2A<2n = 0.01276 rrrVs,
and 2A<2,,, = 0.05289 m3/s. This gives 2Ql = 0.1671 nWs, 2Q2 = 0.05108
nrVs, 2 03 = 0.03603 m J /s, 2Q4 = 0.03121 nWs, 2QS = 0.1001 m3/s, 2Qb
= 0.05276 m3/s, and 2Q7 = 0.2729 m3/s. Using these discharges, the/and
R values of pipes for iteration 2 are obtained and Equations (3) . . . (5) are
rewritten using the revised R values. Equation (6) is then reformulated,
solved, and the z values are obtained. The iterative procedure is continued
and the answers given in Example 8.6 are finally obtained.

9.4.4 NETWORKS WITH VALVES

The formulation of AQ equations for networks containing check valves

and pressure reducing valves is described in Chapter 7. The loops are se-
lected such that these valves are externalized.
For check valves, the AQ correction is restricted so that the discharge in
the pipe containing a check valves does not change direction.
A PRV is initially assumed to be in the operative mode and is replaced
by source and sink nodes. A pseudo loop is introduced connecting this
source node to a given source node, and AQ equations are formed and
solved. The correction, however, must be applied so that continuity is
maintained. When a PRV is inoperative or when it prevents flow in the
 Ixiop Equations 255

reverse direction, the procedure described earlier for check valves is fol-
lowed. A check must be made to ascertain the mode of the PRV.

Example 9.8

Solve Example 8.7 by the Newton-Raphson method based on loop equa-

tions.

Solution

It is presumed that the PRV is in the operative mode. It is replaced by

source and sink nodes and the two loops shown in Figure: Example 8.7 are
considered. The first-iteration pipe discharges are assumed the same as
those assumed in Example 8.7. These values from Table: Example 8.7 are
as follows: , £ , = 2.0 irrVs, ,Q2 = 1.2 rrrVs, ,£>., = 0.5 nxVs, ,Q4 = 0.2
nWs, , g 5 = 0.9 nWs, ,(?6 = 0.5 m'/s, and ,Q7 = 0.1 m3/s. As calculated
in Example 8.7, /?PRV = 55.06. It has been ascertained in Example 8.7 that
the PRV is in the operative mode for the assumed pipe discharges.
The AQ equations for loops I and II are

F, = 6(1.2 + AQ,)2 - 40(0.2 - Ag, + A0,,) 2

- 4(0.5 - AQ,)2 = 0 (1)

F,, = 2(2.0) 2 + 6(1.2 + AG,)2 + 9(0.9 + AQn)2

+ 12(0.5 + AQ,,)2 - 50(0.1 - AQu)2 + (80 - 100) = 0 (2)

Note that, for maintaining continuity, the correction AQn is applied to the
pipes of loop III, i.e., pipes 4 . . . 7. Therefore, the correction AQn is ap-
plied to pipe 4 in Equation (1), but is not applied to pipes 1 and 2 in Equa-
tion (2).
The correction equations are

(3)
256 NEWTON-RAPHSON METHOD

For the first iteration taking ,A£?, = iAQu = 0, Equation (3) becomes

34.4 -16.0 6.04

(4)
14.4 38.2 6.43

which yields ,z, = -0.2160 m3/s and ,z n = -0.0869 m3/s. Therefore,

3 3
2AQ, = -0.2160 m /s and 2AQ2 = -0.0869 m /s. The iterative proce-
dure is continued and the answers of Example 8.7 are obtained in the end.
If the PRV becomes inoperative, as stated in Example 8.7, loop III of
Figure: Example 8.7 replaces loop II, and the resistance constant of pipe
7 becomes 50 + 55.06 = 105.06. The AQ equations for loops I and III,
are, respectively,

F, = 6(1.2 + A0,) 2 - 40(0.2 - AQ, + AQm)2

- 4(0.5 - AQ,)2 = 0 (5)

and

Fm = 40(0.2 - Afl + AQmf + 9(0.9 + Afi,,,)2

+ 12(0.5 + AQ,,,)2 - 105.06(0.1 - AQm)2 = 0 (6)

Simultaneous solution of these equations by the Newton-Raphson

method gives 2A<2, = -0.1109 m3/s, and 2AQm = -0.1390 rrrVs. Con-
tinuing the iterative procedure the final solution of Example 8.7 is ob-
tained.
For the increased nodal flows as in Example 8.7, when the PRV behaves
like a check valve, pipe 7 and loop II (or loop III) are omitted, and only
loop I is considered for which the answers given in Example 8.7 are ob-
tained.

9.5 LABELLING NETWORK ELEMENTS

In the Hardy Cross method described in the last chapter, each loop or
node was considered separately, and the effect of adjacent loops or con-
necting nodes was neglected. Thus, each equation was an independent
equation containing only one unknown and therefore could be solved
separately. On the other hand, in the Newton-Raphson method described
in this chapter, all nodes or loops are considered simultaneously. Simi-
larly, in the linear theory method described in the next chapter, all pipes
 Labelling Network Elements 257

or nodes are considered simultaneously. Therefore, all H equations, AQ

equations, or Q equations are considered and solved simultaneously in the
Newton-Raphson and linear theory methods.
Each loop or node of a water distribution network would be connected
to a few loops or nodes and therefore the resultant coefficient matrix would
be sparse as stated in the solution of Example 9.1. In the small illustrative
networks considered in this text, the sparseness of the coefficient matrix
does not have much significance. However, the networks in practice may
have hundreds of loops, nodes, and pipes, and therefore, the sparseness of
the coefficient matrix attains significance. Even though the matrix is
sparse, the nonzero elements may not be concentrated in a particular part
of the matrix. If the nonzero elements are concentrated as close to the
main diagonal as possible, the time and effort required for the solution of
the equations is reduced and at the same time the accuracy of the solution
is increased.
Starting from the main diagonal, if we label 1, 2, . . . , w the parallel
diagonals that must be traversed in any row of the matrix so that only zero
elements remain in the remaining diagonals, w is the half band width and
2w + 1 is the bandwidth of the matrix. It is therefore necessary for the
bandwidth of the coefficient matrix to be as small as possible. Thus, the
loops, nodes, and pipes of the network should be labelled such that the
bar 'width of the coefficient matrix is minimum.
Several algorithms are available for labelling the network elements but
none of them can guarantee the minimum bandwidth of the coefficient ma-
trix in all situations. Martin and Peters [116] discussed bandwidth reduc-
tion and showed that by proper node-labelling of the illustrative network in
their paper, the bandwidth could be reduced from 24 to 6. However, they
have not provided an algorithm for bandwidth reduction. Grooms [120]
has provided an algorithm for bandwidth reduction for structural net-
works.
Some algorithms suggested for the bandwidth reduction of the coeffi-
cient matrix for H, AQ, and Q equations encountered in the analysis of
water distribution networks are now described.

9.5.1 ALGORITHM OF EPP AND FOWLER [107]

Epp and Fowler applied the Newton-Raphson method for the solution of
AQ equations, and therefore proposed a labelling algorithm for the loops
of a network. Their algorithm is explained with the help of Figure 9.2 in
which the loops are initially labelled with letters for reference.
Their algorithm is as follows:
A list is made of all loops that contain a node of degree 2. (The degree
of a node is n if n is the number of pipes connected to it.) Usually, these
258 NEWTON-RAPHSON METHOD

L M N 0 P
IY VII X XUI XV

F G H J K
II V VIII XI XIV

A B
c D E
I III
YI IX XII

Figure 9.2 Labelling loops of a network using an approach suggested by Epp and Fowler.

loops occur at the corners of the network (Loops A, E, L, and P). All
pseudo loops are then added to this list. The first loop of this list is labelled
I (one). Its neighbors are labelled II, III, IV, . . . etc. (Loop A is labelled
I and adjacent loops F and B are labelled II and III, respectively.) Now
consider loop II. Its unlabelled neighbors are L and G. The labelled neigh-
bors of L consist of single loop II, whereas loop G has two labelled neigh-
bors, namely II and III. Thus, loop L which has the fewest labelled neigh-
boring loops is labelled IV and loop G is labelled V. All the neighbors of
loop II have now been labelled and therefore loop HI is considered next.
As loop III has only one unlabelled neighbor C, loop C is automatically
labelled VI. The process is continued until all loops are labelled as shown
in Figure 9.2.
In this simple illustrative network, the half bandwidth is three and is op-
timal. However, the above algorithm may not always produce a minimum
bandwidth. Therefore, the above procedure is repeated starting with each
loop having a node of degree 2. The numbering that produces the smallest
bandwidth is then selected as the best loop labelling.

9.5.2 ALGORITHM OF JEPPSON AND DAVIS

Jeppson and Davis [83] have suggested an algorithm that can be used to
band the coefficient matrix of Q, AQ, or H equations. Initially, the pipes,
loops, or nodes are arbitrarily numbered and the algorithm suggests how
they should be renumbered for minimum bandwidth. The algorithm is ex-
plained for H equations for a one-source, nine-demand node network
shown in Figure 9.3. The source node is labelled 0 (zero) and the demand
nodes are labelled 1 . . . 9 as shown so that the H unknowns are //, . . .
H9. The nine //equations obtained for the node-flow continuity relation-
ship at nodes 1 . . . 9 are labelled (1) . . . (9), respectively, so that the
equation numbers are identical with the node numbers for which they are
written. (This is the reason why the source node is labelled 0 herein). The
unknowns in Equations (1) . . . (9) are shown by • in Figure 9.4(a). This
 i

(2)
U) (1)

17) C5) (3)

7

19) (8) C6)

Figure 9.3 Network for illustrating the algorithm of Jeppson and Davis for banding a coefficient
matrix.

Unknown Unknown
1 2 3 4 5 6 7 8 9 A; AI 1 2 3 4 5 6 7 8 9
1 • 9 3 1 3 * •
it 2 • • 82 2-2 • • •
S.3 • • • 6 1 3-5.1 • • •
• • • • 10 4 i'5 4 • • • •
• • 10 5 sl5 • • •
• • • • 14 8 6 iG 7 • • • • •
§7 • • • • • 106 7i 8 • • •- •
,3"8 • • • 10 7 8^6 • • • •
9 • • • 15 9 9 9 • • •
9 7 5 7 12 14 10 12 16
(0) 0
1
B! 4 3 1 2 6 8 5
(bl
7 9
Position Position
On ^
1 2 3 4 5 6 7 1 2 3 4 5 6 7 8 9
Unknown 8 9 £
Unknown
3 4 2 1 7 5 8 3 4 2 17 5 8 6 9 A-
3 6 9 Ai
4 Ai * ~ 3 * • •
1 1 4
62 2 ^ 2 • • • • 6
&
•S.1
• 10 4 3 -%_ 4 0 • • • 7
• 7 3 4 b 1 • • • 10
-05 • 10 5 5 -§ 5 • • • 10
£7 10 6 6 2 7 • • • • • 10
£8 • • 13 7 7 ;§ 8 • • • • 13
iS"6 14 8 8
^ 6 • • • • 14
9 • • 16 9 9 9 • 16
Bi 4 7 7 9 10 11 13 13 16
(c) d)
Figure 9A Procedure for obtaining a banded coefiicient matrix: (a) original coefficient matrix;
(b) matrix after changing positions of rows (equations); (c) matrix after changing positions of col-
umns (unknowns); (d) finally obtained banded coefficient matrix.
260 NEWTON-RAPHSON METHOD

figure, therefore, also depicts the nonzero elements in the coefficient ma-
trix. The numbers across the top denote the positions of the nine un-
knowns, Hi . . . H9, while the numbers at the left denote the positions of
the nine equations, Equations (1) . . . (9).
The iterative procedure for reducing the bandwidth consists of two
steps: (1) changing the positions of the equations, and (2) changing the
positions of the unknowns.
1. Changing Positions of Equations. Each row (equation) is given a
weight equal to the sum of the position numbers of the first and last non-
zero elements in that equation, as shown in column At [Figure 9.4(a)]. The
weights in column At are next arranged in ascending order of magnitude
(column A /) and the equations are changed to positions given by this order
[Figure 9.4(b)].
When weights of two or more arrays are equal, they are assigned posi-
tions arbitrarily. It is presumed that the ascending order is maintained as
long as the weight of an array is not less than the weight in the earlier
array.
2. Changing Positions of Unknowns. Each column (unknown) is given a
weight equal to the sum of the position numbers of the top and bottom non-
zero elements in the column as shown by row fi, [Figure 9.4(b)]. The
weights in row 5, are next arranged in ascending order of magnitude (row
B-) and the positions of the unknowns are reassigned [Figure 9.4(c)].
These two steps of giving weights and reassigning the positions of the
equations or unknowns in the ascending order of the weights is repeated
iteratively until all the equations or unknowns are arranged in ascending
order and no change in their positions occurs, or until some maximum al-
lowed number of such iterations have been completed to avoid alternation
of the positions of two or more equations or unknowns between consecu-
tive iterations.
One more iteration of changing positions of the rows of the matrix in
Figure 9.4(c) results in the matrix shown in Figure 9.4(d). For this matrix
it is observed that all #, values are in the ascending order. Since all At val-
ues are already in the ascending order, as shown again by column At in
Figure 9.4(d), the iterative procedure is complete.
The original coefficient matrix of Figure 9.4(a) is symmetrical while that
in Figure 9.4(d) is not. For a symmetrical matrix, the relative positions of
each equation and corresponding unknown are identical. Therefore, to
maintain the identical positions, the ordering of the unknowns is the same
as that of the equations. Therefore, only At (or Bt) are obtained and the
positions of the equations and unknowns are identically changed. For the
network of Figure 9.3, the different symmetrical coefficient matrices at dif-
ferent ordering stages are shown in Figure 9.5. The node labels as obtained
from the final coefficient matrix for the symmetrical matrix with minimum
bandwidth [Figure 9.5(b)] are shown in parentheses in Figure 9.3.
 Problems 261

Position Position
4 5 6 89 I
Unknown Unknown
8 6 9 AiA'i S. 3 2 4 1 7 5 8 6 9 A ;
3 • • • 5 1 1_ 3 • • • 4
£2 • • • • 7 2 2^2 • • • • 6
• • • 9 4 3-S.4 • • • 7
• • • • 7 3 4 b 1 • • • 9
• • • 12 6 5 "§7 • • • • • 10
5 • • • • • 105 6*2 5 • • • 11
28 • • • • 12 7 7z 8 • • • 13
• • • • 148 8uf 6 • • • • 14
• • • 169 9 9 • • • 16
Bi 5 7 9 7 12 10 12 14 16 Bi 4 6 7 9 10 11 13 14 16
(a) (b)
Figure 9.5 Procedure for obtaining symmetrical banded coefficient matrix: (a) after first order-
ing; (b) finally obtained symmetrical banded coefficient matrix.

Pseudo loops are necessary in the formulation of Q and AQ equations.

Therefore, in large multisource networks, the bandwidth of the coefficient
matrix would be decided by the pseudo loops and would be large as
pointed out by Epp and Fowler [107]. Pseudo loops should be selected so
that they pass through the smallest number of real pipes.

9.6 PROBLEMS

Problems 9 . 1 . . . 9.8: Analyze the networks described in Problems 7.1

. . . 7.8 by the Newton-Raphson method based on (a) head equations and
(b) loop equations.
 Linear Theory Method

10.1 INTRODUCTION

E have seen in Chapter 7 that at least some (if not all) equations in-
W volved in pipe network analysis are nonlinear. Because no direct
method is available for their solution, these equations are linearized and
then solved. Naturally, the solution is approximate, and therefore it is cor-
rected, and the iterative procedure is continued until satisfactory accuracy
is reached.
The Newton-Raphson method described in the previous chapter expands
the nonlinear terms in Taylor's series, neglects the residues after two
terms, and thereby considers only the linear terms. As such, the Newton-
Raphson method linearizes the nonlinear equations through partial differ-
entiation. This method is general and works even when the nonlinear
equations are transcendental containing exponential, trigonometric, hyper-
bolic, or logarithmic terms. The nonlinearity in the equations for pipe net-
work analysis, however, is algebraic, uniform and simple: the variables are
raised to the same, nonunity exponent. For example, the nonlinear Q equa-
tions obtained from the loop-head loss relationship contain terms, RXQ",
and the H equations obtained from the node-flow continuity relationship
contain terms, [(//, - //,• )//?<,•]"". In both these types of nonlinearity, the
value of n is the same for all terms—1.852 in the Hazen-Williams pipe-
head loss formula and 2.0 in the Darcy-Weisbach and Manning formulas.
This is a helpful feature and the nonlinear term can be easily linearized by
merging a part of the nonlinear term into the pipe resistance constant as in
Equation (6.5). This principle was first suggested and used by Mcllroy
[121], Marlow et al. [122], and Muir [123]. Later, Wood and Charles [124]
developed it, and it is now widely used in practice [124-129].
Even though the principle can be used to linearize all types of equations
including A£> equations [121,130], the method is applied in practice to
pipe-discharge equations, i.e., Q equations, and the nodal-head equations,
i.e., H equations.
263
264 LINEAR THEORY METHOD

10.2 PIPE DISCHARGE EQUATIONS

The pipe discharges are taken as the basic unknowns in formulating the
Q equations. The application of the linear theory method for the solution
of these equations is described herein for different situations.

10.2.1 NETWORKS WITH KNOWN PIPE RESISTANCES

The Q equations are formulated for single- and multiple-source net-

works with known pipe resistances, as explained in Chapter 7. The Q
equations for the node-flow continuity relationship are already linear. The
Q equations for the loop-head loss relationship for the real, and pseudo
loops, if any, are nonlinear. These nonlinear equations are linearized in the
form of Equation (6.15). Thus, the nonlinear loop-head loss equations are
written as

")G. = 0, c = 1 . . . C (10.1)

in which Rax = known resistance constant of pipe x; ,QX = assumed dis-

charge in pipe x for the t'h iteration; and Qx = the unknown parameter.
Equation (10.1) expressed in the linearized form becomes

2 MO, = 0, c = 1 . . . C (10.2)

in which ,/?,' = modified resistance constant of pipe x for the t'h iteration
and is given by

.*,' = R~ I .& I"- (10.3)

Muir [123J and Wood and Charles [124] have suggested that the pipe dis-
charge ,QX should be taken as unity for the first iteration. (Muir has also
suggested that ,QX can be alternatively taken as some other arbitrarily
chosen figure.) Thus, for the first iteration, we have

,l?x' = Rox,x = 1. . .X (10.4)

and Equation (10.1) modifies to

2 R^Q: = 0, c = 1 . . . C (10.5)
 Pipe Discharge Equations 265

The simultaneous solution of the linear node-flow continuity equations

and the linearized loop-head loss equations [Equation (10.5)] would give
the pipe discharges at the end of the first iteration, i.e., Q,,,,, x = 1 . . .
X. [The discharge in pipe x obtained after t'h iteration is denoted herein by

Muir suggested that from the second iteration onwards

,Gx = I , M , a X a , - . ) i 0 5 , t = 2, 3 . . . (10.6)

According to the procedure suggested by Wood and Charles, the Qx

values obtained from the first iteration are taken as the assumed values for
the second iteration for evaluating 2 / ? x . Thus,

2& = Q,w,x = 1... X (10.7)

If this procedure of taking the obtained values of the previous iteration

as the assumed values for the next iteration is continued further, it is
observed that the Qx values start oscillating. To prevent oscillations, Wood
and Charles proposed that the average of the obtained Qx values in the pre-
vious two iterations should be taken as the assumed value for the next iter-
ation. Thus, from the third iteration onwards,

Q ( 2) gx
,6x = "" 2 """ , x = 1 . . . X (10.8)

Instead of taking the average of the obtained values in the previous two
iterations, taking the average of the assumed and obtained values in the
previous iteration gives rapid convergence as shown later and also by Ex-
ample 10.1. Therefore,

,6, = " " ) g l 2 &

""> , x = 1 . . •X (10.9)

The method is illustrated in Example 10.1.

Example 10.1

Solve Example 8.3 by the linear theory method based on Q equations.

Solution

In the network of Figure: Example 8.3, reproduced herein as Figure:

266 LINEAR THEORY METHOD

Head, demand,O-2rry's

Example 10.1, the node-flow continuity linear relationships for nodes

3 . . . 6 give, respectively,

Q, - Q3 - 0.3 = 0 (1)

i + & - Qs - 0.2 = 0 (2)

0 5 + 0 7 - 0.4 = 0 (3)

0 6 - 0 7 - 0.1 = 0 (4)

From the two basic loops I and II and the pseudo loop HI, we have, re-
spectively,

20<2£85 - 300V 85 - 10£>!85 - 4 0 0 ! 8 5 = 0 (5)

4O0 5 85 - 5O0}:85 - 2O0J-"5 + 3 0 0 } 8 5 = 0 (6)

4 0 0 ! 8 5 + 1O0} 85 + (95 - 100) = 0 (7)

[Equations (1) . . . (7) are the same as Equations (7.12a). . . (7.12g) ex-
cept that n is replaced by 1.85.]
For the first iteration, XQX = . . . = xQi = 1 and therefore, Equations
(5) . . . (7) become, respectively,

2O02 - 3O04 - 1003 - 400, = 0 (8)

4O0 5 - 5O0 7 - 2O06 + 3O04 = 0 (9)

4 0 0 , + 1003 - 5 = 0 (10)
 Pipe Discharge Equations 267

Solving Equations (1) . . . (4) and (8) . . . (10) simultaneously, we get

<2un = 0.16 m3/s, Q2(l) = 0.37 nrVs, 0 3 ( 1 ) = -0.14 mVs. 0 4 ( 1 ) = 0.08
m 3 /s, 0 5 ( 1 ) = 0.25 m3/s, 0 6 ( 1 ) = 0.25 m3/s, and 0 7 ( 1 ) = 0.15 m3/s. These
values are used as 2Q, values to calculate 2 /?/ values for the second itera-
tion. Thus, 2R[ = 40(0.16)°85 = 8.42 and 2R2 = 20(0.37)°-8S = 8.59.
Similarly, 2 /? 3 ' = 10 I -0.14 |0-85 = 1.88, and so on. Thus, the linearised
form of Equations (5) . . . (7) for the second iteration becomes

8.5902 - 3.510 4 - 1.88ft - 8.420, = 0 (11)

12.310 5 - 9.970, - 6.160 6 + 3.510 4 = 0 (12)

8.420, + 1.880.3 - 5 = 0 (13)

Solving Equations (1) . . . (4) and (11) . . . (13) simultaneously, we get

0 1 ( 2 ) = 0.5402 nWs, 0 2 ( 2 ) = 0.5464 mVs, 0, ( 2 ) = 0.2402 nrVs, 0 4 ( 2 ) =
-0.0872 m3/s, 0 5 ( 2 ) = 0.2593 mVs, 0 6 ( 2 ) = 0.2407 nrVs, and 0 7 ( 2 ) =
0.1407 rrrVs.
Following the averaging procedure suggested by Wood and Charles
[Equation (10.8)], for the third iteration the averages of the obtained values
in the previous two iterations, i.e., iterations 1 and 2, are taken. Thus,
30i = [0KD + 01(2)1/2 = (0.16 + 0.5402)/2 = 0.3501 m3/s, and soon.
The iterative procedure is continued and the first five iterations are shown
in Table 1: Example 10.1. The final solution, with an accuracy of 0.0001
m3/s in pipe discharges, is obtained in 14 iterations as 0, = 0.3246 m3/s,
Q2 = 0.4715 m3/s, 03 = 0.0246 m3/s, 0 4 = -0.0219 m3/s, 0 5 = 0.2496
m3/s, 0 6 = 0.2504 m3/s, and 0 7 = 0.1504 m3/s.
The first five iterations in the alternative procedure of taking the average
of the assumed and obtained values in the previous iteration, from third
iteration onwards are shown in Table 2: Example 10.1. This procedure re-
quires six iterations to obtain the final solution.
From the two tables it can be observed that for any iteration, the average
of the assumed and obtained values in the previous iteration is closer to the
final value than the average of the obtained values in the previous two itera-
tions. Thus, the convergence is faster in Table 2 than in Table 1.
In the text that follows, Equation (10.9) is used instead of Equation (10.8)
in the iterative procedure.

10.2.2 NETWORKS WITH UNKNOWN PIPE RESISTANCES

As illustrated in Chapter 7, we can formulate either the Q-R equations

or 0 equations for networks with unknown pipe resistances. When Q-R
equations are formulated and solved, the number of the unknowns is X +
 TABLE 1: EXAMPLE 10.1.

Discharge in Cubic Meters per Second

Iteration 1 Iteration 2 Iteration 3 Iteration 4 Iteration 5

Pipe Assumed Obtained Assumed Obtained Assumed Obtained Assumed Obtained Assumed Obtained

1 1 0.16 0.16 0.5402 0.3501 0.3048 0.4225 0.2632 0.2840 0.3631

2 1 0.37 0.37 0.5464 0.4582 0.4844 0.5154 0.4404 0.4624 0.4793
3 1 -0.14 -0.14 0.2402 0.0501 0.0048 0.1225 -0.0368 -0.0160 0.0631
4 1 0.08 0.08 - 0.0872 -0.0036 -0.0397 -0.0634 0.0057 -0.0170 -0.0261
5 1 0.25 0.25 0.2593 0.2546 0.2448 0.2520 0.2461 0.2454 0.2532
6 1 0.25 0.25 0.2407 0.2454 0.2552 0.2479 0.2539 0.2545 0.2468
7 1 0.15 0.15 0.1407 0.1454 0.1552 0.1479 0.1539 0.1545 0.1468
 TABLE 2: EXAMPLE 10.1.

Discharge in Cubic Meters per Second

Iteration 1 Iteration 2 Iteration 3 Iteration 4 Iteration 5

Pipe Assumed Obtained Assumed Obtained Assumed Obtained Assumed Obtained Assumed Obtained

1 1 0.16 0.16 0.5402 0.3501 0.3048 0.3274 0.3223 0.3248 0.3244

2 1 0.37 0.37 0.5464 0.4582 0.4844 0.4713 0.4715 0.4714 0.4715
3 1 -0.14 -0.14 0.2402 0.0501 0.0048 0.0274 0.0223 0.0248 0.0244
4 1 0.08 0.08 -0.0872 -0.0036 -0.0397 -0.0216 -0.0219 -0.0217 -0.0219
5 1 0.25 0.25 0.2593 0.2546 0.2448 0.2497 0.2496 0.2496 0.2496
6 1 0.25 0.25 0.2407 0.2454 0.2552 0.2503 0.2504 0.2504 0.2504
7 1 0.15 0.15 0.1407 0.1454 0.1552 0.1503 0.1504 0.1504 0.1504
270 LINEAR THEORY METHOD

XUR and the approach is implicit. When the Q equations are formulated
and solved, the number of the unknowns is X and the approach is explicit.
In the loop-head loss equations for the implicit approach, the terms indi-
cating the head loss through pipes of known resistance are linearized as
usual for pipe discharges and those indicating head loss through pipes of
unknown resistance are linearized for pipe resistances. Thus,

A = (Rox • ,Q"'x)Qx (linearized for basic unknown Qx) (10.10)

A = G<2") " Rx (linearized for basic unknown Rx) (10.11)

It is not necessary to initialize the unknown pipe resistances, but their

values are obtained at the end of each iteration. Thus, in the implicit ap-
proach, the unknown pipe resistances are also evaluated along with the
unknown pipe discharges for each iteration.
When a pipe with unknown resistance is internal, it appears in more
than one loop-head loss equation in the formulated Q-R equations and
therefore the unknown resistance is evaluated implicitly. However, if a pipe
with unknown resistance is external, the unknown resistance appears in
only one equation, and therefore this equation can be set aside, and the
unknown resistance can be evaluated in the end.
In the explicit approach, the pipes with unknown resistance are external-
ized and only Q equations are formulated and solved. The unknown pipe
resistances are evaluated in the end.
Example 10.2

Solve Example 8.4 by the linear theory method based on pipe-discharge

equations using the implicit and explicit approaches.
Solution

Implicit Approach. The network of Figure: Example 8.4 reproduced

herein as Figure: Example 10.2 has six pipes of which pipe 4 has an un-

o-smys c O-7
t 4
R4

node.1^
a
evil
to m
WO-OOmpN. a'
to m
0-6
^ ^

0-6 J -51 98-00m

 Pipe Discharge Equations 271

known resistance. Thus, the number of unknowns for Q-R equations is

seven. The seven Q-R equations are

Qi ~ Q3 + Qe - 0.6 = 0 (1)

Qi + Q3 - QA - 0.5 = 0 (2)

Q4 + Q, - 0.7 = 0 (31

-Qs ~ Q6 + 0.6 = 0 (4)

82? - 27.5QS - \0Ql = 0 (5)

27.523 + R4Q\ - 40QI + SQl = 0 (6)

- 8QI + (98 - 100) = 0 (7)

For the first iteration, i2i = . . . = ,Q 6 = 1- The term R4Ql is

linearized for R4. Therefore, taking i<24 = 1, R4Ql becomes R4 for the
first iteration. Thus, Equations (5) . . . (7) in the linearized form are

8 2 , - 27.5Q3 - 10& = 0 (8)

27.52a + R4 - 4OGs + 82a = 0 (9)

1022 - 82a - 2 = 0 (10)

Equations ( 1 ) . . . (4) and ( 8 ) . . . (10) are solved simultaneously. Equa-

tions (8) . . . (10) are updated and the first three iterations are shown in
Table: Example 10.2. The final solution is obtained in five iterations.
Explicit Approach. For the explicit approach, Equations (1) . . . (5) and
(7) provide six equations for the six unknowns Qt . . .Q6. These six equa-
tions are continuously updated and solved until satisfactory Q values are
obtained. The value of R4 is obtained in the end.
In this particular example, R4 appears in only one equation [Equation
(6)] in the implicit approach. If this equation is kept aside, the remaining
equations are the required equations for the explicit approach. However,
as in Example 8.5, if R3 is unknown instead of R4 (R4 = 14.85) the un-
known R3 would appear in Equations (5) and (6) and must be evaluated im-
plicitly.
For the explicit approach, to obtain only one equation involving R3, pipe
3 is externalized as explained earlier, by considering the overlapping loop.
Therefore, one equation from Equations (5) and (6) is retained and the
 TABLE: EXAMPLE 10.2.

Iteration 1 Iteration 2 Iteration 3

Pipe •*. <0x Ox,,, 2Q, 2R; Ox,2) 3Q, 3 « ; Q*,3,

1 8 1 8 0.7872 0.7872 6.298 0.5514 0.6693 5.354 0.7002

2 10 1 10 0.4128 0.4128 4.128 0.6486 0.5307 5.307 0.4998
3 27.5 i 27.5 0.0788 0.0788 2.167 0.3670 0.2229 6.130 0.1789
4 1 0.3660 0.3660 0.4184 0.3922 0.3791
5 40 1 40 0.3340 0.3340 13.360 0.2816 0.3078 12.312 0.3209
6 8 1 8 0.2660 0.2660 2.128 0.3184 0.2922 2.338 0.2791

Pipe Ro 1O4 iO2 0

4(1| 2O4 205 ft 4(2) 3O4 3Ql ft 4(3)

4 — 1 1 9.061 0.3660 0.1340 17.09 0.3922 0.1538 14.32

 Pipe Discharge Equations 273

other is replaced by an equation for the overlapping loop. Therefore, re-

taining Equation (5), Equation (6) is replaced by

80? + 14.8503 - 400^ + 8QI - 1 0 0 = 0 (11)

or by

80? + 14.8503 - 400^ + (98 - 100) = 0 (12)

for the pseudo loop comprising pipes 1, 4, and 5. Thus, Equations (1) . . .
(4), (7), and (12) provide six 0 equations for the six unknowns 0 , . . . 0 6 .
These equations are then solved by the linear theory method to obtain the
final 0 values and subsequently the value of R3.

10.2.3 NETWORKS WITH PUMPS

The formulation of 0 equations for networks with pumps is quite simple

as shown in Chapter 7. The head supplied by the pump h,, is expressed in
terms of the discharge through it, and therefore, in terms of the discharge
in the pipe in which the pump is located. Equation (5.4), transformed to
either Equation (5.6) or Equation (5.7), can be used. When Equation (5.6)
is u "\, instead of introducing a new variable Gp and thereby increasing
one unknown and adding one equation [14], Gp can be expressed in terms
of 0 P [Equation (5.5)] and therefore in terms of Qx, and can be used in
forming the 0 equations. This keeps the number of the basic unknowns
and available equations as X.
When Equation (5.7) is used, n is taken equal to the exponent of 0 in the
pipe head loss relationship so that all nonlinear terms are similar.

Example 10.3

Solve Example 8.6 by the linear theory method based on pipe-discharge

equations.

Sump
W.L.=80-00
274 LINEAR THEORY METHOD

Solution

The network has seven pipes and therefore the basic 0 unknowns are
Qi . . . Q7. From the node-flow continuity relationship at nodes 3, 4, 5,
and 6, of Figure: Example 8.6, reproduced herein as Figure: Example 10.3,
we get, respectively,

G. - Gi - a - 0.08 = o (i)

Qi - &. + Qs - 0.12 = 0 (2)

03 + Q* + ft - 0.12 = 0 (3)

-Qs - 06 + 07 - 0.12 = 0 (4)

Taking/ = 0.02 for all pipes, thefirst-iterationR values of the pipes are
obtained as shown in Table: Example 10.3. Using these R values, the head
loss equations for loops I . . . Ill are, respectively,

2O6602 + 326503 - 870701 = 0 (5)

-326503 - 77501 + 326501 = 0 (6)

1570? + 77501 - 20660? - 2040? + hp + (80 - 100) = 0

(7)
The head delivered by the pump, hp, in Equation (7) can be expressed
in terms of 0,. Thus, from Equation (2) of Example 9.7, we have

hp = 12.05 - 38.2(0, - 0.0353)2 (8)

Incorporating the value of hp from Equation (8), Equation (7) becomes

1570? + 77501 - 20660? - 2040?

+ 12.05 - 38.2(0, - 0.0353)2 - 20 = 0 (9)

On simplification Equation (9) becomes,

1570? + 77501 - 20660? - 242.20?

(10)
+ 2.6970, - 7.998 = 0
 TABLE: EXAMPLE 10.3.

Iteration 1 Iteration 2 Iteration 3

Pipe iOx i ' . iRx ifli Q«,, if, 2RX 2R'X Ox(2) 3Ox »f. 3RX 3R'X 0x,3,

1 1 0.02 204 204 0.1467 0.1467 0.01859 189.7 27.82 0.1779 0.1623 0.01853 189.0 30.68 0.1617
2 1 0.02 2066 2066 0.0433 0.0433 0.02180 2252 97.51 0.0581 0.0507 0.02167 2238 113.48 0.0511
3 1 0.02 8707 8707 0.0234 0.0234 0.02351 10235 239.5 0.0398 0.0316 0.02324 10118 319.72 0.0306
4 1 0.02 3265 3265 0.0350 0.0350 0.02316 3782 132.4 0.0292 0.0321 0.02323 3792 121.7 0.0327
5 1 0.02 775 775 0.1117 0.1117 0.02121 821.7 91.8 0.0911 0.1014 0.02125 823.3 83.48 0.1016
6 1 0.02 3265 3265 0.0615 0.0615 0.02388 3899 239.8 0.0510 0.0562 0.02392 3905 219.46 0.0567
7 1 0.02 157 157 0.2933 0.2933 0.01774 139.5 40.9 0.2621 0.2777 0.01776 139.7 38.80 0.2783
276 LINEAR THEORY METHOD

Taking ,<2, = . . . = , 0 7 = 1 for the first iteration, linearization of

Equations (5), (6), and (10) gives

206602 + 326504 - 870703 = 0 (11)

- 3 2 6 5 0 4 - 7750 5 + 32650 6 = 0 (12)

1570 7 + 7750 5 - 2O6602 - 242.20, + 2.6970, - 7.998 = 0

(13)

Note that , 0 , = 1 is used only for the linearization of nonlinear terms

and therefore 2.6970, in Equation (10) remains as 2.6970,.
Equations (1) . . . (4), and (11) . . . (13) are solved simultaneously and
the 0,(1, . . . 0 7 ( 1 ) values are obtained. Using the assumed 0 values, the
/ a n d R values are updated, and Equations (5), (6), and (10) are linearized
and solved simultaneously along with Equatioas (1) . . . (4). The iterative
procedure is continued and the first three iterations are shown in Table:
Example 10.3. The assumed values for iteration 4 would be practically the
same as the final answers obtained in Example 8.6.

10.2.4 NETWORKS WITH VALVES

The formulation of 0 equations for networks containing check valves

and pressure reducing valves is described in Chapter 7. The loops are se-
lected such that these valves are externalized.
In the case of a check valve, the flow through it is initially assumed to
be in the permitted direction, and the 0 equations are formulated. How-
ever, when the direction of flow reverses, the discharge in the pipe contain-
ing the check valve is made zero in the node-flow continuity relationships
for the end nodes of this pipe, and the loop-head loss equation for the loop
containing this pipe is dropped.
In the case of a PRV, it is initially assumed to be in the operative mode
and is therefore replaced by source and sink nodes. A pseudo loop is intro-
duced and the 0 equations are formulated and solved. For the assumed
discharge in the pipe with a PRV, a check must be made to ascertain that
the PRV has remained in the operative mode. However, if it is observed
that the PRV has become inoperative or the flow direction has reversed,
the procedure described earlier for check valve is followed.

Example 10.4

Solve Example 8.7 by linear theory method based on pipe-discharge

equations.
 Pipe Discharge Equations 277

Solution

The network of Figure: Example 8.7 is reproduced herein as Figure: Ex-

ample 10.4. Assuming that the PRV is in the operative mode, the PRV is
0-5m)s
" 5 0-4

5Q 6 o-6
replaced by source and sink nodes and a pseudo loop (loop II in Figure:
Example 8.7) is formed. Thus 0 equations for nodes 2 . . . 6 and loops I
and II are, respectively,
G. - Qi - & - 0.3 = o (1)

Q3 - Q< - Qy - 0.2 = 0 (2)

Qi + & - ft - 0.5 = 0 (3)

Qs - & - 0.4 = 0 (4)

Q6 + Q7 - 0.6 = 0 (5)

6QI - 40QI - 4QI = 0 (6)

20? + 6QI + 9 0 i + 12QI - 50Q27 + (80 - 100) = 0 (7)

Equations (6) and (7) are linearized and solved simultaneously with
Equations (1) . . . (5). Equations (6) and (7) are modified and the iterative
procedure is continued. Finally, the answers given in Example 8.7 are ob-
tained. As initially assumed, the PRV is in the operative mode.
When the PRV becomes inoperative for another pressure setting given
in Example 8.7, loop II is replaced by loop III shown in Figure: Example
8.7. Therefore, Equation (7) is replaced by

4003 + 901 + 1201 - 105.060? = 0 (8)

Note that the resistance constant of pipe 7 now includes the resistance
constant of the PRV also.
278 LINEAR THEORY METHOD

Equations (1) . . . (6) and (8) are solved by linear theory and the an-
swers given in Example 8.7 are obtained.
For the changed nodal flows in Example 8.7, the PRV behaves as a check
valve. Here, pipe 7 containing the PRV is dropped. Therefore, the term Q7
is dropped in Equations (1) . . . (6). Further, as loop II containing pipe 7
is now absent, Equation (7) or (8) for loop II is also absent. Therefore, the
required Q equations for basic unknowns Qi . . . Q6 are provided by
Equations (1) . . . (6) in which Q7 is absent. These six equations are
solved by linear theory and the answers given in Example 8.7 are obtained.

10.3 NODAL HEAD EQUATIONS

The nodal heads are taken as the basic unknowns in the formulation of
//equations as explained in Chapter 7. The application of the linear theory
method for the solution of these equations for different situations is ex-
plained herein.
10.3.1 NETWORKS WITH KNOWN PIPE RESISTANCES

Pseudo loops are not necessary in the formulation of H equations. These

equations, formulated for single- and multiple-looped networks are non-
linear, but are linearized using Equation (6.12). From the node-flow con-
tinuity relationship at nodes of known flow,

H — H\l/"
J
—^ -\ + qoj = 0, for a l l ; of known q3 (10.12)
/
connected to
j through x

in which //, = Hoi and //, = Hoj for source nodes. Therefore, for the tth
iteration, Equation (10.12) can be written as

RUn .(H, - Hj) + qoj = 0 , (10.13)

' for all j of known qy
connected to
j through x

in which Rox = known resistance constant of pipe x\,//,,,//, = known or

assumed nodal heads, for the t'h iteration, at nodes /, j , respectively; //,,
Hj = unknown nodal heads. ,
Expressing Equation (10.13) in the linearized form, we get

r C x '(//, - Hj) + qoj = 0, for ally of known qj (10.14)

i
connected to
j through x
 Nodal Head Equations 279

in which ,CX = modified conductance of pipe x for the fh iteration and is

given by

C" — ,x = 1 . . .X (10.15)
1^ x —

or

,0, x = l (10.16)

in which ,QX, thx = discharge and head loss in pipe x for the fh iteration,
respectively.
To begin the iterative procedure, it is necessary to initialize the dis-
charge and head loss in a pipe, i.e., to select the values of XQX and thx,
x = 1 . . . X. Collins and Johnson [125] suggest that tQx should be
chosen to correspond to the Reynolds number of 200,000. As Re =
AQI-KDV, we get

= 5O,0OOTDXI»,JC = 1 (10.17)

Taking r = l x 10" 6 m 2 / s , w e g e t

lrD
r> ' 1 v (10.18)
i Q* = - j o " , x = l . . . X

in which the discharge and diameter are in cubic meters per second and
meters, respectively.
Issacs and Mills [126] suggest that the initial discharge may be set to the
same value for all pipes. They observed that an initial pipe discharge rang-
ing from 0.001 m3/s to 1 m3/s, when used in a test network of sixty pipes,
had no significant effect on the solution [126]. Alternatively, they also sug-
gest using an initial velocity of 1 m/s in all pipes to initialize the pipe flows
[126-127]. (The initial velocity in the approach of Collins and Johnson is
0.2/D,.) The value of the initial head loss xhx corresponding to the
assumed discharge i Qx is then evaluated from

I A, = Ra = 1 . • •X (10.19)

Equation (10.16) now gives ,C,', x = 1 . . . Xand thus the nonlinear

node-flow continuity equations are linearized in the form of Equation
(10.14) for the first iteration. These linearized equations are solved to ob-
280 LINEAR THEORY METHOD

tain Hj values at the end of iteration 1, i.e., Hjm values. These Hj{X)
values are then used to determine C^(i) values from Equation (10.15) or
Equation (10.16). In the iterative procedure, if the values of ,C'X are taken
equal to ,C,' (t - 1) , overcorrection occurs as observed by Collins and John-
son. Therefore, they suggest the use of the average of the assumed and ob-
tained values. Thus, for the t'h iteration,

c, = , t - u c ; + <:;,_„ x=l x (1020)

The iterative procedure is continued until the assumed and obtained C'x
values and therefore the //, values are sufficiently close. The unknown
nodal flows are obtained in the end.
If the unknown nodal flows are also considered as basic unknowns in ad-
dition to the unknown nodal heads as done by Collins and Johnson, we get
H-q equations, one for each node. These equations are linearized and
solved iteratively so that all the unknown nodal heads and unknown nodal
flows are obtained simultaneously at the end of each iteration.
As shown by Collins and Johnson [125], a similarity exists between a
structural network and a hydraulic network such as a water distribution
network. These similarities are as follows:
(1) The algebraic sum of the forces at a joint of a structural network
must be zero to maintain equilibrium. Similarly, the algebraic sum of
the flows at a node of a hydraulic network must be zero to maintain
flow continuity.
(2) The displacement of the ends of the members meeting at a joint of a
structural network must be the same. Similarly, the heads at the ends
of the pipes meeting at a node of a hydraulic network must be the
same.
(3) The force-displacement relationship specified by the geometric and
elastic properties of the members must be satisfied in a structural
network. Thus,

F = Ku (10.21)

in which F = force in member with stiffness K, and u = displacement of

the member. In a hydraulic network, from Equation (10.16) we have

Q = C •h (10.22)

which is the linearized form of the discharge-head loss relationship for a

pipe. As seen from Equations (10.21) and (10.22), the discharge-head loss
 Nodal Head Equations 281

relationship of a pipe in a hydraulic network is similar to the force-

displacement relationship of a member in a structural network. Thus, sim-
ilarity exists between structural and hydraulic networks.
For structural networks, a finite-element method is well developed to
formulate and simultaneously solve Equation (10.21). Therefore, a finite-
element method can also be used for hydraulic networks to formulate and
simultaneously solve Equation (10.22). Collins and Johnson and later
other investigators [128,129] used this approach and therefore termed the
linear theory method based on linearized nodal-head equations as the
finite-element method.
The linear theory method based on H, and H-q equations is illustrated
in Example 10.5.

Example 10.5

Solve Example 8.3 (Example 8.9) by the linear theory method based on
(1) H equations and (2) H-q equations.

Solution

1. H Equations

In the network of Figure: Example 8.3, reproduced herein as Figure:

Head

Example 10.5, the linearized H equations written for the node-flow conti-
nuity relationship at nodes 3 . . . 6 are, respectively,

.d'OOO - H3) - ,Ci(H3 - 95) - 0.3 = 0 (1)

,C 2 '(100 - H4) + , C ( 9 5 - H4) - ,CUH4 - H5) - 0.2 = 0

(2)
282 LINEAR THEORY METHOD

,C 5 '(// 4 - Hs) + ,Ci{Hb - Hs) - 0.4 = 0 (3)

,C.'(95 - H6) - ,Ci(H6 - Hs) - 0.1 = 0 (4)

Equations (1) . . . (4) can be rearranged and expressed as, respectively,

i.Ci + ,Ci) • H3 = 100,C + 9 5 , a - 0.3 (5)

tCi + tCi) • H4 - ,CL = 100rC2' + 95 t C 4 ' - 0.2

(6)

-,d • H4 + (tCi + ,C4) • H5 - ,Ci • H6 = -0.4 (7)

- , d • Hs + (,Ce + ,C 7 ') • Ht = 95,C 6 ' - 0.1 (8)

Equations (5) . . . (8) expressed in the matrix form give

a +a o o o H3
o a +a +a -a o H4
o -CL c; + a -a Hs
o o -a a + c; H6

lOOC; + 95C3' - 0.3

100C2' + 95C4' - 0.2
(9)
-0.4
95C6' - 0.1

As the pipe diameters are not given, we assume the discharges in pipes
to be 0.1 m 3 /s. The corresponding hx and iC« values are obtained from
Equations (10.19) and (10.16), respectively, and are shown in Table: Exam-
ple 10.5. Therefore, for the first iteration Equation (9) becomes

0.885 0 0 0 H3 84.66 '

0 0.767 -0.177 0 H4 57.62 (10)
0 -0.177 0.319 -0.142 Hs -0.40
0 0 -0.142 0.496 H6 33.53
Solving Equation (10) we get H3O) = 95.66 m, HM1) = 96.98 m,
#5<n = 94.71 m, and H6{1) = 94.72 m. The corresponding values of ,/i x ,
tQx [using Equation (10.19)] and Cx'(1, [using Equation (10.16)] are ob-
 TABLE: EXAMPLE 10.5.

Iteration 1 Iteration 2

2 c;
R<>
Pipe A ic; *«2, Q*,2,
C
x(2,

1 40 0.1 0.5650 0.177 4.34 0.301 0.0694 0.1232 4.52 0.308 0.0681
2 20 0.1 0.2825 0.354 3.02 0.360 0.1192 0.2366 3.37 0.382 0.1133
3 10 0.1 0.1413 0.708 0.66 0.230 0.3486 0.5283 0.48 0.194 0.4036
4 30 0.1 0.4238 0.236 -1.98 -0.230 0.1162 0.1761 -1.63 -0.207 0.1271
5 40 0.1 0.5650 0.177 2.27 0.212 0.0934 0.1352 2.31 0.214 0.0927
6 20 0.1 0.2825 0.354 0.28 0.100 0.3554 0.3547 0.53 0.141 0.2651
7 50 0.1 0.7063 0.142 0.01 0.010 1.0013 0.5717 0.15 0.043 0.2885
284 LINEAR THEORY METHOD

tained. The average of , C'x and C*'m is taken as 2CX' for the second itera-
tion. Therefore, Equation (9) now becomes

0.6515 0 0 0 H3 62.208 '

0 0.5479 -0.1352 0 H4 40.189
0 -0.1352 0.7069 -0.5717 Hs -0.400
0 0 -0.5717 0.9264 Hb 33.596
(11)

Solving Equation (11), we get H3{1) = 95.48 m, H4(2) = 96.63 m,

#s<2) = 94.32 m, and H6{2) = 94.47 m. The first two iterations are
shown in Table: Example 10.5. The iterative procedure is continued and
the answers given in Example 8.3 are obtained.

2. H-q Equations

In the approach using H-q equations, the node-flow continuity relation-

ship is also applied to the nodes of unknown flow, i.e., to nodes 1 and 2.
Therefore, H-q equations for nodes 1 and 2 are, respectively,

- H3) - ,C 2 '(100 - H qt = 0 (12)

- 95) - ,C 4 '(95 - H4) - ,C,'(95 - + q2 = 0 (13)

Equations (12), (13), and (1) . . . (4) give the required linearized H-q
equations. These equations are expressed in the matrix form as

Ci + d 0 -Ci -Ci 0 0
0 a + c; + c6' -Ci -c; 0 -Ci
— Ci -Ci ci + a 0 0 0
-Ci C '
L-4
0 a +-Cia + a -Ci 0
0 0 0 C :.' + c; -Ci
0 — Ci 0 0
-a Ci + Ci

100"
95
H3 -0.3
(14)
H4 -0.2
H, -0.4
H6 -0.1
 Nodal Head Equations 285

Using the XC'X values from Table: Example 10.5, Equation (14) becomes

0.531 0 -0.177 -0.354 0 0

0 1.298 -0.708 -0.236 0 -0.354
0.177 -0.708 0.885 0 0 0
0.354 -0.236 0 0.767 -0.177 0
0 0 0 -0.177 0.319 -0.142
0 -0.354 0 0 -0.142 0.496

100 ' •

95 q2
H3 = -0.3
X (15)
H4 -0.2
H5 -0.4
H6 -0.1

Solving Equation (15), we get qt = 1.84 m3/s, q2 = -0.84 m3/s, H3 =

95.66 m, H4 = 96.98 m, H, = 94.71 m, and H6 = 94.72 m. The iterative
procedure is continued and after several iterations, the answers given in
Example 8.3 are obtained.
In the network of this example, at the different nodes, either the flow is
unknown or the head is unknown. This is the usual case in practice where
the heads are known and flows unknown at source nodes and flows are
known and heads unknown at demand nodes. For such networks, the for-
mulation of linearized H-q equations in the matrix form of Equation (14)
is simple and the equations can be written directly as follows:
In the coefficient matrix of Equation (14), the element on the main diag-
onal au, i = j , represents the sum of the modified conductances of the
pipes connected to a node under reference, and the other nondiagonal ele-
ments atJ, i =£ j , are either zero if the two nodes are not connected or rep-
resent the negative modified conductance of the pipe connecting the two
nodes. For example, for node 4, the diagonal element (/ = j = 4) is
C2' + C4' + C5' because node 4 is connected to nodes 1, 2 and 5, through
pipes 2, 4, and 5 with modified conductances of C 2 \ C4' and C s \ respec-
tively. Similarfy, as node 4 is connected t' • node 1 through pipe 2, the ele-
ment a,-4 = - C 2 ' . Thus, the coefficien. matrix can be written directly
from the geometry of the network. Naturally, the coefficient matrix is sym-
metrical on the main diagonal, and the row sums and column sums are
each zero.
The second matrix in Equation (14) is a column matrix for nodal heads,
286 LINEAR THEORY METHOD

known at source nodes and unknown at demand nodes. The column matrix
on the right-hand side represents the known flows at demand nodes and
unknown flows at source nodes, the preceding sign for the nodal flows be-
ing negative for outflows and positive for inflows.
The procedure for writing the coefficient matrix as described above for
H-q equations [Equation (14)] can also be followed for writing the coeffi-
cient matrix for H equations [Equation (9)]. However, when the node
under reference, node^, is connected to another node of known head, node
i through pipe ij, the corresponding element of the column matrix on the
right-hand side is increased by the product of Hoi and C,y. For example, as
node 4 is connected to nodes 1 and 2 of known heads of 100 m and 95 m,
through pipes 1 and 4, respectively, the corresponding element in the
matrix on the right-hand side becomes lOOCV + 95C4' - 0.2 as shown in
Equation (9). The coefficient matrix for H equations is symmetrical but
the row sums and column sums are no longer zero.

10.3.2 NETWORKS WITH UNKNOWN PIPE RESISTANCES

The formulation of H-q-R equations, reducible to H-R, H, and H-q

equations is described in Chapter 7. When H-q-R equations are formu-
lated, the formulation is implicit and the R values are updated at every iter-
ation. In the explicit approach, the R values are obtained after satisfactory
convergence is reached for H values.

10.3.2.1 Implicit Approach

Equation (6.2) giving the head loss in a pipe is linearized for Rx and
therefore can be expressed as

or

f R, (10.24)

or

Qx = | ,B, | • Rx (10.25)
 Nodal Head Equations 287

in which

BB- iQ' (10.26)

• * - \x
The linearized node-flow continuity equations are formulated using
these equations as illustrated in the following example.

Example 10.6

Solve Example 8.4 (Example 8.10) by the linear theory method based on
(1) implicit and (2) explicit approaches.

Solution

1. Implicit Approach

In the network of Figure: Example 8.4, shown herein as Figure: Exam-

ple 10.6, the basic unknowns are qit H3, H4, R4 and Hs. The linearized H-
0-5 m%.

q-R equations written for the node-flow-continuity relationship at nodes

1 . . . 5 are, respectively,

,C,'(100 - H4) + t C 2 '(200 - H3) = q, (1)

,C 5 '(98 - H5) + ,C 6 '(98 - H3) = 0.6 (2)

,C 2 '(100 - H3) - ,CUH3 - HA) + ,C 6 '(98 - H3) = 0.6 (3)

,C,'(100 - H4) + ,C3{H3 - H4) - ,B4 • R4 = 0.5 (4)

,B4 • R4 + ,C 5 '(98 - Hs) = 0.7 (5)

288 LINEAR THEORY METHOD

Equations (1) . . . (5) can be expressed in the matrix form as

1
c( + a — C -ci 0 0 ioo'
o — C' 0 0 -Ci H3
a + a\ a. -Ci 0 0 H4
-Ci c; + a B4 0 /?4
0 o -B4 Ci . Us ,

0.6 - 98(C5' + i
98C6' - 0.6 (6)
-0.5
98C5' - 0.7

Initially, let the t Qx values be 0.1 m3/s. The corresponding xhx and i C'x
values are evaluated and shown in Table: Example 10.6. Therefore, for the
first iteration Equation (6) becomes

2.25 -1 -1.25 0 0 100

0 -1.25 0 0 -0.25 H3
-1 2.614 -0.364 0 0 H4
1.25 -0.364 1.614 0.005 0 R*
0 0 0 -0.005 0.25 , , H5 ,

146.4
121.9 (7)
-0.5
23.8

Solving Equation (7) we get q1(l) = 1.2025 m3/s, H3{1) - 98.81 m,

#4(1) = 99.99 m, /?4(i, = -183.17, and H5(1) = 91.54 m. Using these
values, the second iteration is carried out from which we get qH2) =
1.200 nrVs, // 3 ( 2 ) = 98.685 m, H4{2) = 99.853 m, R4(1) = -150.33,
and #5(2) = 90.551 m. The first two iterations are shown in Table: Exam-
ple 10.6. Continuing the iterative procedure, the answers given in Example
8.6 are obtained.

2. Exploit Approach

In the explicit approach, the linearized H equations are formulated at

nodes 2, 3, and merged-node 4-5. These equations are, respectively,

- , C 6 ' ( 9 8 - H3) - t C 5 '(98 - H5) + 0.6 = 0 (8)

 TABLE: EXAMPLE 10.6.

Iteration 1 Iteration 2

Pipe iO, A ,c; "x,1» Ox,,, C

x(1) 2 c; "x,2, Ox,2) C"'

1 8 0.1 0.08C 1.250 0.01 0.0354 3.536 2.3930 0.147 0.1356 0.9221
2 10 0.1 0.100 1.000 1.19 0.3450 0.290 0.6450 1.315 0.3626 0.2758
3 27.5 0.1 0.275 0.364 -1.18 -0.2071 0.1755 0.2697 -1.168 -0.2061 0.1764
5 40 0.1 0.40C 0.250 6.46 0.4019 0.0622 0.1561 7.449 0.4315 0.0579
6 8 0.1 0.08C 1.250 -0.81 -0.3182 0.3928 0.8214 -0.685 -0.2926 0.4272

Iteration 1 Iteration 2

Pipe iff. lOx iS* Ox,,) 8

X(1) 2BX Hx{2) e,(2>
4 20 0.1 0.005 -183.17 8.45 -0.2148 0.00117 0.00308 -150.33 9.302 -0.2488 0.00165
290 LINEAR THEORY METHOD

t C 2 '(100 - H3) - tCi(H3 - H4) + t C 6 '(98 - H3) - 0.6 = 0 (9)

,C,'(100 - H4) + tCi(H3 - H4) + ,C S '(98 - Hs) - 1.2 = 0 (10)

These equations expressed in the matrix form are

-C«' 0 -C5' H3

c2' + a + c6' -a o H4

^--3 (-'l ~r C3 C-5

0.6 - 98(C5' + Ci)

100C2' + 98C6' - 0.6 (11)

lOOC + 98C5' - 1.2 ,

These equations are iteratively solved and the answers given in Example
8.6 are obtained.

10.3.3 NETWORKS WITH PUMPS

The formulation of nodal head equations for networks with pumps has
been explained in Chapter 7. As stated therein, it is preferable to use the
head-discharge relationship that is consistent with the head loss relation-
ship for the pipes.

Example 10.7

Solve Example 8.6 (Example 8.11) by the linear theory method based on
head equations.

1
Sump ESR.
W.L=8O-0Om WL=i00-00m

0-12
 Nodal Head Equations 291

Solution

The linearized node-flow continuity equations for nodes 3 . . . 6 of Fig-

ure: Example 8.6, reproduced herein as Figure: Example 10.7 are

,C,'(80 + hp - H3) - tCHH3 - HA) - ,C3'(H3 - Hs) - 0.08 = 0

(1)

,Ci(H3 - H4) - ,Cl(H4 - Hs) + ,Ci(H6 - H4) - 0.12 = 0 (2)

,d(H3 - Hs) + ,d(H4 - Hs) + ,Ci(Ht - Hs) - 0.12 = 0 (3)

,C,'(100 - He) - ,Cs(H6 - H4) - tCi(H» ~ Hs) - 0.12 = 0 (4)

Equations (1) . . . (4) can be expressed in the matrix form as

c + a + c, -a -a o
-a a + a + CL -C; -a
-a -a a +a +a -a
o -c -a CL + a + a

Hi C,'(80 + hp) - 0.08

-0.12
-0.12
(5)
H6 100C7' - 0.12

The H equations in the matrix form as given by Equation (5) could have
been directly written as explained in Example 10.5. Herein, the sump head
of 80 m is increased by the pump head hp in the first term of the matrix on
the right-hand side.
Initially, the Reynolds number is taken as 200,000 in all pipes as pro-
posed by Collins and Johnson [125]. Therefore, discharges in pipes, i.e.,
the iQ, values, are obtained using Equation (10.18), and are shown in
Table: Example 10.7. Corresponding to these discharges, the lfx values are
obtained from the Swamee-Jain pipe-head loss explicit relationship. The
corresponding thx values are obtained from the Darcy-Weisbach head loss
relationship, and finally the XC'X values are obtained using Equation
(10.16).
The head developed by the pump is given by

hp = 12.05 - 38.2(&, - 0.0353) 2 (6)

 TABLE: EXAMPLE 10.7

Iteration 1 Iteration 2

Pipe A '*> C
x(1> 'x(2)

1 0.04712 0.01976 0.4476 0.10527 0.693 0.02009 0.05815 0.08390 0.09458 1.115 0.01902 0.07581 0.06799
2 0.03142 0.02213 2.2570 0.01392 -1.072 0.02245 -0.02150 0.02006 0.01699 0.353 0.02354 0.01205 0.03413
3 0.02356 0.02350 5.6785 0.00415 1.889 0.02400 0.01345 0.00712 0.00563 3.428 0.02359 0.01827 0.00533
4 0.02356 0.02350 2.1295 0.01106 2.961 0.02318 0.02798 0.00945 0.01025 3.075 0.02316 0.02852 0.00927
5 0.03142 0.02213 0.8463 0.03713 4.516 0.02129 0.07400 0.01639 0.02676 5.437 0.02124 0.08129 0.01495
6 0.02356 0.02449 2.2192 0.01062 7.477 0.02391 0.04377 0.00585 0.00823 8.512 0.02388 0.04673 0.00549
7 0.05498 0.01930 0.4590 0.11978 3.065 0.01804 0.14697 0.04795 0.08386 4.001 0.01794 0.16837 0.04208

*P
— — 12.04 — 12.03 — — — — 11.99 — — —
 Nodal Head Equations 293

Since Qp = ,Qt = 0.04712 m3/s, ,/zp = 12.04 m. Substituting the values

for ihp and Equation (5) becomes

0.12334 -0.01392 -0.00415 0

0.01392 0.06211 -0.01106 -0. 03713
0.00415 -0.01106 0.02583 -0. 01062
0 -0.03713 -0.01062 0.16753

H3 9.609
H4 -0.120
(7)
H5 -0.120
H6 11.858
Solving these equations H3(1) = 91.347 m, # 4 ( , , = 92.419 m,
#5(i> = 89.458 m, and H6(l) = 96.935 m. These HjW values are used to
determine hx{1) values, fz(1) values with the help of Equation (2.40), and
then QxM values with the help of Equation (2.38). The Cx'(i> values are
obtained from Equation (10.16).
For the second iteration the 2CX values are taken as the average of the
XC'X and C7(i) values as given by Equation (10.20) and the iterative proce-
dure is continued. Table: Example 10.7 shows the first two iterations.
The iterative procedure is continued and the answers given in Example
8.6 are obtained.

10.3.4 NETWORKS WITH VALVES

The formulation of H equations for networks with check valves is de-

scribed in Chapter 7. These equations are linearized so that the discharge
in pipe ij, connecting nodes / andy, and containing a check valve with per-
missible direction of flow from / toy is given by C,y(//, — / / , ) . During the
iterative procedure if (//, — H}) becomes negative, C'tj is made zero so
that the discharge in the pipe is zero.
For a network with a pressure reducing valve, it is initially assumed that
the PRV is in the operative mode and therefore it is replaced by source and
sink nodes, and the H equations are formulated as explained in Chapter 7.
These equations are linearized and iteratively solved. A check is made at
the end of each iteration to see that the PRV has remained in the operative
mode. If the PRV becomes inoperative, or the direction of flow reverses,
the PRV behaves like a check valve.

Example 10.8

Solve Example 8.7 (Example 8.12) by the linear theory method based on
H equations.
294 LINEAR THEORY METHOD

Solution

The network of Figure: Example 8.7 is reproduced herein as Figure: Ex-

ample 10.8. Initially it is assumed that the PRV is in the operative mode.

5 0-4

100-00m

0-2"3 50 g 0-6

Using the known head of 100 m at node 1 and the set pressure head of 80
m at d, the downstream of the PRV, the linearized H equations for nodes
2 . . . 6 of the network for the t'h iteration are, respectively,

,C,'(100 - H2) - tCi{H2 - H4) - tCi(H» - H3) - 0.3 = 0 (1)

,C 3 '(fl 2 - H3) - ,Ci(H3 - H4) - ,C.U(80 - H6) - 0.2 = 0 (2)

,C2'(H2 - H4) + , c ; ( f t - H4) - ,C 5 '(// 4 - Hs) - 0.5 = 0 (3)

,CUH4 - Hs) - , C 6 ' ( # 5 - H6) - 0.4 = 0 (4)

t C;_ 6 (80 - H6) + tCi(Hs - H6) - 0.6 = 0 (5)

Equations (1) . . . (5) are expressed in the matrix form as

Ci — Ci 0 0
c +a + a -Ci -C'd.
-c: ci + a 0
-Ci a + a + ci -Ci 0
0
-a0 -Ci
0
-ci a + Ci
0 0 a +c
ft 100C - 0.3
ft • -80CJ-6 - 0.2
X ft ,= -0.5 (6)
ft -0.4
80C'-6 - 0.6 .

As the PRV is situated just downstream of node 3, C j . 6 = C 7 '. For the

first iteration, taking XQX = . . . = VQ7 = 0.1 m 3 /s, the ,C X ' values are
 Relationship between Newton-Raphson and Linear Theory Methods 295

determined and solving Equation (6) we get the values of H2 <, ( . . . H6{1).
The discharge in pipe 7 is evaluated and the pressure head at the down-
stream end of the PRV is determined to check the mode of the PRV. For
the given nodal flows and the pressure head setting of the PRV, the PRV
remains in the operative mode throughout the iterative procedure and the
final solution as given in Example 8.7 (Example 8.12) is obtained.
At the pressure head setting of 89 m, the PRV becomes inoperative dur-
ing the iterative procedure. At this stage, CJ. 6 is replaced by C 7 ', corre-
sponding to the increased resistance from R7 to R7 + /?PRV, and the
known pressure head setting of 89 m is replaced by unknown pressure
head H3. Equation (6) therefore becomes

:» + ci •Ci -C2' 0 0
Ci a +-Cia + a -Ci 0 -Ci
Ci a +-Cia + a -Ci 0
0 0 ci + a -Ci
0 -Ci 0 -Ci ci + a
H2 ' 100C,' - 0.3
H3 -0.2
X Ht -0.5 (7)
H5 -0.4
u -0.6

Equation (7) can be written directly as explained earlier, remembering

that Ci includes the effect of the PRV.
For the changed nodal demands, H3 becomes less than H6 during the
iterative procedure. Herein the PRV behaves as a check valve and prevents
flow from node 6 to node 3. Therefore, CJ. 6 and C7' become zero in Equa-
tions (6) and (7) respectively.

10.4 RELATIONSHIP BETWEEN NEWTON-RAPHSON AND LINEAR

THEORY METHODS [131]

As seen earlier, the Newton-Raphson and linear theory methods simul-

taneously solve the concerned nonlinear equations by linearizing them.
The Newton-Raphson method linearizes the nonlinear equations through
partial differentiation while the linear theory method linearizes the non-
linear equations by merging a part of the nonlinear term in the resistance
or conductance of the pipes. Let us now study the relationship between the
Newton-Raphson and linear theory methods.
296 LINEAR THEORY METHOD

10.4.1 PIPE DISCHARGE EQUATIONS

Consider a pipe with resistance constant Ro carrying discharge Q so that

the head loss in the pipe is ROQ". Considering the first iteration, let the as-
sumed discharge in the pipe be i Q and the correction be AQ so that

Ro(iQ + AQ)" = ROQ" (10.27)

According to the Newton-Raphson method, we get

RoUQ" + n • .Q"'1 • A0 N R ) = R0Q" (10.28)

or

= g". Q3
~3" (10.29)

in which AQNR = correction obtained according to Newton-Raphson

method.
According to the linear theory method, we get

^.(.G""')G(.) = RoQ" (10.30)

in which Q( i, is the obtained discharge in the pipe at the end of iteration

1, as stated earlier. If AQLT represents the correction to , Q according to
the linear theory, then Q(l) = ,<2 + AQu, and thus Equation (10.30)
becomes

K o d G - ' X i Q + AGLT) = R°Q" (10.31)

from which

AGLT =

Thus, from Equations (10.29) and (10.32), we get

AQLT ,= n • A0 NR (10.33)

As n = 1.85 or 2, it is evident that the absolute value of the correction

obtained by the linear theory method is larger than that obtained by the
Newton-Raphson method and therefore oscillations occur. Thus, to pre-
vent oscillations, if the correction is taken equal to that given by the
 Relationship between Newton-Raphson and Linear Theory Methods 297

Newton-Raphson method, the assumed discharge for the second iteration

in the linear theory method can be taken as

2<2=IQ+^AQLT (10.34)

or

2 0 = ,<2 +£[&.> - .Gl (10.35)

For the Darcy-Weisbach and Manning head loss formulas, n = 2, so

that Equation (10.35) gives

( 1 0 .36)
z* - 2

Thus, for the fh iteration, the assumed discharge in pipe x is given by

,Q, = 2 (10.37)

When the Hazen-Williams head loss formula is used, n = 1.852 so that

Equation (10.35) leads to

,QX = 0.46,,-,, Qx + 0.540,,,-,, (10.38)

However, because the difference between Equations (10.37) and (10.38)

is not much, Equation (10.37) can be uniformly used throughout the linear
theory method irrespective of the head loss formula.
Equation (10.37) is the same as Equation (10.9). Since Newton-Raphson
method has quadratic convergence, the use of Equation (10.9) improves
convergence, as stated earlier.

10.4.2 NODAL HEAD EQUATIONS

Consider a pipe with resistance constant Ro connecting nodes / and j

with heads //, and Hj (Hi > //,) so that the discharge Q in the pipe is given
by
298 LINEAR THEORY METHOD

Let the head Ht be known and fixed and the assumed head at nodey for
the first iteration be tHj with correction A//, so that

or
- lHJ) - AH^'"
(10.41)

According to the Newton-Raphson method, we get

(10.42)

Therefore, we have

(H — HY'n — (H —lJ ,HV"

= -n - ^ ^ ^i ^— (10.43)

in which the subscript NR is introduced for A J/, to indicate that it is a cor-

rection according to the Newton-Raphson method.
According to the linear theory method, we get

:F M - it,.,l - U0.44,

Since HJ(i) = ,//, -I- A//,_LT, from Equation (10.44), we get

(Hi - .Hjf'1 • KH - tHj) - A//y_LT] = (//, - // 7 ) 1

(10.45)

from which we get

AHj_LT = _ (//- 46)

 Relationship between Newton-Raphson and Linear Theory Methods 299

Thus, from Equations (10.43) and (10.46), we get

AHj.LT = £ • AHj. NR (10.47)

Because n = 1.85 or 2, it is evident that the absolute value of the cor-

rection obtained by the linear theory method is numerically less than that
obtained by the Newton-Raphson method. Although this does not produce
oscillations, the convergence is delayed. Therefore, to enhance conver-
gence in the linear theory method, the obtained //, value, i.e., HJ{i),
should be modified by taking the correction n • AHj instead of the ob-
tained correction AHj (the subscript LT is now dropped for simplicity).
Thus, the modified HJW value, i.e., HJO)m value is given by

HJiUm = lHJ + n- AHj (10.48)

or

HJWm = ,Hj + n[HJO) - tHj] (10.49)

These HJ{1)m values are used to calculate the modified values, i.e.,
hx{i)m, Qx{nm and Cx'(1)m values, so that

lC; + C;
2 C = 2 """ (10.50)

Further, instead of taking n = 1.85 or 2 depending upon the head loss

formula, if n is uniformly taken as 2, from Equation (10.15) we get

Similarly, from Equation (10.48), we get

_ [ / / • ( , / / , + 2AHj)]
*-x(l)m — DO.5 (IV.

or

R'o
300 LINEAR THEORY METHOD

Substituting the values of ,C I ' and CV(1)m from Equations (10.51) and
(10.52), respectively, in Equation (10.20), we get

s
(H, - tHj)-°- + [(//, - 1HJ) -
C - - (10.54)

Expanding the term within brackets in the Taylor's series and consider-
ing the first two terms, we get

_ 1
2
* 2
(10.55)

or

(10.56)

Now, the term on the right-hand side of Equation (10.56) represents the
first two terms of expansion in Taylor's series of

i.e., Cx'(i). Therefore, we get

2C; = C;,,, (10.57)

Thus, for the /'* iteration, instead of using the average of the assumed
and modified obtained values in the previous iteration, the obtained value
in the previous iteration can be used. Therefore, for the fh iteration,

,c; = c;,,-!, (io.58)

Thus, for fester convergence, Equation (10.58) should be used in place of

Equation (10.20). This is similar to the observation of Greenlaw and Clack
[132] when they stated that in some cases the most recent result is closer
to the final result than the average of the two most recent results as pro-
posed by Wood and Charles.
This principle can also be used for unknown pipe resistances.
Table 10.1 shows the results obtained by using Equations (10.20) and
(10.58) for the next two iterations, i.e., iterations 3 and 4 of Example 10.6.
 Problems 301

TABLE 10.1 . Comparison of Convergence Characteristics for Example 10.6.

rC; according to

Equation (10.20) Equation (10.58)

H3 H4 R* «s H3 HA RA H5
Iteration m m m m m m

3 98.53 99.68 -98.36 89.28 98.19 99.24 10.64 86.21

4 98.37 99.46 -37.36 87.96 97.93 98.24 1.50 86.17

As can be seen, convergence is improved when Equation (10.58) is used in-

stead of Equation (10.20). However, as stated earlier, overcorrection may
occur as observed by Collins and Johnson [125]. It is therefore suggested
that Equation (10.58) should be initially used for faster convergence, and
when oscillations occur, Equation (10.20) should be used. This approach
is similar to the application of the reduced correction when the correction
sign changes in the Hardy Cross method [Equation (8.31b)].

10.5 PROBLEMS

Problems 10.1 . . . 10.8: Analyze the networks described in Problems

7.1 . . . 7.8 by the linear theory method based on (a) pipe-discharge equa-
tions and (b) nodal-head equations.
 111

Other Analysis Methods

11.1 INTRODUCTION

N addition to the commonly adopted methods of network analysis, i.e.,

I the Hardy Cross, Newton-Raphson and linear theory methods described
earlier in Chapters 8, 9, and 10, respectively, some other methods have
also been tried and suggested. As described by McPherson [133], attempts
have been made to predict the prototype network behaviour through stud-
ies on scale models using pinchcocks [134], short capillary tubes within
large tubing [135], small orifices [136], and glass tubing [137]. However,
these models can provide qualitative information only, because the flow in
them is mostly laminar while the flow in the prototype pipes is almost al-
ways turbulent. Naturally, these methods have no practical significance.
Similarly, the approaches using special accessories consisting of wooden
disks and paper scales, as developed by Cobb [138,139] have not found
favour in practice.
Of some interest, however, are three methods: (1) analysis using electri-
cal analyzers, (2) analysis through unsteady behaviour during start-up, and
(3) analysis through optimization. These methods are briefly described in
this chapter.

11.2 ANALYSIS USING ELECTRICAL ANALYZERS

11.2.1 HYDRAULIC AND ELECTRICAL ANALOGIES [140-142]

Electrical analyzers are based on certain analogies that exist between

hydraulic networks and electrical networks. These analogies are explained
with the help of Figure 11.1. Figure ll.l(a) shows a two-source, six-demand
node, two-loop hydraulic network while the analogous electrical network
is shown in Figure ll.l(b). The various analogies are as follows:
303
304 OTHER ANALYSIS METHODS

4 0
(a) . (b)
Figure 11.1 Networks: (a) hydraulic network; (b) analogous electrical network.

(1) The discharge Q through the pipe of a hydraulic network corre-

sponds to the current / through a resistor of an electrical network.
Thus, the discharge Qy in pipe 1 of the hydraulic network shown in
Figure ll.l(a) corresponds to the current lx in resistor 1 of the electri-
cal network shown in Figure ll.l(b). The discharge Q in a pipe can
be measured by inserting a flowmeter in it; similarly, the electric
current through a resistor can be measured by inserting an ammeter
in series with the resistor.
(2) The resistance constant R of a pipe corresponds to the electric resis-
tance R of a resistor.
(3) The head loss h in a pipe due to its resistance constant R corresponds
to the voltage drop V along a resistor due to its electrical resistance
R. Thus, head loss /z, in pipe 1 corresponds to voltage drop K, in
resistor 1. Just as the head loss in a pipe can be measured by a
manometer connected at its ends, the voltage drop in a resistor can
be measured by a voltmeter connected across its terminals.
(4) The node-flow continuity relationship—the algebraic sum of the
flows at a node of a hydraulic network is zero—corresponds to
Kirchoffs first law—the algebraic sum of the currents towards a junc-
tion of an electrical network is zero. Thus, Q3 + Qs — Qb —
Q7 = 0 for node 5 of the hydraulic network corresponds to / 3 +
/ 5 — h — l7 — 0 for junction 5 of the electrical network.
(5) The loop-head loss relationship—the algebraic sum of the head
 Analysis Using Electrical Analyzers

losses along a loop of a hydraulic network is zero—corresponds to

KirchofFs second law—the algebraic sum of the voltage drops around
a closed circuit of an electrical network is zero. Thus, h' + h, +
h3 — hs = 0 for loop III of the hydraulic network corresponds to
V + V, + V3 - Vs = 0 of circuit III of the electrical network.
Because of these analogies, it is possible to use electrical networks to
simulate hydraulic networks. Such electrical networks, termed electrical
analyzers can therefore be used for the analysis of flow through water dis-
tribution networks. However, there is one basic difference between the be-
haviour of a hydraulic network and that of an ordinary electrical network.
The head loss in a pipe of a hydraulic network is given by a nonlinear rela-
tionship, h = RQ" in which the exponent/! is usually 1.85 or 2.00. For or-
dinary electrical networks, however, the voltage drop is given by a linear
relationship

V = RI (11.1)

11.2.2 ELECTRICAL ANALYZERS WITH LINEAR RESISTORS

Camp and Hazen [143] were perhaps the first to introduce electrical ana-
lyzers for the analysis of flow through hydraulic networks. The analyzers
used linear resistors (V = RI) to represent the pipes of the hydraulic net-
works. To obtain the analogy of the behaviour of hydraulic networks in
these analyzers, the resistances of the linear resistors are successively ad-
justed by trial and error so that ultimately the resistances of these resistors
correspond to RQ"'1 for all pipes. Thus, the network analysis through ana-
lyzers with linear resistors is similar to network analysis through linear
theory. The linear theory linearizes and solves the nonlinear equations
mathematically, while an electrical analyzer with linear resistors linear-
izes and solves the nonlinear equations electrically. Because of the neces-
sity of successive adjustments, these analyzers give the result indirectly.
Therefore electrical analyzers with linear resistors are also termed as in-
direct-reading electrical analyzers.
Although satisfactory results can be obtained with these analyzers for
networks having a large number of nodes and pipes, these analyzers suffer
from the following limitations:
(1) The analyzers involve a trial-and-error procedure.
(2) There is no guarantee of balancing and therefore of convergence.
(3) Because each pipe is represented by different resistors at various
stages of balancing, the resistances of the resistors keep on changing.
No resistor can be assigned a definite resistance. Thus, the storage
and proper selection of the resistors become difficult.
306 OTHER ANALYSIS METHODS

(4) The first settings of the linear resistors give preliminary values of
currents and voltages. These values must be recorded and adjusted
for all the resistors of the network, even though the discharge and
head loss in only a few pipes may be required.

The analyzer of Camp and Hazen [143] used dc supply but later Camp
[135] alternatively suggested the use of ac supply. Reid and Wolfenson
[144] have described the general features of these analyzers. Perry et al.
[145], and Suryaprakasam et al. [146] have offered valuable techniques for
speedy convergence. Application of these analyzers to gas networks has
been described by Clennon [147].

11.2.3 ELECTRICAL ANALYZERS WITH NONLINEAR RESISTORS

Mcllroy [140-142] developed electrical analyzers using nonlinear resis-

tances. For these nonlinear resistors,

V = Rl" (11.2)

By proper selection of the resistances, it is possible to obtain nearly the

same value of the exponent n as that obtained for hydraulic networks, i.e.,
1.85 for Hazen Williams head loss formula and 2.00 for the Darcy-
Weisbach and Manning formulas. Thus, electrical networks using such
nonlinear resistors, commercially termed fluistors, can represent hydraulic
networks.
A fluistor (Figure 11.2) consists of a straight uncoiled tungsten filament
supported by nickel leads in an evacuated glass bulb. Tungsten is a suitable
material for the filament because tungsten is available in different diame-
ters, is stable, has a long life, and also has a nonlinear resistance property.

Figure 11.2 Nonlinear resistors (fluistors).

 Analysis Using Electrical Analyzers 307

The resistivity of the tungsten filament depends on its temperature, and the
distribution of the temperature along the filament is affected by radiation
of heat to the surroundings and conduction of heat to the leads. As the cur-
rent through a filament increases, the temperature rises which in turn in-
creases the resistance. This causes the voltage to drop as the desired-
nonlinear function of the current. Fluistors resemble incandescent lamps
but are primarily designed to give nonlinear resistance characteristics
rather than to serve as an efficient source of light.
Tungsten filaments provide nonlinear characteristic up to their melting
point reached at about 12 volts. However, the normal working range is kept
between 0 to 5 volts for longer life. Test results have shown that the ob-
tained nonlinear characteristic of the fluistor matches excellently with the
required nonlinear characteristic except for voltages lower than one. How-
ever, these lower voltages would cause an overall error of about 0.5 percent
for the entire network.
The average voltage drop per fluistor is kept at about 2.5 V per fluistor
in the normal operation of the analyzers. Therefore, a hydraulic network
having 40 pipelines in series as in a 20 X 20 grid network can be
simulated by an electrical network having approximately 100 V across its
most widely separated terminals. Networks representing 120 pipes in
series can operate on 300 V without resorting to ac current and isolating
transformers.
Fluistors having approximately 200 different coefficient values ranging
from 0.058 to 500 (a ratio of roughly 1:9,000) would serve as a normal
stock to simulate hydraulic networks. Fluistors are designed so that resis-
tances of successive fluistors are spaced at approximately 5 percent, i.e.,
the resistance of each successive fluistor is approximately 1.05 times that
of the preceding lower one. This practice enables the selection of a fluistor
within an accuracy of about 2.5 percent in its resistance. As the resistance
of a fluistor resembles the resistance constant of a pipe which in turn de-
pends upon its length, diameter, and roughness characteristic, a single
fluistor can represent pipes of different lengths, diameters, and roughness
characteristics, but having the same value for their resistance constants.
For a given voltage drop across the filament, the temperature at the mid-
point of the filament is approximately the same for all fluistors irrespective
of the value of the actual resistance of the fluistor. Therefore, the bright-
ness of the filament can serve as an indication of the voltage drop across
the filament and thereby the head loss in the corresponding pipe. Thus, the
operator, simply by observing the filaments, can know the portions of the
network having large head losses. He may measure only these voltage
drops, estimate corresponding head losses, and suggest suitable changes,
if necessary, in the network.
V. Raman [148] has successfully used small commercial lamp bulbs and
308 OTHER ANALYSIS METHODS

auto lamp bulbs as nonlinear resistors for small networks and provided a
description of the fabrication and performance of such analyzers.
The analyzers with nonlinear resistors are analogous to hydraulic net-
works in their nonlinear behavior and can directly give the required results.
They are therefore called direct-reading electrical analyzers. Because they
were first suggested by McDroy, they are also known as Mcllroy's electrical
analyzers.
Mcllroy has described in detail several aspects such as the preparation
of the network map, selection of voltage-head loss ratios, current-flow rate
scales, operation of the analyzers, and interpretation and accuracy of the
analyzer results.
Network analysis through electrical analyzers does not involve any com-
putational work, provides immediate visual indication of the sections of a
network where head losses are excessive, and also provides ease in consid-
ering alternate designs and different operating conditions. The Hardy
Cross method was the only systematic method prevalent for network anal-
ysis when electrical analyzers were developed in the 1940s, and therefore
electrical analyzers became quite popular. According to McPheron [133],
eighteen Mcllroy electrical analyzers were in use in 1961 in the U.S.A.
These included those at Cornell, Washington, and other universities, used
for research and consultancy work; at Baltimore and Philadelphia, used
for water distribution networks; at southern California and other gas com-
panies, used for gas utility networks; and at the U.S. Bureau of Mines,
used for mine ventilation studies. Several investigators have described the
performances of these electrical analyzers [149-152].
The electrical analyzers require a special setup and thus have high initial
cost. They also require the services of trained personnel. Their accuracy
is within ± 3 percent of the total head loss in a network. Therefore, when
digital computers became readily available, because they provided faster,
cheaper, and more accurate results and could also consider elements other
than pipes in network analysis, they replaced electrical analyzers. Because
of the further advancement in computer technology and also the introduc-
tion of Newton-Raphson and linear theory methods as additional network-
analysis tools, electrical analyzers are almost obsolete as far as the analy-
sis of flow in water distribution networks is concerned.

11.3 ANALYSIS THROUGH UNSTEADY BEHAVIOUR

DURING START-UP

11.3.1 BASIC PRINCIPLES

Fox and Keech [153] considered the steady state of flow in a distribution
network as an ultimate stage of the unsteady flow developed in it after the
start-up of the network. Therefore, they suggested an approach based on
 Analysis through Unsteady Behaviour during Start-Up 309

unsteady flow equations for the steady state analysis of distribution net-
works. Usually, the methods of solving unsteady flow problems are more
complex and take more time as compared to steady state problems. How-
ever, as only the ultimate steady state behaviour is required and not the in-
termediate transient states, the approach can be simplified.
The basic unsteady flow equations for flow of water in a pipeline of neg-
ligible slope are [153,154]

dH aV dV fV V|
_L
T"
1/
V + = 0 (11.3)
dx ~~dx~ ~dl ID
and
dh r cl dv
dx + dt
-o (11.4)
g ox
in which V = velocity of flow; H = piezometric head; g — gravitational
acceleration; D = pipe diameter; and c = wave speed. Since we are not
interested in the transient behaviour at intermediate stages, conversion of
Equations (11.3) and (11.4) into finite difference form yields two character-
istic equations for the ultimate steady state flow in a pipe. These two equa-
tions are [153]
c(Vd - Vu) fLVu | V,
Hd = //„ — (11.5)
g 2gD

c(Vu - Vd) , /LK d | V.

= nd -\ (11.6)
8 2gD
in which the subscripts u and d denote the upstream and downstream ends
of a pipe, respectively. Equations (11.5) and (11.6) represent the standard
form and could have been derived from elementary considerations: "Head
difference over a pipe length represents the sum of inertia head and friction
head." The last term in Equations (11.5) and (11.6) represents the head loss
in a pipe as obtained from the Darcy-Weisbach equation. It can therefore
be replaced by the general head loss term, Equation (2.37), so that any
head loss formula can be used. Replacing Kby <\QlicD2, Equations (11.5)
and (11.6) can be written as
Ac
Hd = Hu — (Q* - Qu) - (11.7)

u = Hd Q* I"'1 01.8)

in which R is the resistance constant of the pipe.

310 OTHER ANALYSIS METHODS

In the characteristic equations, the wave speed c depends upon the fluid
properties, pipe material, pipe thickness, and pipe anchorage conditions.
The actual value of c would be necessary in an unsteady flow analysis.
However, for the ultimate steady state condition, any value of c would give
the same result, and therefore c can be taken as a round value of 1,000 m/s
for water.
When a reservoir is located at the upstream end or the downstream end
of a pipe, the appropriate characteristic must be used. Consider, for exam-
ple, the three-reservoir network shown in Figure 11.3 in which three reser-
voirs 1, 2, and 3, having water levels //„,, Ho2, and Ho3 are connected to a
junction J through pipes, 1, 2, and 3, respectively. For pipe 1, the forward
characteristic gives

Ac
Hj = Hol — (11.9)

For pipes 2 and 3, the backward characteristic gives

Ac
j = Ho2 + 2» - Q2d) + Q2d I""1 (11.10)

Ac
j = Ho3 + 3» - Q3d) 6 M I"'1 (11.11)

At junction J, the continuity equation must be satisfied, and therefore,

Qid = Q*» + Q3u (11.12)

Successive application of Equations (11.9) . . . (11.12) would yield the

desired steady state results. To begin the iterative procedure, initially Qlu

Figure 11.3 Three-reservoir network.

 Analysis through Unsteady Behaviour during Start-Up 311

in Equation (11.9), the forward characteristic originating at reservoir 1, and

Q2d and ft, in Equations (11.10) and (11.11), the backward characteristics
originating at reservoirs 2, and 3, respectively, are taken to be zero. Suc-
cessively, these Q values are taken equal to the values obtained in the pre-
vious iteration at the other end. The method is illustrated for a simple
three-reservoir network in Example 11.1.

Example 11.1

Obtain the steady state solution of the three-reservoir network of Exam-

ple 5.2 through the unsteady flow approach.

Solution

For the three-reservoir network of Example 5.2(Figure 11.3), Hoi = 120

m, Ho2 = 92 m, Ho3 = 80 m, Dol = 0.3 m, Do2 = 0.2 m, and Do3 =
0.3 m. The R values as obtained earlier are Rol = 222.91, Ro2 = 493.90,
and Ro3 = 148.61. Therefore, taking c = 1,000 m/s, Equations (11.9) . . .
(11.11) become, respectively,

Hj = 120 - 1442.1(2., - ft») - 222.91(2.* I Gi. I 0 8 " (D

Hj = 92 + 3244.7(ft a - Q2d) + 493.90ft, | Q2d | 0 8 5 2 (2)

Hj = 80 + 1442A(Q3u - Q3d) + 148.61ft, | ft, | 0 8 5 2 (3)

For the first iteration, taking ftu = Q2d = Q3d = 0, we get

120 - Hj HJ - 92 Hj - 80
1442.1 ^" ~ 3244.7 •ana ^ - ~ 1442.1

Substituting these values in the continuity equation for node J, i.e., Equa-
tion (11.12), and solving it we get Hj = 98.55 m. Therefore, Qld = 0.01488
m 3 /s, ftu = 0.00202 m3/s, and Q3u = 0.01286 m3/s. These values are sub-
stituted in Equations (1) . . . (3), the equations are resolved and new val-
ues of Hj, Q,d, ftu, and ftu are obtained. The iterative procedure is con-
tinued until the discharges in all pipes at their upstream and downstream
ends are nearly equal and the steady state condition is reached. After sev-
eral iterations, the answers given in Example 5.2 can be obtained.
312 OTHER ANALYSIS METHODS

11.3.2 GENERAL DISCUSSION

The method can be applied to looped networks and extended so that

pumps and valves can also be incorporated in the networks [153]. How-
ever, the method has not been popular because of the following reasons.
(1) The method converges extremely slowly. Even for the simple three-
reservoir network of Example 11.1, several iterations are required.
However, because the final result does not depend upon the actual
value of c (the inertia term becomes negligible when the iterations
are terminated and the steady state condition is practically reached),
convergence can be made faster by taking a smaller value of c. For
example, taking c = 100 m/s, iteration 1 would give the same value
of///, i.e., 98.55 m, but the Q values would be ten times larger, i.e.,
Qld = 0.1488 m3/s, Q2u = 0.0202 m7s, and Q3u = 0.1286 m7s. This
would, however, introduce oscillations in later iterations. It has been
observed that for large networks oscillations are almost always pres-
ent [155], and special measures are necessary to prevent them [153].
(2) As stated earlier, the iterations are terminated when the velocities
and discharges at both ends of all pipes are nearly equal, i.e., when
the inertia term in the characteristic equations becomes negligible.
In the actual steady state condition, the velocities and discharges at
both ends of all pipes are exactly equal, the value of the inertia term
is zero and therefore it is absent. Thus, the characteristic equations
reduce to the well-known steady state condition: "Head difference
over a pipe length represents the head loss due to friction (and other
causes) in it." Therefore, instead of using the inertia term and making
it gradually negligible, it is preferable not to consider it at all. This
is the approach taken in the solution of Example 5.2 and the solution
could be obtained quite easily. Therefore, as rightly pointed out by
Demuren and Ideriah [155], it is inefficient to employ the unsteady
flow approach for steady state problems because the unsteady ap-
proach takes considerably more computational effort and time and
has convergence problems.

11.4 ANALYSIS THROUGH OPTIMIZATION

11.4.1 INTRODUCTION

Analysis of hydraulic networks can also be achieved by treating it as an

optimization problem as shown by Arora [156], Hall [157], and Collins et
al. [75]. Arora considered a simple two-piped loop while Collins et al.
 Analysis through Optimization 313

have based their approach on rigorous theoretical background and devel-

oped nonlinear optimization models, solutions of which yield the hydrau-
lic network analysis. As optimization techniques are beyond the scope of
this text, only a simple elementary treatment is presented for network
analysis through optimization.

11.4.2 SYSTEM ENERGY APPROACH

Arora [156] suggested an approach based on the principle of conserva-

tion of energy. This principle states: "Flows in the pipes of a hydraulic net-
work adjust so that the expenditure of the system energy is minimum." The
expenditure of energy Ex when water flows in pipe x at the rate of Qx and
undergoing a head loss K is given by

Ex = KQxhx (11.13)

in which K = appropriate constant for conversion of energy to Joules or

kilowatt-hours. Therefore, the total expenditure of system energy, SE, in a
hydraulic network can be expressed as

SE
= Ti I KQ'h' I (U-14)

The modulus sign in Equation (11.14) guarantees the expenditure of energy

to be positive irrespective of the direction of flow in a pipe.
Expressing the head loss hx in terms of the known pipe resistance con-
stant Rox and discharge Qx [Equation (2.37)], Equation (11.14) can be ex-
pressed as

SE = £ KROX | Qx |"+1 (11.15)

The discharges in pipes, i.e., the Qx values, should be such that they
satisfy the node-flow continuity relationship at all demand nodes. There-
fore, the optimization model for a single-source network can now be ex-
pressed as

Minimize: SE = £ KROX | Qx |"+1 (11.16a)

Subject to: ^ Qx + qoj = 0, for all demand nodes j (11.16b)

connected
toy
314 OTHER ANALYSIS METHODS

Figure 11A Single-source, two-demand node looped network.

Because K is constant throughout, it can be dropped from Equation

(11.16a) and the system energy can be expressed in discharge-head loss
units (m 3 min" 1 m; Ls""'m, etc.). Therefore, the optimization model can
be expressed as

Minimize: SE = £ Rox \ Qx (11.17a)

Subject to: J^ Qx + qoj = 0, for all demand nodes j (11.17b)

connected
toj

Instead of giving a rigorous mathematical proof to show that the solution

of this optimization model yields the hydraulic network analysis solution,
the proof is demonstrated herein through a simple three-pipe, one-loop hy-
draulic network.
Consider a single-source (labelled 0), two-demand node (labelled 1 and
2) looped network of Figure 11.4. The known demands at nodes 1 and 2
are qol, and qo2, respectively. The pipes are labelled 1, 2, and 3 and their
known resistance constants are Rol, Ro2, and Ro3, respectively. Let the pipe
discharges, which are the decision variables, be Qu Q2, and Q3 in pipes
1,2, and 3, in the directions as shown in the figure, respectively. The op-
timization model for this network is given by

Minimize: SE = /?„, | Qt |"+1 + Ro2 | Q2 |"+1 + Ro3 | Q3 |"+1

(11.18a)

Subject to: Qx - Q3 - qol = 0 (11.18b)

and Q2 + Q3 - qo2 = 0 (11.18c)

 Analysis through Optimization 315

Rearranging the constraints of Equations (11.18b) and (11.18c) so that Q,

and Q2 are expressed in terms of Q3, we get

<2> = qol + Q3 (11.19a)

Q* = q*-Q* (11.19b)

Incorporating the values of Qt and Q2 in Equation (11.18a), the optimiza-

tion model now becomes

Minimize: SE = /?„, | qol + Q3 |"+1 + Ro2 \ qol - Q3 |"+1

+ Ro3\Q3\"+i (11.20)

which is now an unconstrained optimization model in a single decision

variable Q3. Therefore, for expenditure of system energy to be minimum,

= 0 = (n + l)[Rol | qol + Q

- R02 | qoi - Q3 |" + /?o3 | ft |"J (11.21)

[It can be shown that d2SE/dQl > 0, Equation (11.20) is strictly convex
and therefore, Equation (11.21) gives the value of Q3 for the minimum value
of SE.]
Reintroducing Qt and Q2, from Equations (11.19a) and (11.19b) in Equa-
tion (11.21) and simplifying, we get

Roi | Gi |" - Roi | Qi |" + Ro3 | ft |" = 0 (11.22)

which is precisely the loop-head loss relationship giving the algebraic sum
of the head losses in the three pipes forming the loop as zero. Thus, when
the expenditure of the system energy for a loop is minimum, the pipe dis-
charges are adjusted so that the loop-head loss relationship is satisfied. As
the node-flow continuity relationships are already satisified and as the
loop-head loss relationship is also now satisfied, the solution of the op-
timization model yields the hydraulic network analysis.
The proof can be extended to single-source, multiple-looped networks.
Selecting one pipe discharge in each basic loop as a basic decision varia-
ble, the optimization model can be recast to an unconstrained optimization
model with the number of basic decision variables equal to the number of
basic loops. By partially differentiating with respect to each basic decision
316 OTHER ANALYSIS METHODS

variable, and simplifying, we get the loop-head loss equations for the
loops. Thus, the solution of the optimization model having pipe discharges
as the decision variables yields the hydraulic network analysis.
The proof can also be extended to multiple-source networks. Herein also
one pipe discharge is selected as a basic decision variable from each real
loop, and one source-node flow as a basic decision variable from each
pseudo loop. The optimization model is then recast as an unconstrained
optimization model and is partially differentiated and simplified to get the
loop-head loss equations as a necessary condition for the optimal solution.

Example 11.2

Carry out the analysis of the single-source, three-demand node, two-

loop network shown in Figure 1: Example 11.2. The head loss in a pipe is
given by h = RQ18S2 (this example is the same as Example 8.2).

= O-8rn/s

0-9

Solution

In this example, Ro, = 5, Ro2 = 8, Ro3 = 8, Ro4 = 6, and RoS = 5.

The outflows at the demand nodes are qo2 = 0.8 m3/s, qo3 = 1.3 m3/s, and
qo4 = 0.9 m3/s.
Assuming the flow directions as shown in Figure 1: Example 11.2, and
expressing the expenditure of system energy in m V - m units, the op-
timization model becomes

Minimize: SE = 5 | <2i + 8 8 | Q3 |2-852

+ 6 | Q4 (1)

Subject to: Qx + Q3 - Q4 - 0.8 = 0 (2)

 Analysis thmugh Optimization 317

04 + 05 - 1.3 = 0 (3)

Q2 - Qs - Qs - 0.9 = 0 (4)

Taking Q3 and 0 5 as the basic decisions variables, we get

0 , = 2.1 - Q3 - Qs (5)

0 2 = 0.9 + 0 3 + Qs (6)

& = 1.3 - 05 (7)

Therefore, the unconstrained optimization model is

Minimize: SE = 5 | 2.1 - Q3 - Q5 | 2 8 5 2

+ 8 I 0.9 + 03 + 05

+ 8 I 03 + 6 I 1.3 - 05

+ 5 I 05 (8)

The values of the system-energy expenditure, SE are calculated for dif-

ferent values of 0 3 and 0 5 and system energy contours are plotted in Fig-
ure 2: Example 11.2. The system-energy expenditure is minimum when

0-1 0-2 0-3 0-4 0-5 0-6 07 0'8

318 OTHER ANALYSIS METHODS

Figure 11.5 Two-source, two-demand node looped network.

Q3 = -0.2216 m3/s, and Qs = 0.6471 mVs as in Example 8.2. The mini-

mum expenditure of system energy is 42.9537 m V - m . Therefore, the
steady state pipe discharges are Qx — 1.6745 m3/s, Q2 = 1.3255 nfVs,
Q3 = -0.2216 m3/s, Q4 = 0.6529 m 3 /s, and Q5 = 0.6471 m3/s.
11.4.3 CONTENT MODEL APPROACH

Collins et al. [75] have proposed a general optimization model, termed

by them as the content model, solution of which yields the network analy-
sis. They have provided a rigorous mathematical proof showing that the
solution provides network analysis. A simplified, but slightly different ver-
sion is presented herein and explained with the help of a simple multi-
source network.
Consider the two-source, two-demand node network shown in Figure
11.5. Nodes 1 and 2 are source nodes with known HGL values Hol, and
Ho2, respectively, and nodes 3 and 4 are demand nodes with known de-
mands qo3, and q^, respectively. Let the five pipes labelled 1 . . . 5 have
known resistances, / ? „ , . . . Ro5, respectively. Let the unknown inflows at
nodes 1 and 2 be qu and q2, respectively. Consider a pseudo ground node
G with a fixed known level Hoc as shown in Figure 11.6. Let source nodes

Figure 11.6 Network with ground node and pseudo pipes for the content model.
 Analysis through Optimization 319

1 and 2 be connected to the pseudo ground node G through pseudo pipes

1G and 2G. These pseudo pipes 1G and 2G are now assumed to carry in-
flows <7, and q2, respectively, as shown in the figure.
Now consider the following optimization model:

n+\ 1 «+l
Minimize: Content C = Ro G. R02 G2

1G3 n+l 1
R«\ G4

Gs "+1 + (n +
/ TJ TJ
1)^2 {noa — no 2 )

*• (n + - Hol) (ii .23a)

Subject to the node-flow continuity constraints

- G. - G2 = 0 (ii .23b)

q>~ - G 4 —Gs = 0 (ii .23c)

G. -- G3 + G4 -q* = 0 (ii .23d)

Q^ -f G3 + Gs - qo* = 0 (ii .23e)

Herein the first five terms of the objective function, Equation (11.23a)
represent the energy loss (in discharge-head loss units) in real pipes 1 . . .
5 of the network, respectively, and the last two terms represent (n + 1)
times the energy loss in the pseudo pipes. Note that the node-flow conti-
nuity constraints are written for all nodes of the network.
To prove that the solution of this optimization model gives the network
analysis, consider Qu Qs, and q2 as the basic decision variables. Using the
node-flow continuity constraints, let us express G2, Qi, Q*, and #, in terms
°f Qu Qs, and q2 so that

Qi = qo3 + <?o4 - qi ~ Qi (11.24a)

G s = G . + <?2 - Gs - <7o3 (11.24b)

G4 = ? 2 - G s (11.24c)

qt = qo3 + <?o4 - q2 (11.24d)

Substituting the values of Q2, Q3, Q4, and q, from Equations (11.24a)
320 OTHER ANALYSIS METHODS

. . . (11.24d), respectively, in the objective function of Equation (11.23a),

we get
Minimize: Content C = Rol \ Qx |"+1 + Ro2 \ qo3 + qa4

- qa3 | "+1 + RO4 | qi - Qs |"+1

+ Ros I Gs |"+1 + (n + \)q2(HoC - Ho2)

+ (n + \){qo3 + qo4 - ? 2 )(// o C - Hol) (11.25)

The objective function of Equation (11.25) now represents an uncon-

strained optimization model in three basic decision variables: Qy, Qs, and
q2. Therefore, for minimization of the optimization model, partially differ-
entiating Equation (11.25) with respect to Qu Qs, and q2, we get, respec-
tively,

-rpr = (n + !)[/?„! | Gi " — Roi \ qo3 + <?»4 - q2

- G. I" + Ro3 | G. + qi ~ Qs - qo3 I"] = 0 (11.26a)

dC
= (n + \)[-Ro3 - Qs -

q2- Qs\" Qs |"J = 0 (11.26b)

•T— = (it + l)[~Ro2 | <?o3 + <7o4 - ^2 - Qx | "

Gl Ro

+ (HoC - Ho2) - (HoG - Hol)] = 0 (11.26c)

Reintroducing G2, Qs, Q*, and q,, we get

G2 |" + Ro3 I G3 I" = 0 (11.27a)

-Ro3 \ Q3 Q4 |" + Ro5 "= 0 (11.27b)

- R 0 2 I G 2 |" + R o 3 \ Q3 \" + R°4 I Q* I" - Hol + / / „ , = 0

(11.27c)
 Analysis through Optimization 321

Adding Equations (11.27b) and (11.27c) and changing sign, we get

R02 | & |" - Ros I Qs I" + Ho2 - Hoi = 0 (11.27d)

Equations (11.27a), (11.27b) and (11.27d) are precisely the loop head loss
equations for the two real loops 1-3-4-1, 3-2-4-3, and the pseudo loop 1-4-
2-G-l, respectively. Since all the node-flow continuity and loop-head loss
equations are satisfied, the solution of the optimization model gives the
network analysis. For simplicity, H^ can be taken as zero, so that the gen-
eral content optimization model can be expressed as

Minimize: Content C = Ro, I 0,

L
X

- (n + 1) E<ijH
j
OJ (j e source nodes)

(11.28a)

Subject to: E Qij + qj = 0 , for all 7 (including source nodes)

j
(11.28b)
in which qy = qoj for demand nodes. Note that the second summation in
the objective function is only for nodes with known heads, i.e., source
nodes, while the constraints are written for all nodes including source
nodes.
Example 11.3

For the network of Figure 11.5, Rol =2, Rol = 1.2, Ro3 = 3, Ro4 =
1.6, Ro5 = 2.4, qo3 = 6 L/s, q^ = 4 L/s, //„, = 100.00 m, and Hol =
95.00 m. The head loss in a pipe is given by h = RQl 85 in which h is in
meters, and Q in liters per second. Analyze the network using the content
model approach.
Solution

Taking H^ = 0, n = 1.85, and substituting the various known values

for the network of Figure 11.6, the content model of Equations (11.28a) and
(11.28b) is expressed as

Minimize: Content C = 2 \ Qt | 2 8 5 + 1.2 | Q2 | 2 8 5 + 3 | Q3 I 2.85

+ 1.6 | 0 4 |2-85 + 2.4 | 0 5 |2-85

- 2.85 X 95<?2 - 2.85 X 100?, (1)

322 OTHER ANALYSIS METHODS

Subject to: q, - 0 , - Q2 = 0 (2)

<?2 - Q< - Qs = 0 (3)

0 . - Q3 + 04 - 6 = 0 (4)

and 0 2 + Q3 + 0 5 - 4 = 0 (5)

The solution of this nonlinear optimization model would give the mini-
mum value of the content C. Alternatively, taking 0 , , 0 5 , and q2 as the
basic decision variables, and expressing Q2, Q3, 0 4 , and qx in terms of Qu
Qs, and q2, we get the unconstrained optimization model given by

Minimize: Content C = 2 \ 0.
Q, 2.8S _|_
1.2 10 - 0. - 92

3 a + < ? 2 " G. - 6
2 ,

1.6 <?2 - 0s
2.8S
+ 2. 4 c ) 1 2.85

- 270.75^ 2 - 285(10 - q2) (6)

The solution of the unconstrained minimization model of Equation (6)

gives the minimum value of the content as —2692.303 Ls"'-m when
0i = 2.751 L/s, Qs = 1.598 L/s, and q2 = 3.984 L/s. Therefore, the
final solution is as follows:

Q1 = 2.751 L/s, <?, = 6.016 L/s,

02 = 3.265 L/s, <?, = 3.984 L/s,

03 = - 0 . 8 6 3 L/s, H3 = 87.00 m,

04 = 2.386 L/s, //4 = 89.29 m

05 = 1.598 L/s,

11.4.4 CO-CONTENT MODEL APPROACH

The content model approach described in the earlier section was based
on equations having pipe discharges and nodal flows as basic unknowns,
i.e., on Q-q equations. Colins et al. [75] have also proposed a model,
termed the co-content model, that is based on equations having the un-
known nodal heads as the basic unknowns, i.e., based on //equations. The
 Analysis through Optimization 323

unknown pipe flows are expressed in terms of the nodal heads and the
known pipe resistances, so that the energy loss in pipe x is given by

Ex = Q , - h x = 1V
(11.29)
ox

in which //, = Hoi and //, = Hoj for source nodes.

Again consider the network of Figure 11.5, with the known and unknown
parameters as shown therein. Let the unknown nodal heads at nodes 3 and
4 be H3 and H4, respectively. Herein also consider a ground node G with
fixed known level //oG, as shown in Figure 11.7. The nodes 3 and 4 are con-
nected to the ground node G with pseudo pipes, carrying the known nodal
outflows qo3 and qo4 as shown in the figure.
The co-content optimization model is expressed as

Minimize:
H.t - H3 Hol - H4
Co-Content (CC) -
Rir
H3 - H4 Ho2 - H3

U _

°2

qo4{H4 - HoG) (11.30)

The first five terms of the objective function [Equation (11.30)] represent

Figure 11.7 Network with ground node and pseudo pipes for the co-content model.
324 OTHER ANALYSIS METHODS

the energy loss in real pipes 1 . . . 5 of the network, respectively, and the
last two terms represent [(1/n) + 1J times the energy loss in the pseudo
pipes. Note that there are no constraints and therefore we get directly an
unconstrained optimization model in two basic decision variables H2 and
H3. Therefore, for minimization of the optimization model, partially dif-
ferentiating Equation (11.30) with respect to H2 and H3, we get

+ — H3 H3 - H4
t£-(i '"- oi

— H3
Ro + J =o (11.31a)

dCC
dK HHl- — HA H3 — H4

Ho2 — H4
= 0
(11.31b)

Reintroducing the Qx parameters, Equations (11.31a) and (11.31b) give,

respectively,

-0. + & - = 0 (11.32a)

and
= 0 (11.32b)

which are precisely the node-flow continuity equations for nodes 3 and 4,
respectively. Therefore, the solution of the co-content optimization model
gives the values of the unknown nodal heads H3 and H4 such that the node-
flow continuity relationships are satisfied. When the nodal heads are the
basic decision variables, the loop-head loss constraints are automatically
satisfied. Further, as the solution now also satisfies the node-flow conti-
nuity relationships, the solution of the co-content optimization model
gives the network analysis. For simplicity, Hoc can be taken as zero so that
the general co-content optimization model can be expressed as

Minimize:

Co-Content CC -
ft -
+ - +
(11.33)
 Problems 325

in which //, = Hoi and Ht = Hoj for source nodes. Note that the second
summation is only for nodes with known flows.

Example 11.4

Solve Example 11.3 by using the co-content model approach.

Solution

Taking // oC = 0, n = 1.85, and substituting the known values for the net-
work of Figure 11.7, the co-content model is expressed as

Minimize:

Co-Content CC =
100 - ft I ' 5 4 100 - fl, I
(1.2) 0 5 4
(2)0.54
95 - H3
LJ u I 1.54 (1.6)° 54

(3)054
1.54(67/3
54
95 - H4 | '
(24)0.54
The solution of the optimization H the minimum value of the
model gives
co-content as 1455.299 Ls"1- m for H3 = 87.00 m, and H4 = 89.29 m.
Other answers are as given in Example 11.3.
The number of the unknowns in the general content model (Example
11.3) is seven: five for the real pipes and two for the pseudo pipes. In addi-
tion, it has four constraints, one for each node. Thus, the number of un-
known decision parameters is maximum as in the formulation of Q equa-
tions in network analysis. When the content model is converted to an
unconstrained one, the number of decision variables reduces to three, one
for each loop, as in the formulation of AQ equations. The co-content
model is an unconstrained one and has only two decision variables for two
unknown nodal heads as in the formulation of H equations.

11.5 PROBLEMS

Problem 11.1: Obtain the steady state solution of the three-reservoir

network of Problem 5.2 using the unsteady flow approach.
Problem 11.2: Obtain the steady state solution of the two-loop network
of Figure: Example 8.2 using the unsteady flow approach.
Problem 11.3: Solve Example 8.1 using the system energy approach.
326 OTHER ANALYSIS METHODS

Problem 114: Solve Problem 5.2 using the system energy approach.
Problem 11.5: Solve Problem 5.2 using the content model approach
and find the minimum content of the network.
Problem 11.6: Using the solution of Example 8.2, find the minimum
content of the network, and, by slightly changing the solution, verify that
the content of the network is minimum for the obtained solution.
Problem 11.7: Solve Problem 5.2 using the co-content model approach
and find the minimum co-content of the model.
Problem 11.8: Using the solution of Example 8.3, find the minimum
co-content of the network, and, by slightly changing the solution, verify
that the co-content of the network is minimum for the obtained solution.
 Dynamic Analysis

12.1 INTRODUCTION

N the network analysis methods described so far, it was presumed that

I the nodal demands and the reservoir water levels remained constant.
This gave a network analysis for steady state, i.e., static conditions, and
thus it was a static analysis. The assumption of constant water demands
and reservoir water levels can be valid for a short period. However, in ac-
tual practice, neither the nodal demands nor the reservoir water levels re-
main constant over a long period. The nodal demands fluctuate during the
day depending upon the type of demand. A typical variation in nodal de-
mands for residential, commercial, office buildings, and hospitals as ob-
served by Rao, Markel and Bree [158] for Greenridge zone in the San Jose
Water Works Service area in California is shown in Figure 12.1. Naturally,
the water levels in the reservoirs would also change, and this change would
depend upon whether water is withdrawn from a reservoir or is supplied
to it from an external source or by the distribution network itself, if the
reservoir is floating on the system. Further, when a zone of a large network
is considered, the inflows from and outflows to the surrounding pressure
zones connected with the zone under study would also change.
To ensure an adequate level of service to the consumers under the vary-
ing demand conditions, proper operation and planning of the water supply
system is necessary. The service provided should ensure (1) adequate
maintenance of flow rates and pressures (residual heads) at all nodes at
various times and (2) proper management of the storage to balance the
supply and distribution of the water. From the operational point of view,
the knowledge of the impart of demand fluctuations on the level of service
would help in evolving proper control strategies to maintain the desired
level of service. From the planning point of view, it would be necessary to
evaluate the adequacy of the storage or the proposed network additions to
effectively provide for the expected increase in total demand. Both these
327
328 DYNAMIC ANALYSIS

Type of Demand:
Residential
Commercial
Office Buildings
Hospitals
00 —
0 2 6 8 10 12 2 L, 6 8 10 12
Midnight A.M. Noon RM. Midnight

Figure 12.1 Typical daily variation in nodal demands.

objectives can be achieved by carrying out the analysis of the network over
a period of 24-48 hours under changing demand patterns, reservoir water
levels, and boundary conditions. Such an analysis of the network is called
an extended period simulation, extended period analysis, or simply dy-
namic analysis.
The necessity of dynamic analysis for water distribution networks was
recognised by Shamir and Howard [88], Tart [159], and Demoyer, Gilman,
and Goodman [160]. However, Rao and Bree [161] and Rao, Markel, and
Bree [158] were perhaps the first to suggest a general, systematic readily
available procedure for carrying out a dynamic analysis of water distribu-
tion networks. Their approach is described in the next section and is fol-
lowed, in a subsequent section, by an improvement suggested by Bhave
[162].

12.2 ITERATIVE METHOD [161,158]

12.2.1 INTRODUCTION

The dynamic analysis of a water distribution system basically depends

upon the fundamental water-balance principle: "For a particular time in-
terval, water supplied to a system less water taken out from it equals the
 Iterative Method 329

change in water stored in the system." The application of this principle to

a single-reservoir water distribution network without any boundary condi-
tions is quite simple. The total volume of water supplied to the consumers
during a particular time interval can be obtained from the demand curve
for the entire network. Water supplied, if any, to the reservoir during this
time interval can also be evaluated. Thus, the change in volume of water
in the reservoir and the corresponding change in its water level can be de-
termined. The initial water level in the reservoir plus the change in its
water level would give the water level at the end of the time interval.
The problem, however, is complicated for a multireservoir water distri-
bution system. The total volume of water supplied to the consumers and,
therefore, the total volume of outflow from the entire system during a par-
ticular time interval can be determined. This will equal the total volume of
water supplied by the reservoirs to the network. However, the actual vol-
ume of water supplied by each reservoir is not known and, therefore, nei-
ther its water level nor its actual supply rate at the end of the time interval
is known. As neither the supply rates nor the water levels in the reservoirs
at the end of the time interval are known, the static analysis of the network
at the end of the time interval cannot be directly obtained by the different
formulations of equations and their solutions with the analysis methods
studied so far. Therefore, Rao and Bree [161], and Rao, Markel, and Bree
[158] suggested an iterative procedure for obtaining the static analysis at
the end of the time interval to obtain the dynamic analysis for the time in-
terval.

12.2.2 PROCEDURE

The period of dynamic analysis is subdivided into several time intervals,

and the successive static solutions at the end of these time intervals are
linked with each other through an iterative integration procedure.
A static solution is obtained for a particular instant, say time t. This
solution gives the reservoir flow rates at time t. Assuming that these reser-
voir flow rates remain constant for the time interval, say At, the volume
change in each reservoir during this time interval At is predicted. Using
the known capacity-elevation curves for the reservoirs, the changes in
water levels of the reservoirs are predicted. Using the known water levels
at time t and the predicted changes in water levels during the time interval
At, the water levels in the reservoirs at time t + At are predicted. These
predicted water levels in the reservoirs are then used to carry out the static
analysis of the network for time t + At. Since the flow rates at the reser-
voirs as obtained from the static analysis for time t + At will be different
from those used in the predictor calculations, a correction would be neces-
sary. For the second iteration, the average of the flow rates at times t and
330 DYNAMIC ANALYSIS

t + At are used to calculate the volume change in the reservoirs and the
predictor iteration is repeated. The predictor-corrector iterative procedure
is continued until the predicted and corrected reservoir water levels agree
to the desired degree of accuracy. Thus, the static analysis of the network
from time t is linked to that for time t + At through an iterative, pre-
dictor-corrector integration step, thereby giving the dynamic analysis of
the network for time interval (t, t + At). The schedule of the pump and
valve settings and the effect of the boundary conditions, if any, are also
considered during this step. The reservoir flow rates and water levels for
time t + At obtained at the end of the iterative integration step become the
input for the next time interval.
The procedure for carrying out the dynamic analysis of a network by the
indirect method consists of the following steps:
Step 1. At time t, the following data are available:
a. Reservoir HGL values Hr(t) and reservoir water volumes Vr{t) in
which r 6 M, the set of reservoirs
b. Nodal demands <£,(?) and qj(t + At) in which y E N, the set of demand
nodes
c. If the network contains boundary elements such as pumps that bring
water into the network from external sources and booster pumps that
bring water into the network from neighboring pressure zones, then the
flows qb(t) or the HGL values Hb(t) and the relationship between qb and
Hb at all the boundary elements, b, b G B, the set of boundary ele-
ments. The flows and the HGL values at the boundary elements may be
a function of the network demand and of the water purchase contracts.
Step 2. Using the data in step 1, the static analysis of the network is ob-
tained for time t. This solution provides the HGL values Hj(t),j <E N for
the demand nodes and the outflow rates qr(i), r € Mfor the reservoirs.
Step 3. Assuming the outflow rate qr(t) for reservoir r to be constant
during the time interval (t, t + At), the volume decrease in reservoir water
is obtained as
AVr(t, t + At) = qr(t)At, r = 1 . . . M (12.1)

Step 4. The net total outflow from all the reservoirs is computed as
M

AVr = £ qr(t)At (12.2)

r=l

Similarly, the net outflow from all the boundary elements is computed as
B
AV
» = £ ?»«A* (12-3)
 Iterative Method 331

As the total outflow from the reservoirs and boundary elements is the in-
flow to the distribution network, the total inflow to the network is LAVr +
LAVb.
Step 5. By plotting a curve of the demand rate versus time for the entire
region under consideration, the cumulative total demand D(t, t + At) is
obtained from the area under the demand curve for time interval (t,
t + At).
Step 6. The error in water volume balance for the network is obtained as

Ep(t, t + At) = L qr(t)At + 2^ q»(t)&t + D(t, t + At)

rEM h£B
(12.4)
in which Ep = predicted error for the entire network. Assuming that this
error is shared by the reservoirs only and not by the boundary elements
[161], the error is distributed to all reservoirs in proportion to the with-
drawal rate at each reservoir. Thus, for reservoir r,

erp(t, t + At) = J?r(t) X Ep(t, t + At) (12.5)

rEM

in which erp = predicted error for reservoir r.

Step 7. The reservoir volume at time t + At is then predicted as

Vrp(t + At) = Vr(t) + qr{t)At + erp(t, t + At) (12.6)

Step 8. Using the known capacity-elevation relationship for reservoir r,

i.e., Vr = fr(Hr), the predicted volume Vrp(t + At) provides the predicted
HGL, Hrp(t + At) for reservoir r at time t + At.
Step 9. Using the Hrp(t + At), r E M and q^l + At),j E N and the
relevant data for the boundary elements, the static analysis of the network
is performed for time / + Af to obtain qr(t + At) values for all r, r E R.
Step 10. The system is checked for preset switch points for controlling
valves and pumps. If one of these is switched in the time interval (t, t +
At), then the next step in the integration procedure is step 13; otherwise,
the error in water volume balance is recomputed using the new flow rates.
Thus,

Ec(t, t + At) = Y, M O + qr(t + At)] Y

rEM

At
[qb(t) + qb(t + At)] Y + D(t, t + At)
bEB

(12.7)
332 DYNAMIC ANALYSIS

in which Ec = corrected error for the entire network. This error is reallo-
cated to reservoir r, r G M, in proportion to the average flow rate. Thus,

+ + At)
eM, t + At) = - ^ *<' X Ec(t, t + AO
L '(f + )
(12.8)

Step 11. The volume of water in reservoir r, r E M, is corrected using

the corrector equation

Vrc(t + At) = Vr(t) + [qr{f) + qr(t + At)] y + ejf, t + At)

(12.9)

Using this corrected reservoir volume, the HGL for reservoir r for time
t + At, i.e., Hr(t + At), is computed for all r, r E M, from the respec-
tive capacity-elevation relationship Vr = fr(Hr), r G M.
Step 12. The difference between the predicted and corrected HGL values
is estimated for all r, r G M. If the difference is more than the permissible
limit, the predictor-corrector integration step is repeated for the same time
interval. If the difference is within the permissible limit for all reservoirs,
the dynamic analysis for time interval (t, t + At) is complete. Finally, the
obtained corrected values Hrc(t + At) and Vrc(t + At) are set to Hr(t +
At) and Vr(t + At) respectively, for all reservoirs. These values then serve
as the starting values for the next time interval.
Step 13. If there is a switch of a pump or valve in the time interval (t,
t + At), then the time t + A't (t < t + A't < t + At) at which the
switching occurs is determined. The time interval is reduced from At to
A't and the integration procedure is carried out for two time intervals
(f, t + A't) and (t + A't, t + At).
The entire procedure is cycled through the various time intervals to
cover the entire period of analysis. This completes the dynamic analysis of
the network.

Example 12.1

The network shown in Figure: Example 12.1 has nodes 1 and 2 as reser-
voirs and nodes 3 . . . 6 as demand nodes. The resistance constants in the
Hazen-Williams head loss formula, hf = RQ1852 for pipes 1 . . . 7, are
shown underlined in the figure. The elevation-capacity data for the reser-
voirs are given in Table 1: Example 12.1, while the nodal demand rates at
an interval of 4 hr are given in Table 2: Example 12.1. From the demand
 Iterative Method 333

Node,1
I rj

curve for the entire network, total network demand volumes are obtained
and given in Table 3: Example 12.1. Reservoir 1 is filled from an external
source during time interval 4-12 hr at a constant rate of 14 m 3 /min, while
reservoir 2 is filled during time interval 12-20 hr at a constant rate of 10
m 3 /min. At t = 0 hr, the water levels in reservoirs 1 and 2 are 101.200 m,
and 102.000 m, respectively. Both reservoirs are floating on the network.
Carry out the dynamic analysis of the network from 0 hr to 4 hr taking
it as one time step. (Complete dynamic analysis for the total period of
0-24 hr is carried out in Example 12.4.)

Solution

Step 1. At t = 0 hr, //,(0) = 101.200 m, K,(0) = 2,000 m 3 , H2(0) =

102.000 m, V2(0) = 5,000 m 3 , ^3(0) = 1.000 m3/min, <?4(0) = 0.800
mVmin, ^5(0) = 1.000 nWmin, and <?6(0) = 0.600 mVmin. Neither of the
two reservoirs is filled from an external source during the time step 0-4 hr.
Step 2. The static analysis of the network is carried out for t = 0 hr.
From this analysis, for the flow directions in pipes as shown in Figure: EX-

TABLE 1: EXAMPLE 12.1.

Hydraulic Gradient Level

in Meters for Reservoir at
Reservoir Capacity
in Cubic Meters Nodei Node 2

0 100.00 98.00
1,000 100.60 98.80
2,000 101.20 99.60
3,000 101.80 100.40
4,000 102.40 101.20
5,000 103.00 102.00
6,000 103.60 102.80
334 DYNAMIC ANALYSIS

TABLE 2: EXAMPLE 12.1.

Nodal Demand Rate in Cubic Meters per Minute at Time

Demand
Node 0 hr 4 hr 8hr 12 hr 16 hr 20 hr 24hr a

3 1.000 2.000 3.000 3.000 2.000 2.000 1.000

4 0.800 1.600 2.000 2.600 1.200 1.600 0.800
5 1.000 3.000 4.000 2.000 4.000 2.000 1.000
6 0.600 1.600 1.000 1.200 2.000 1.400 0.600

Total Network 3.400 8.200 10.000 8.800 9.200 7.000 3.400

Demand Rate
a
Nodal demand rates are same as those at ( = 0 hr.

ample 12.1, Q,(0) = -0.3853 m3/min, Q2(0) = -0.1996 nvVmin,

03(O) = -1.3853 mVmin, Q4(0) = 1.2799 mVmin, Q5(0) = 0.2804
nfVmin, Q6(0) = 1.3196 nvVmin, Q7(0) = 0.7196 nWmin, H3(0) =
101.268 m, // 4 (0) = 101.210 m,// 5 (0) = 101.172 m, and// 6 (0) = 101.499
m. For the assumed pipe flow directions, as reservoir 1 is being depleted,
q, = - ( ( ? , + Q2) = - ( - 0 . 3 8 5 3 - 0.1996) = 0.5849 m3/min. Simi-
larly, for reservoir 2, q2 = Q3 — Q4 — Qb = —3.9849 m3/min. As qt
is positive, reservoir 1 is being filled by the network at t = 0 hr, at a rate
of 0.5849 m3/rnin. Reservoir 2 is being emptied at t = 0 hr at a rate of
3.9849 mVmin., to meet the total network demand of 3.400 nrVmin and
the supply of 0.5849 m3/min to reservoir 1.
Step 3. Assuming the flow rates of 0.5849 m3/min and —3.9849 nvVmin
for reservoirs 1 and 2, respectively, to remain constant for time interval
0-4 hr, the reservoir volume changes are AKt(0, 4) = 0.5849 X
240 = 140.38 m3 (reservoir 1 is filled by 140.38 m3); and AK2(0, 4) =
-3.9849 X 240 = -956.38 m3 (reservoir 2 is depleted by 956.38 m 3 ).
Step 4. The total flow from the two reservoirs during time interval 0-4

TABLE 3: EXAMPLE 12.1.

Total Network
Time Interval Demand Volume
in Hours in Cubic Meters

0-4 1,488
4-8 2,256
8-12 2,304
12-16 2.184
16-20 1.992
20-24 1,296
 Iterative Method 335

hr is EAF(O, 4) = 140.38 - 956.38 = -816.00 m3; i.e., total outflow of

816.00 m3. There is no flow from any boundary element, and therefore the
total supply to the network by the two reservoirs is 816.00 m3.
Step 5. The total network demand volume for time interval (0-4 hr)
shown by Table 3: Example 12.1, is 1,488 m3.
Step 6. The error in water volume balance (outflow) for the entire net-
work, £ p (0, 4) = -816 + 1,488 = 672 m3.

1 8
.-. elp(0, 4) = ^ X 672 = 115.60 m3

and

elp(0, 4) = g 5 ^" 38 X 672 = - 7 8 7 . 6 0 m3

Step 7. The reservoir volumes at t = 4 hr are predicted as

Vlp(4) = K,(0) + AK,(0, 4) + eip(0, 4)

= 2,000 + 140.38 + 115.60 = 2255.98 m3

and

V2p{4) = V2(Q) + AK2(0, 4) + elp(0, 4)

= 5,000 - 956.38 - 787.60 = 3256.02 m 3

Step 8. Using the elevation-capacity data for reservoirs 1 and 2, given in

Table 1: Example 12.1, the reservoir water levels are predicted as

i« no
Hip(4) = 101.20 + y ~ X (101.80 - 101.20) = 101.354 m

and

2
#2P(4) = 100.40 + . 5 ^ 2 X (101.20 - 100.40) = 100.605 m
1 ,UUU

Step 9. Using the values Hlp(4) = 101.354 m, H2p(4) = 100.605 m,

= 2.000 nrVmin, g4(4) = 1.600 nWmin, qs(4) = 3.000 mVmin, and
= 1.600 nrVmin, the static network analysis is carried out. This
336 DYNAMIC ANALYSIS

gives reservoir flows as q,(4) — —4.2086 rrrVmin, and g2(4) = —3.9914

rrvVmin. (Both reservoirs are being depleted at t = 4 hr.)
Step 10. Because there are no pumps and valves in the system, checking
for their switching does not arise. Using the new values for the reservoir
flows, the error in volume balance (outflow) is recomputed for each reser-
voir as
Ec(0, 4) = [(0.5849 - 4.2086) + (-3.9849 - 3.9914)]

240
X - y - + 1,488 = 96 m3

[The volume of 96 m3 represents the difference between the actual total

consumption of the network as obtained from the demand curve and that
obtained from the average network demand basis. Thus, the volume differ-
ence = 1488 - (3.4 + 8.2) X 240 -^ 2 = 96 m3, and remains constant
throughout the iterative procedure for this time step.]
. 0.5849 - 4.2086
€u{ X
•• ' ' ~ (0.5849 - 3.9849) + (-4.2086 - 3.9914)

= 29.99 m3 (outflow)
and
-3.9849 - 3.9914
X
e2c(0, 4) - (Q 5 g 4 9 _ 3 9 g 4 9 ) + ( _ 4 2086 - 3.9914)

= 66.01 m3 (outflow)
Step 11. The reservoir volumes are corrected using the corrector
equation:
M4) = ^.(0) + fa,(0) + 9l(4)] X y + elc(0, 4)

= 2,000 + (0.5849 - 4.2086) X 120 - 29.99

= 1,535.17 m3
and
K2c(4) = V2(0) + fo2(0) + q2(4)] X y + e2c(0, 4)

= 5,000 + (-3.9849 - 3.9914) X 120 - 66.01

= 3,976.83 m3
 Direct Method 337

From the elevation-capacity data for the reservoirs, we get H,c(4) =

100.921 m and H2c(4) = 101.181 m.
Step 12.

Hlp{4) - Hlc(4) = 101.354 - 100.921 = 0.433 m

and

H2p(4) - Hu(4) = 100.605 - 101.181 = - 0 . 5 7 6 m

As the absolute values of the difference between the predicted and cor-
rected water levels for the reservoirs are more than 0.001 m, an assumed
permissible value, the corrected water levels are taken as the predicted
values and the next predictor-corrector integration iteration (steps 9 . . .
12) is carried out. The iterative procedure (steps 9 . . . 12) is continued
and the iteration results are shown in Table 4: Example 12.1. The solution
of the dynamic analysis for time interval 0-4 hr is <2i(4) = 0.9525
nrVmin, Q2(4) = 2.1981 nrVmin, Q3{4) = -1.0474 nrVmin, QA(4) =
1.3983 nrVmin, Qs(4) = 1.9962 m3/min, Q6(4) = 2.6036 m3/ min,
G,(4) = 1.0037 nWmin, q,{4) = -3.1506 nrVmin, q2(4) = -5.0494
nrVmin, K,(4) = 1671.7 m3, V2{4) = 3841.2 m \ //,(4) = 101.003 m,
#2(4) = 101.073 m, H3(4) = 100.637 m, H4(4) = 100.143 m, // 5 (4) =
98.704 m, and H6(4) = 99.308 m.

12.3 DIRECT METHOD [162]

In the predictor-corrector iterative procedure described in the earlier

section, qr(t + At) values are taken as qr{i) for the first predictor-correc-

TABLE4: EXAMPLE 12.1.

Predicted Flow in Cubic

Meters pei Minute at Corrected Head in Meters
Time t = 4 Hr, for at Time t = 4 Hr for

Iteration Reservoir 1 Reservoir 2 Reservoir 1 Reservoir 2

Initial (steps 3 . . .8) 0.58493 -3.9849 100.921 101.181

1 (steps 9 . . .12) -4.2081 -3.9918 101.021 101.048
2 (steps 9 . . .12) -2.9072 -5.2928 100.998 101.079
3 (steps 9 . .12) -3.2065 -4.9935 101.004 101.072
4 (steps 9 .. .12) -3.1376 -5.0624 101.002 101.073
5 (steps 9 .. . 12) -3.1534 -5.0465 101.003 101.073
a
Reservoir 1 is being filled by the distribution network.
338 DYNAMIC ANALYSIS

tor integration iteration (steps 3 . . .8). These values are then successively
corrected (steps 9 . . . 12) so that the predicted reservoir HGL values Hrp
(t + At) and the predicted reservoir flows qn,(t + At) reasonably agree
with the corresponding corrected values Hrc(t + At) and qrc(t + At), re-
spectively, for all reservoirs. However, as has been recently shown by
Bhave [162], it is possible to obtain a relationship between Hr(t + At) and
qr(t + At) and use it in the analysis so that the static analysis for time
t + At can be obtained directly from the static analysis for time t thereby
avoiding the predictor-corrector iterative procedure.

12.3.1 THEORY

From the known capacity-elevation relationship, Vr = fr(Hr) for reser-

voir r, we get

AVr = f'r{Hr)AHr, r = 1. . .M (12.10)

in which AVr = change in volume of reservoir r corresponding to the

change in HGL, AHr for reservoir r, and//(//,.) = first derivative, repre-
senting the cross-sectional area of reservoir r at HGL, Hr. Therefore, the
outflow from reservoir r in time interval (r, t + At) can be expressed as

AVr(t, t + At) = f:\H,(t)][Hr(t + At) - Hr(t)] (12.11)

Equation (12.11) on rearrangement of terms can also be expressed as

AK ( A
Hr(t + At) = Hr(t) + f !:f, * ° (12.12)
JrlJirU)\

Now AVr(t, t + At) is the volume outflow from reservoir r in time

interval (f, t -\- At) and is equal to the average outflow rate \qr(t) +
qr(t + At)] -s- 2, multiplied by the time interval A?, plus the share of the
error in the water volume balance [see Equation (12.9)].
From the static analysis for time t, the total rate of supply to the network
from all reservoirs and boundary elements at time t is Lqr(t) + Lqb(t).
This total rate of supply at time t must be equal to the total rate of demand
for the entire network at time t, i.e., Eqj(t). Similarly, this must be true
for time t + At and therefore, Lqr(t + Af) + Lqb(t + At) must be equal
to Lqy(t + At). Therefore, during the time interval (t, t + At), the average
rate of supply from all reservoirs and boundary elements must equal the
 Direct Method 339

average rate of demand for the entire network. If we multiply these values
by time interval At, we get the total volumes, and therefore,

= £ «/(') + E «/(' + A?) y (12.13)

Equation (12.13) states that the total volume of water supplied by the res-
ervoirs and boundary elements during the time interval (t, t + At) as ob-
tained from the average rate of supply is equal to the total volume of out-
flow from the demand nodes of the network during the same time interval
(t, t + At) as obtained from the average rate of demand.
Now, from the network demand curve, the area under the curve for time
interval (t, t + At) gives the total volume of water consumed and therefore
the total volume of water actually supplied by the demand nodes. This ac-
tual volume is D(t, t + At) instead of

as estimated from the average rate basis. If C(t, t + At) is taken as a cor-
rection factor to convert the estimated total supply to the actual total sup-
ply by the demand nodes, we have

„,
(12.14)
At
2rf qAt) + (t + A?) T
j j

Distributing the difference between the actual and estimated total net-
work demand to all reservoirs and boundary elements in proportion to
their withdrawal rates, the supply of volume of water by reservoir r in time
interval (f, t + At) can be expressed as

AVr(t, t + At) = C(t, t + At)\qr{t) + qr(t + At)] y

(12.15)
340 DYNAMIC ANALYSIS

Substituting the value of AVr(t, t 4- At) from Equation (12.15) in Equa-

tion (12.12), we get

C(t, t + At)[qr{t) + qr{t + At)] y

H,(t, t + At) = HAt) + jr{m]
(12.16)

On rearrangement of terms, Equation (12.16) can also be expressed as

At) = * " '

C(t, X + AO y
(12.17)

As stated earlier, the term/ r '[// r (/)] represents the cross-sectional area
of reservoir r at HGL Hr(t). For reservoirs having constant cross-sectional
area, the term ft[Hr(t)] remains constant throughout the entire dynamic
analysis. For reservoirs having variable cross-sectional area, the value of
fr[Hr(t)] is computed for each starting time t of each time step, and it is
assumed that the value remains constant during the time step, i.e., the time
interval it, t + At). The correction factor C(t, t + At) is the same for all
reservoirs and boundary elements for a particular time interval (f, t + At),
but will be different for different time steps and therefore is computed for
each time step.
All terms except qr(t + At) and Hr(t + At) in Equations (12.16) and
(12.17) are either known or evaluated. Thus, these equations give relation-
ships between reservoir HGL values and reservoir outflow rates at time
t + At.
If a reservoir is filled at a constant rate of qfr during time interval (t,
t + At), Equations (12.16) and (12.17) would modify to, respectively,

C(t, t + At)[qr(t) + qr(t + At)] y + q/rAt

Hrit + At) = H,(t) + j r ^ ^
(12.18)

and
it + At) - Hr{t)] X MHAt)]
A
'> = ^ Ar
C(f, f + AO y
(12.19)
 Direct Method 341

For a boundary element such as a pump, a direct relationship between

the flow rate and the head as given by Equations (5.4) . . . (5.12) can be
obtained. Thus,

Hb{i) = Fb\qb(t)] (12.20)

qb(t) = fb[Hb(t)] (12.21)

Therefore, for time t + At,

Hb(t + At) = Fb[qb(t + At)] (12.22)

qb(t + At) = fb[Hb(t + At)] (12.23)

The volume supply during time interval (t + At) is given by

AVb(t, t + At) = C(t, t + At)[qb(t) + qb(t + At)] y (12.24)

If the supply rate from a boundary element remains constant during time
interval (t, t + At), i.e., if it is qb(f) throughout the time interval (t, t +
At), a correction must not be applied to it. Therefore, while evaluating the
correction factor given by Equation (12.14), this known and fixed volume
qb(t)At is subtracted from the numerator and denominator of the right-
hand side of Equation (12.14).
If the supply to the network is from a clearwell having constant water
level, this known and fixed water level is used in the analysis for time
t + At. [Herein, the question of establishing and using a relationship be-
tween Hr(t + At) and qr{t + At) does not arise.] The value of qr(t + At)
for the clearwell is obtained from the analysis. The supply of water from
the clearwell during the time interval (t, t + At) is obtained using Equa-
tion (12.15).
Equations (12.16). . . (12.24) are used in formulating the required equa-
tions.

12.3.2 PROCEDURE

The procedure for carrying out the dynamic analysis of a network by the
direct method consists of the following steps.
Step 1. Same as step 1 of the indirect method.
Step 2. Same as step 2 of the indirect method.
Step 3. Using the nodal demand rates at times t, and t + At, determine
342 DYNAMIC ANALYSIS

the total outflow from the network during the time interval (f, t + At), as
obtained from the average rate of demand.
Step 4. Same as step 5 of the indirect method.
Step 5. Compute the value of the correction factor C(t, t + At) using
Equation (12.14).
Step 6. Formulate the appropriate equations from Equations (12.16) . . .
(12.24).
Step 7. Using the equations obtained in step 6 and other equations, as
described in Chapter 7, formulate the equations required for carrying out
the static analysis for time t + At.
Step 8. Carry out the static analysis for time t + At.
Step 9. The system is checked for preset switch points for controlling
valves and pumps. If one of these is switched in the time interval (t, t +
At), then go to step 10; otherwise, the dynamic analysis for time step (t +
At) is complete and go to step 3 to continue the dynamic analysis for the
next time step.
Step 10. If there is a switch of a pump or a valve in the time interval (t,
t + At) then determine the time t + A't (t < t + A't < t + Af) at which
the switching occurs. Reduce the time interval from At to A't and carry
out the dynamic analysis (steps 3 . . . 9) for time interval (t, t + A't) and
then for time interval (t + A't, t + At).
The direct method for carrying out dynamic analysis is illustrated in Ex-
ample 12.2.

Example 12.2

Carry out the dynamic analysis of the network of Example 12.1 by the
direct method for time interval (0, 4 hr), taking it as one time step.

Solution

Step 1. Same as step 1 of Example 12.1.

Step 2. Same as step 2 of Example 12.1.
Step 3. The total outflow volume from the network during time interval
0-4 hr as obtained from the average rate of demand = (3.4 + 8.2) X
240/2 = 1,392 m3.
Step 4. The total network demand volume as obtained from the demand
curve for time interval 0-4 hr is 1,488 m3.
Step 5. Correction factor C(0, 4) = 1,488/1,392 = 1.06897.
Step 6. The formulation of Equations (12.16) . . . (12.24) is as follows:
For reservoir 1, /,'[//,(())] = reservoir area at HGL 101.200 m =
1,000/0.6 = 1,666.67 m2. Similarly, for reservoir 2,/ 2 '[// 2 (0)] = reservoir
area at HGL 102.000 m = 1,000/0.8 = 1,250 m2. Therefore, from Equa-
 Direct Method 343

tions (12.16) and (12.17), we get, for reservoir 1, //,(4) = //,(0) +

1.06897[0.5849 + <?,(4)] X (240/2) H- 1,666.67, or

//,(4) = 101.200 + 0.076966^,(4) + 0.5849] (1)

and

4,(4) = 12.9929[//,(4) - 101.200] - 0.5849 (2)

Similarly, for reservoir 2,

// 2 (4) = 102.000 + 0.10262[g2(4) - 3.9849] (3)

and

q2(4) = 9.74462[// 2 (4) - 102.000] + 3.9849 (4)

As seen from Figure 1: Example 12.1, for the assumed directions of flow
in pipes, the rate of flow (inflow positive, outflow negative) for the reser-
voirs is given by

9.(4) = - I G , ( 4 ) + Gi(4)] (5)

and

<?2(4) = G3(4) - G4(4) - Q.(4) (6)

Steps 7 and 8. Equations (1) . . . (6) are used in formulating the equa-
tions necessary for carrying out static analysis at t = 4 hr. Replacing
1.852, the index in the pipe-head loss relationship by n, for simplicity, we
can derive the necessary equations as follows:
HEquations. The //equations are formulated by writing node-flow con-
tinuity equations at all nodes of the network. Therefore, using Equation
(2), the H equation for node 1 is given by

0.4 | [ 0.2

- (12.9929[//,(4) - 101.200] - 0.5849)

= 0 (7a)
344 DYNAMIC ANALYSIS

Similarly, using Equation (4), the H equation for node 2 is

\H3(4) - H2(2)V'" \H2(4) - # 4 <4)1"" \H2{4) - 1

0.4 | [ 0.5 J [ 0.3 J

- j9.74462[// 2 (4) - 102.000] + 3.9849}

= 0 (7b)

[Note that q, and q2 represent flows for reservoirs. Because they are out-
flows, as seen from Equations (5) and (6), respectively, they are preceded
here by a negative sign.]
For nodes 3 . . . 6, respectively,

H2{4) - #4(4)1"" f// 4 (4) -

0.2 I L 0.5 J [ 0.4 J
- 1.600 = 0 (7d)

» (7e,

-1.600-0 (7f,

Equations (7a) . . . (7f) provide six independent H equations for six

basic unknowns #i(4) . . . H6(4). These equations can be solved simulta-
neously by the Newton-Raphson method or by the linear theory method to
obtain the static analysis for t = 4 hr and thereby the dynamic analysis for
time interval 0-4 hr. Note that for ordinary static analysis for t = 4 hr, the
values of #i(4) and H2(4) would be known, and therefore Equations (7a)
and (7b) would be absent. Equations (7c) . . . (7f), with known values of
#i(4) and # 2 (4) therein, would provide four equations for four unknowns
# 3 (4) . . . // 6 (4). Thus, in dynamic analysis the number of required H
equations increases by M, the number of reservoirs.
Q Equations. From node-flow continuity relationships at nodes 3 . . . 6,
we get, respectively,

Gi(4) - fia(4) - 2.000 = 0 (8a)

 Direct Method 345

62(4) + fi4(4) - Gs(4) - 1.600 = 0 (8b)

Cs(4) + fi,(4) - 3.000 = 0 (8c)

06(4) - 07(4) - 1.600 = 0 (8d)

From the loop-head loss equations for the basic loops I and II, we get,
respectively,

O.2[02(4)]" - 0.5[Q4(4)]" - 0.4[G3(4)]- - 0.4[g,(4)]- = 0

(8e)

0.4[Q5(4)]- - 0.6[fi7(4)]" - 0.3[&(4)]" + O.5[04(4)]" = 0

(8f)

Similarly, from the loop-head loss relationship for the pseudo loop, i.e.,
loop III,

0.4[G,(4)]" + 0.4[G3(4)]" + H2{4) - H,(4) = 0 (8g)

Now from Equation (1), ff,(4) = 101.200 + 0.076966[tf,(4) + 0.5849].

Substituting the value of #,(4) from Equation (5) and simplifying we get

#,(4) = 101.245 - 0.076966[Q,(4) + Q2(4)] (8h)

Similarly,

H2(4) = 101.591 + 0.10262[ft(4) - Q4(4) - Q6(4)]

Substituting the values of //,(4) and H2(4) from Equations (8h) and (8i),
respectively, in Equation (8g) and simplifying, we get

0.4[Gi(4)]" + 0.4[Gi(4)]" + 0.10262[G3(4) - Q4(4) - Q6{4)\

+ 0.076966[G.(4) + Gi(4)] + 0.346 = 0

Simultaneous solution of Equations (8a) . . . (8f) and (8j) by the linear

theory method provides the dynamic analysis solution for time interval 0-
4 hr. Note that in ordinary static analysis for t = 4 hr, the value of //i(4)
and H2(4) would be known, and therefore, the simultaneous solution of
Equations (8a) . . . (8g) would yield the ordinary static analysis solution.
In dynamic analysis, the values of Ht(4) and H2(4) are expressed in terms
°f G*(4), and therefore, Equation (8g) is replaced by Equation (8j). Thus,
346 DYNAMIC ANALYSIS

only the Q equation for the pseudo loop has changed. The total number of
Q equations for the dynamic analysis has remained the same, i.e., seven,
as in the ordinary static analysis.
AQ Equations. The values of Qi(4) . . . Q7(4) are assumed so that they
satisfy the node-flow continuity equations, Equations (8a) . . . (8d). Let us
denote these values by suffix o as usual, and let us omit 4 and the enclosing
parentheses for simplicity. If AQ, . . . AQm are the loop-flow corrections
(clockwise positive) for loops I . . . I l l , respectively, Equations (8e) . . .
(8g) become, respectively,

0.2(Qo2 + AQ,)" - 0.5(Qo4 - AQ, + AQ,,)"

- 0.4(Qo3 - AQ, + AQ,,,)"

- 0.4(0,, - AQ, + AQ,,,)" = 0 (9a)

0.4(Qo5 + AQ,,)" - 0.6(Qo7 - AQ,,)" - 0.3(Qo6 - AQ,,)"

+ 0.5(Qo4 - AQ, + AQ,,)"= 0 (9b)

0.4(Q ol - AQ, + AQ,,,)" + 0.4(Qo3 - AQ, + AQ,,,)" + # 2 (4)

- /f,(4) = 0 (9c)

Substituting the values of 7/^4) and H2(4) from Equations (8h) and (8i),
respectively, in Equation (9c), we get

0.4(0,, - AQ, + AQ,,,)" + 0.4(Qo3 - AQ, + AQ,,,)"

+ 0.10262[(Qo3 - AQ, + AQ,,,) - (Qo4 - AQ, + AQ,,)

- (Q»6 - AQ,,)] + 0.076966[(Qol - AQ, + AQ,,,)

+ (Qo2 + AQ,)] + 0.34600 = 0 (9d)

On simplification, Equation (9d) becomes,

0.4(0.. - AQ, + A0.II)" + 0.4(0,3 - AQ, + AQ,,,)"

+ 0.10262(0.3 - Qo4 - Q»6 + AQ,,,) + 0.076966(Qol

+ Q.2 + AQ,,,) + 0.34600 = 0 (9e)

 Direct Method 347

The solution of Equations (9a), (9b), and (9e), either simultaneously by

the Newton-Raphson method or individually by the Hardy Cross method
[retaining only A(?i in Equation (9a), AQtt in Equation (9b) and AQm in
Equation (9e)] would yield the required dynamic analysis solution. Note
that only AQm, the correction term for the pseudo loop under considera-
tion, is present in the third and fourth terms (corresponding to H2 and Hu
respectively) of Equation (9e), while the correction terms for loops I and
II are absent. Further, herein also the solution of Equations (9a) . . . (9c)
yields the ordinary static analysis for time t = 4 hr. For the dynamic anal-
ysis, Equation (9c) is replaced by Equation (9e). Thus, herein also, only
the equation for the pseudo loop has changed, and the number of equations
for the dynamic analysis is also three as in the static analysis.
AH Equations. Assuming suitable initial values for //i(4) . . . H6(4),
AW equations can be written from Equations (7a). . . (7f). Retaining only
the correction for the node for which the H equation is formulated, the
equations can be solved iteratively by the Hardy Cross method. Herein
also the number of AH equations increases by M, the number of reser-
voirs.
Step 9. Because there are no pumps and valves in the network, checking
for their switching does not arise. Therefore, the solution obtained in step
8 for the static analysis at t = 4 hr is the required dynamic-analysis solu-
tion for time interval 0-4 hr. The various answers are the same as obtained
in step 12 of Example 12.1.
Note that in the formulation of Q and AQ equations, as the number of
equations required to be solved in dynamic analysis remains the same as
that in ordinary static analysis, the time required for dynamic analysis by
the direct method would be practically the same as that required for one
iteration of the iterative method. Thus, there are considerable savings in
time and effort in the direct method. The number of equations, however,
increases by M when H and AH equations are formulated and solved.
Therefore, depending upon the number of reservoirs relative to the num-
ber of other nodes, the time required for the dynamic analysis by the direct
method could be greater than that needed by the predictor-corrector itera-
tive method.

Example 12.3

Obtain the dynamic analysis for the network of Figure: Example 8.6 (re-
produced herein as Figure: Example 12.3) for time interval 0-2 hr. The
nodal demands at t = 0 hr are as given in Example 8.6. The nodal de-
mands at t = 2 hr are q3(2) = 0.12 m 3 /s, qA{2) = 0.10 m3/s, qs(2) = 0.18
nvVs, and q6(2) = 0.10 m3/s. The reservoir at node 2 has a constant cross-
348 DYNAMIC ANALYSIS

Sump Reservoir

sectional area of 1,000 m2. The water level in the sump at node 1 remains
constant at 80 m. The total network-demand volume, as obtained from the
demand curve, during the time interval 0-2 hr is 3640 m3.

Solution

We shall carry out dynamic analysis by the direct method using the lin-
ear theory method based on Q equations.
Steps 1 and 2. The static analysis solution of the network for time t = 0
hr is already obtained in Example 8.6, and is as follows:/^O) = 0.01852;
Rl(0) = 189.0, (2,(0) = 0.1619 m'/s; / 2 (0) = 0.02167, R2(0) = 2238,
Q2(0) = 0.0508 m 3 /s,/ 3 (0) = 0.02325; R3(0) = 10,123, Q3(0) = 0.0311
m 3 /s; / 4 (0) = 0.02322, R4(0) = 3,791, Q4(0) = 0.0324 nrVs; / 5 (0) =
0.02125, /?5(0) = 823.3, Qs(0) = 0.1016 nv7s,/ 6 (0) = 0.02392, R6(0) =
3905, <26(0) = 0.0565 mVs, / 7 (0) = 0.01776, R7(0) = 139.7, Q7(0) =
0.2781 m 3 /s; and hp(0) = 11.44 m.
Step 3. The total outflow from the network during time interval 0-2 hr
as obtained from the average rate of demand = f(0.44 -I- 0.50)/2] X
7,200 = 3,384 m3.
Step 4. The total network demand volume, as obtained from the demand
curve, during the time interval 0-2 hr = 3,640 m3.
Step 5. Correction factor C(0, 2) = 3,640/3,384 = 1.07565.
Step 6. For the sump at node 1, //,(2) = 80.00 m. For the reservoir
at node 2, as q2(t) = -Q7(i), we get H2(2) = 100.00 + 1.07565 X
[-0.2781 - Q7(2)] x 3600 •* 1,000, i.e.,

H2(2) = 100.00 - 3.87234[G7(2) + 0.2781] (1)

For the pump in pipe 1 as Qp(t) = Qt(t), hp(t) = 12.05 - 38.2[(2,

(r) - 0.0353J2. Therefore,

hp{2) = 12.05 - 38.2[£>,(2) - 0.0353] 2 (2)

 Direct Method 349

Step 7. Since the Darcy-Weisbach head-loss formula is used, t h e / , R and

R' values will need updating at each iteration during the solution of the 0
equations. For the first iteration instead of taking , Qx(t + At) values as
1.00 m3/s for linearization, it is preferable to assume them to be equal to
those obtained at the end of the previous time step of the dynamic analysis,
i.e., Qx{t). Therefore, herein ,0,(2) = 0,(0), x = 1 . . . 7. Naturally,
./x(2) = /*(0), and ,/?x(2) = Rx(0), x = 1 . . . 7. Therefore, using these
values as given in steps 1-2, the 0 equations for t = 2 hr are

(2,(2) - &(2) - 03(2) - 0 . 1 2 = 0 (3)

Q2(2) - Q4(2) + 0 5 (2) - 0.10 = 0 (4)

0 3 (2) + 04(2) + ft(2) - 0.18 = 0 (5)

-Gs(2) - 0 6 (2) + 0 7 (2) - 0.10 = 0 (6)

2,2380^2) + 3,7910^(2) - 10,12301(2) = 0 (7)

-3,7910^(2) - 823.301(2) + 3,9O50i(2) = 0 (8)

and

139.70,(2) + 823.301(2) - 2,23801(2) - 1890?(2)

+ hp(2) + 80 - H2(2) = 0 (9)

Substituting the values of H2(2) and hp(2) from Equations 1 and 2, re-
spectively, Equation (9). on simplification, becomes

139.70 7 (2) + 823.301(2) - 2,2380i(2) - (189.0 + 38.2)0?(2)

+ 2.6970,(2) + 3.8720 7 (2) - 6.921 = 0 (10)

Successively modifying the pipe resistance constants and linearizing

Equations (7), (8), and (10) and solving them simultaneously with Equa-
tions (1) . . . (4), we get the final solution as/,(2) = 0.01838, /?,(2) =
187.5, 0,(2) = 0.2138 m 3 /s; / 2 (2) = 0.02167, R2(2) = 2,239, 0 2 (2) =
0.0504 m 3 /s; / 3 (2) = 0.02302, R3(2) = 10,022, 0 3 (2) = 0.0434 m 3 /s;
/ 4 (2) = 0.02285, R4(2) = 3,731, 0 4 (2) = 0.0595 m 3 /s; / 5 (2) = 0.02122,
Rs(2) = 822.1, 0 5 (2) = 0.1092 m 3 /s; / 6 (2) = 0.0238, ^ 6 (2) = 3,885,
0 6 (2) = 0.0770 m 3 /s; / 7 (2) = 0.01775, R7(2) = 139.6, 0 7 (2) = 0.2861
nrVs; H2(2) = 100.00 - 3.87234(0.2861 + 0.2781) = 97.815 m; Qp
350 DYNAMIC ANALYSIS

(2) = g,(2) = 0.2138 nvVs; hp(2) = 12.05 - 38.2(0.2138 - 0.0353)2 =

10.38 m; volume of water supplied by the sump at node 1 = 1.07565 X
[0.1619 + 0.2138) x 7,200 -s- 2] = 1,455 m3; and the volume of water
supplied by the reservoir at node 2 = 1000 X (100.00 - 97.815) =
2,185 m3.

Example 12.4

Using the data given in Example 12.1, carry out the dynamic analysis of
the network for 0-24 hr taking a time step of 4 hr.

Solution

The dynamic analysis for 0-4 hr has been carried out in Example 12.1
and 12.2, by indirect and direct methods, respectively. Following either of
the two methods for the other time steps, the dynamic analysis for 0-24 hr
can be obtained.
For time steps of 4-8 hr and 8-12 hr, reservoir 1 is filled at a constant
rate of 14 nrVmin. Therefore, Equations (12.18) and (12.19) should be used
instead of Equations (12.16) and (12.17) in formulating the desired equa-
tions. Similarly, reservoir 2 is filled at a constant rate of 10 nWmin for time
intervals of 12-16 hr and 16-20 hr, and therefore, the equations stated for
reservoir 1 should be used.
The complete dynamic analysis solution is given in Table: Example 12.4
[162].

12.4 PROBLEMS

Solve the following problems by both indirect and direct methods.

Problem 12.1: Solve Example 12.1 for a time interval of 0-4 hr, during
which (a) the reservoir at node 1 is filled at a constant rate of 10 nvVmin,
(b) the reservoir at node 2 is filled at a constant rate of 8 m3/min, and (c)
both reservoirs are filled at the rates given in (a) and (b), respectively.
Problem 12.2: Solve Example 12.3 using (a) AQ equations and (b) H
equations.
Problem 12.3: For the network of Example 12.3, the nodal demands at
t = 4 hr are q3(4) = 0.08 m 3 /s, qA(4) = 0.16 nxVs, qs(4) = 0.12 nWs,
and q6(4) = 0.10 m3/s. The total network-demand volume, as obtained
from the demand curve, during the time interval 2-4 hr is 3180 m3. Carry
out the dynamic analysis for a time interval of 2-4 hr, using the solution
obtained in Example 12.3.
 TABLE: EXAMPLE 12.4.

Static Analysis at Time

Item Ohr 4hr 8hr 12 hr 16 hr 20 hr 24 hr

Oi, m3/min -0.3853 0.9525 2.5150 3.2257 1.8102 0.7376 0.5009

02, m3/min -0.1996 2.1981 3.8316 4.6296 3.1115 1.5682 0.9069
Q3, m3/min -1.3853 -1.0474 -0.4850 0.2260 -0.1896 -1.2626 - 0.4993
Q4, m3/min 1.2799 1.3983 0.7649 -0.4382 0.9499 1.3306 0.5518
Q5, m3/min 0.2804 1.9962 2.5965 1.5917 2.8615 1.2989 0.6582
06, m3/min 1.3196 2.6036 2.4034 1.6085 3.1385 2.1012 0.9420
0?, m3/min 0.7196 1.0037 1.4034 0.4082 1.1386 0.7014 0.3415
Q,, m3/min 0.58493 -3.1506 -6.3466 -7.8553 -4.9217 -2.3058 - 1.4079
q2, m3/min -3.9849 -5.0494 -3.6534 -0.9444 -4.2781 - 4.6943 -1.9931
q3, m3/min 1.0000 2.0000 3.0000 3.0000 2.0000 2.0000 1.0000
q4, m3/min 0.8000 1.6000 2.0000 2.6000 1.2000 1.6000 0.8000
q5, m3/min 1.0000 3.0000 4.0000 2.0000 4.0000 2.0000 1.0000
q6, nn3/min 0.6000 1.6000 1.0000 1.2000 2.0000 1.4000 0.6000
V1, m 3 2,000 1.671.7 3,853.3 5,473.3 3,923.3 3,035.0 2,573.3
V2, m 3 5,000 3,841.2 2,762.5 2,198.8 3,965.0 5.262.5 4,428.8
H,, m 101.200 101.003 102.312 103.284 102.354 101.821 101.544
H2, m 102.000 101.073 100.210 99.759 101.172 102.210 101.543
H3, m 101.268 100.637 100.105 99.784 101.154 101.594 101.433
H4, m 101.210 100.143 99.906 99.867 100.718 101.361 101.377
H5, m 101.172 98.704 97.564 98.921 97.914 100.712 101.192
H6, m 101.499 99.308 98.688 99.036 98.677 101.023 101.275
a
Reservoir 1 is being filled by the distribution network.
352 DYNAMIC ANALYSIS

Problem 12A: Using the data given in Example 12.3 and Problem 12.3,
carry out the dynamic analysis for the time interval 0-4 hr treating it as
one time step, and compare it with that obtained in Problem 12.3.
Problem 12.5: Carry out the dynamic analysis of the network de-
scribed in Example 12.1 for 0-24 hr if reservoir 1 is filled for 0-8 hr and
reservoir 2 for 16-24 hr at the given rates.
 Node Flow Analysis

13.1 INTRODUCTION

HILE designing water distribution networks, it is important that the

W network supplies the anticipated demands with adequate residual
heads at all nodes of the network during the entire design period. How-
ever, in due course of time one or more of the following situations may
arise [163-165]:

(1) The actual nodal demands are more than the anticipated design de-
mands due to an unanticipated accelerated growth over the entire
area. Thus, the system becomes inadequate even during the design
period.
(2) Even though the overall growth is as anticipated, it is an imbalanced
one causing the system to be deficient in some zones and more than
adequate in others.
(3) The withdrawals at some nodes are more than the design ones, due
to large diameter withdrawal connections, installation of pumps, pil-
ferage of water, or excessive leakage. This naturally would create a
deficiency in some other parts.
(4) The distribution system is used beyond its design period and thus
becomes deficient. (The usual design period of a distribution net-
work is 30 years. However, because the cost of replacement of the
entire network is usually enormous, the distribution network is con-
tinued in use beyond the design period by suitably strengthening it
wherever necessary and replacing only those pipes that have exces-
sive corrosion and leakage, thereby creating maintenance problems.)
(5) Some pipes are closed for cleaning, repairs, or replacement. This
may cause a temporary deficiency at some nodes.
353
354 NODE FLOW ANALYSIS

In all these situations either the nodal demands have changed or the dis-
tribution network has been deficient, and therefore it is unable to meet the
demands. In practice, therefore, it would be strengthened or expanded by
a trial-and-error procedure or by using one of the several available op-
timization techniques [85,86,166-1751. It would, however, be worthwhile
to know how the system would actually behave until the necessary modifi-
cations are carried out.
In the distribution network analysis methods that we studied in the ear-
lier chapters, whenever the nodal demands were specified, we presumed
that the network was capable of supplying them, and therefore it actually
supplied them. The primary aim of such analysis is to determine the heads
at the demand nodes and check that they are at least equal to the minimum
required residual heads, and the network is satisfactory. Therefore, Bhave
[163-165] termed this type of analysis as Node Head Analysis (NHA). The
NHA helps us in locating nodes that are deficient in heads and also in de-
termining the head deficiency at such nodes. We can, therefore, decide on
providing additional heads at source nodes, boosting pressures within the
network, strengthening in certain localities, providing check valves and
other similar modifications.
In the NHA it is presumed that the modifications are available, and
therefore, the actual nodal outflows are the same as the nodal demands.
However, such modifications are not available especially when the change
in nodal demands or the deficiency in the network is temporary. In these
situations the network fails to satisfy the demands at all nodes. It may be
able to satisfy demands at some nodes, partially satisfy demands at some
other nodes, while failing to provide even a drop of water at the remaining
nodes. To analyze the network for these situations, it is necessary to con-
sider the network as it is and determine the actual flows that would be
available at different nodes. Because this type of analysis would help in de-
termining the actual available nodal flows, Bhave [163-165] termed it
Node Flow Analysis (NFA).
Because NHA is a steady state analysis, the nodal demands are fixed and
therefore, even though more water may be available at a node, the with-
drawal remains fixed at the specified value. In NFA also, even though more
water may be available at a node, the withdrawal does not exceed the spec-
ified value; however, it may be less or even zero. Thus, the outflow at a de-
mand node in NFA lies between the specified value (the network is capable
of and therefore actually supplies the specified demand) and zero (the net-
work is unable to provide any water).
Before the general NFA principle is developed, we shall study the be-
haviour of a simple two-demand-node serial network under different con-
ditions.
 Two-Node Serial Network 355

13.2 TWO-NODE SERIAL NETWORK [163]

Consider a two-node serial network shown in Figure 13.1. Let S be the

source node (elevated service reservoir) and A and B be the demand-node
outlets with design demands of qAei and qp, respectively. Let the avail-
able head at source node be Hs and the levels of the outlets A and B be HA
and HB, respectively. Neglecting velocity heads as is the practice in net-
work analysis, HA and HB are also the minimum required hydraulic gradi-
ent level values for outflow to occur at A and B, respectively. Let HAl and
HBl be the actual HGL values for the design demands at A and B, respec-
tively, so that SAlBl is the hydraulic gradient for the design demands.
Rarely, if ever, HAt = HA and HB, = HB, and therefore, in practice, the
hydraulic gradient for the design demands SAXBX lies above SAB, the line
depicting the minimum required HGL values. In a rare case, if SAlBl co-
incides with SAB, any increase in the demand beyond the design demand
is not feasible at either A or B, and therefore maximum possible qA at node
A, qT* = qA* and similarly at node B qi?* = qf*. In practice, however,
SAiBi is at least partly above SAB and therefore an increase in nodal de-
mands would be feasible.
The street junctions As and fi, for outlets A and B, respectively, are also
shown in the figure. Let their levels be HAj and HB), respectively.
We shall now study the behaviour of this serial network for three situa-
tions: (1) increase beyond the design demand in the upstream nodal flow
and the corresponding effect on the downstream nodal flow, (2) increase
beyond the design demand in the downstream nodal flow and the corre-
sponding effect on the upstream nodal flow, and (3) increase beyond the
design demands in both nodal flows.

Figure 13.1 Elevated service reservoir and two-demand node serial network.
356 NODE FLOW ANALYSIS

13.2.1 INCREASE IN UPSTREAM NODAL FLOW

Herein we shall study the increase in upstream nodal outflow qA beyond

qAei and its effect on downstream nodal outflow qB.
Points S, At, A, and Bt of Figure 13.1 are shown in Figure 13.2. The
position of outlet B is varied so that different situations giving different rel-
ative positions of A and B can occur. Horizontal lines are drawn from A
and 5 , to intersect the vertical lines from fi, and A in B2 and A2, respec-
tively, as shown in the figure. AB3 is drawn parallel to AiBt. As qA in-
creases beyond qAc% the hydraulic gradient line SA,Bt changes to some
other position such as SCD. (As qB remains same as q$\ CD is parallel
to^.fl,.)
Now consider different positions of the outlet B. If B coincides with Bt
(a rare case), any increase in qA beyond qAc" results in a decrease in qB.
The decrease in qB is less than the increase in qA. (Why?) Ultimately, qB
becomes zero when the HG line becomes SA2Bt. The outflow qA can in-
crease further (qB remains zero) and reaches its maximum value when the
HG line becomes SA.
When B lies between B, and B2, outflow qA can increase beyond qAK
without affecting qB ( = qBc-) until the new HG line SCD is such that D
coincides with B. Thereafter, an increase in qA would result in a situation
similar to that in the earlier case.
When B coincides with B2, qA can increase beyond q^ without affect-
ing qB until D coincides with B2. Thereafter, qB decreases as qA increases
and qB becomes zero when qA becomes qTK and the HG line is SAB2.
When B lies between B2 and B3, qB ( = qlB^) is not affected by an in-
crease in qA until SCD is such that D coincides with B. Thereafter, as qA
increases further, qB decreases and attains its minimum value (>0) when
C coincides with A and qA = qTx- Thus, when B lies between B2 and

Datum
Figure 13.2 Hydraulic grade lines for different positions of downstream node for increase in
upstream nodal flow.
 Two-Node Serial Network 357

Datum
Figure 13.3 Hydraulic grade lines for different positions of downstream node for increase in
downstream nodal flow.

B3, qB may be affected by increase in qA but does not become zero, i.e.,
qp > qB > 0.
When B coincides with or lies below B3, qB ( = q'Bs) is not affected at
all by an increase in qA. When qA attains its maximum value, qT*, qB is
still equal to qf*, and SAB3 is the limiting HG line.

13.2.2 INCREASE IN DOWNSTREAM NODAL FLOW

Herein we shall study the increase in downstream nodal flow qB beyond

qp and its effect on the upstream nodal flow qA.
Points S, Au A, Aj and B, of Figure 13.1 are shown in Figure 13.3 and
the position of B is varied as before to get different situations. As qB in-
creases beyond qBc\ the hydraulic gradient line SAtBi changes to some
other position SCD. When qB increases, discharges in both pipes increase
and therefore both SC and CD become continuously steeper.
When B coincides with Bt (a rare case), qB cannot increase beyond
qt".
When B is lower than Bu as qB increases beyond q%^, the HG line be-
comes steeper and steeper and ultimately becomes SAB4. Thus, for any
position of B above and including B4, qB can increase beyond qi** without
affecting #4 ( = qdf).
When B is lower than B4, as qB increases further, qA decreases. The de-
crease in qA is equal to the increase in qB. (Why?) The HG line remains
SA for pipe 1 but becomes steeper thar\AB4 for pipe 2. When qB increases
further, a stage is reached when qA becomes zero, and the corresponding
358 NODE FLOW ANALYSIS

HG line is SABS. Thus, for B between B4 and B5, qA is between qACh and
zero.
When B is lower than Bs, for a further increase in qB the HG line goes
below SABS. This stage continues until the HG line becomes SAjB6 where
the maximum discharge carrying capacity of pipe 1 is reached. Thus, for
B between Bs and B6, qB can increase further, although qA remains zero.
(Actually in practice, when the HG line approaches SAjB6, air entrainment
takes place and the flow becomes discontinuous, however, we shall neglect
this aspect.)
When B is lower than B6, because subatmospheric pressure at Aj is not
feasible, there is no further increase in qB.
This two-node serial network is shown in Figure 13.4. In Figure 13.4(a),
the two outlets A and B are nearly at the same level and therefore any in-
crease in qA beyond qAcs may affect qB ( = qBa) and qB may even become
zero if outlet B is above or at the same level as that of A. However, any
feasible increase in qB will not affect qA ( = q^). In Figure 13.4(b), outlet
A is much higher than outlet B. Therefore, any increase in qB beyond ^^,es
may affect qA ( = qAcs) and qA may even become zero. However, any feasi-
ble increase in qA will not affect qB ( = qdBcs).
Water distribution networks usually resemble Figure 13.4(a) and there-
fore outflow at downstream nodes may get affected by an increase in out-
flow at upstream nodes. However, for distribution networks in uneven
terrain, for portions resembling Figure 13.4(b), downstream nodes in low-
lying areas may affect the outflow at upstream high-lying nodes. House
plumbing with overhead storage tanks resembles the situation of Figure
13.4(b).

13.2.3 INCREASE IN BOTH NODAL FLOWS

When both qA and qB increase beyond qAK and qp, respectively, de-
pending upon the increase and the relative positions of A and B, different

(Q) (b)
Figure 13.4 Two-node serial system: (a) outlets A and Bat nearly same levels; (b) outlet A much
higher than outlet B.
 Two-Node Serial Network 359

situations occur. Such situations can be analyzed following the procedure

outlined for the earlier two cases.

Example 13.1

In Figure 13.1, Hs = 100.00 m, HA = 94.00 m, HAj = 92.00 m, HB =

90.00 m, HBj = 87.00 m, q^ = 1.8 nWmin, q&* = 1.2 nrVmin, Rol =
0.6, and Ro2 = 2.5. The pipe head loss relationship is h = RQ2 in which
h is in meters and Q in cubic meters per minute. Determine: (1) the maxi-
mum value of qA without affecting qBcs and (2) the maximum value of qB
without affecting qA^.

Solution

For the design demands, the pipe discharges are: Qt = 1.8 -I- 1.2 = 3
nrVmin, and Q2 = 1.2 nrVmin.
.-. ft, = RtQi = 0.6 x ¥ = 5.40 m, and h2 = R2Q1 = 2.5 x
1.22 = 3.60 m.
.-. Hdr = 100.00 - 5.40 = 94.60 m > 94.00 m ( = HA), and Hp =
94.60 - 3.60 = 91.00 m > 90.00 m ( = // B ). Thus, an increase in qA and
qB beyond their design values is feasible.
Case 1. Let AqA be the maximum permissible increase in qA without af-
fecting qBe%. Considering path S4, for outflow to occur at A,

100.00 - 0.6(3 + AqA)2 < 94.00 (1)

Similarly considering path SAB, for outflow to occur at B, 100.00 — 0.6

(3 + A ^ ) 2 - 2.5 X 1.22 < 90.00, i.e.,

100.00 - 0.6(3 + AqAf < 93.60 (2)

For the maximum value of AqA, the condition given by Inequality (1) is
critical. Therefore, 100.00 - 0.6(3 + AqA)2 = 94.00 gives the maximum
value of AqA = 0.1623 nWmin. Thus, qA can increase to 1.8 + 0.1623 =
1.9623 nWmin for which HAl = 94.00 m, and HBl = 94.00 - 3.60 =
90.40 m > 90.00 m (=HB). For the maximum permissible increase in qA,
because Inequality 1 for node A is critical, an increase in qA is governed
by node A itself. Therefore, an increase in qA cannot affect q^.
Case 2. Let AqB be the maximum permissible increase in qB without af-
fecting qAts. From path S4,

100.00 - 0.6(3 + AqBf < 94.00 (3)

360 NODE FLOW ANALYSIS

and from path SAB,

100.00 - 0.6(3 + AqB)2 - 2.5(1.2 + AqB)2 < 90.00 (4)

From Inequality (3), AqB = 0.1623 rrrVmin, and from Inequality (4),
AqB = 0.1009 m3/min. Therefore, the maximum permissible AqB =
0.1009 nrVmin, and the maximum permissible value of qB = 1.2 +
0.1009 = 1.3009 nrVmin. HA, = 100.00 - 0.6(3 + 0.1009)2 = 94.23 m
(>94.00 m), while HB, = 90.00 m. As Inequality (4) for node B is criti-
cal, the maximum permissible increase in qB is governed by node B itself
and thus an increase in qB cannot affect qAcs.
Both nodes are independent in this example and the increase in outflow
at one cannot affect the design outflow at another.

Example 13.2

In Example 13.1, HA = 92.00 m and HA, = 90.00 m, while the other

data remain unchanged. Find: (1) the maximum value of qA without affect-
ing qB^ and the corresponding values of HAl and HB, and (2) the maxi-
mum value of qA notwithstanding its effect on q p, actual value of qB, and
the corresponding values of HAl and HBl.

Solution

Case 1: Let AqA be the maximum permissible increase in qA. Consider-

ing the paths SA and SAB, we get, respectively,

100.00 - 0.6(3 + AqAy < 92.00 (1)

and 100.00 - 0.6(3 + AqAf - 2.5 X 1.22 < 90.00 or

100.00 - 0.6(3 + AqA)2 < 93.60 (2)

Considering Inequalities (1) and (2), the condition of Inequality (2) is

critical. From this condition we get AqA = 0.2660 nrVmin.
.-. Maximum qA = 1.8 + 0.2660 = 2.0660 nWmin; qB = 1.2 mVmin
( = qfr)\ HAl = 93.60 m ( > 92.00 m); and HB, = 90.00 m.
Herein node B is critical and the maximum permissible increase in qA
without affecting qBa is governed by node B.
Case 2. When the effect on qp is neglected, q*t can increase further
and attain a maximum value when HAl = 92.00 m. Thus, if Q, is the dis-
charge in pipe 1 for this condition, 100.00 - 0.6Qi = 92.00 m, giving
Qt = 3.6515 nvVmin. Further because HA, = 92.00 m and HB, =
 Two-Node Serial Network 361

90.00 m, the discharge in pipe 2, Q2, is given by 92.00 - 2.5Q} = 90.00.

Thus, Q2 = 0.8944 nrVmin = qB, and qA = 3.6515 - 0.8944 = 2.7571
m3/min. Therefore, qA can increase to a maximum value of 2.7571 nrVmin,
but qB will reduce to 0.8944 nrVmin. Herein, HAl = 92.00 m and HB, =
90.00 m; and the HG line coincides with the line depicting the minimum
required HGL values.

Example 13.3

In Example 13.1, HB = 62.00 m and HB, = 60.00 m, while the other

data remain unchanged. Determine: (1) the maximum value of qB without
affecting qAei, (2) value of qB so that qA just becomes zero, and (3) maxi-
mum feasible value of qB. Find also the actual nodal HGL values for all
three cases.

Solution

Case 1. Let AqB represent the maximum permissible increase in qB.

Then, for paths SA and SB, respectively,

100.00 - 0.6(3 + AqBf < 94.00 (1)

100.00 - 0.6(3 + AqB)2 - 2.5(1.2 + AqB)2 < 62.00 (2)

From Inequality (1), AqB = 0.1623 nrVmin, and from Inequality (2),
AqB = 1.880 m3/min. Thus, the maximum permissible increase in qB
without affecting qAa is 0.1623 nrVmin. Therefore, qA = 1.8 nrVmin;
maximum qB = 1.2 + 0.1623 = 1.3623 nrVmin; HAl = 94.00 m; and
HBl = 100 - 0.6(3.1623)2 - 2.5(1.3623)2 = 89.36 m.
Herein, node A is critical and the maximum permissible increase in qB
without affecting qAcs is governed by node A.
Case 2. As seen in case 1, when the withdrawal at B increases beyond
qBCi, HA, decreases and just becomes 94.00 m when qB = 1.3623 nrVmin.
When the withdrawal at B increases further, HAt remains at 94.00 m but
qA decreases. Therefore, a stage is reached when HA, is 94.00 m and qA
just becomes zero. Thus, for this situation, HAt — 94.00 m, qA = 0, and
Qi = Q2 = qB- From path SA, 100.00 - O.602 = 94.00 m, which gives
Qi = 3.1623 nWmin = Q2 = qB. HereL the entire withdrawal of
qAcs = 1.8 m3/min at node A is transferred to node B. HB, = 100.00 -
0.6(3.1623)2 - 2.5(3.1623)2 = 69.00 m. Thus, qA = 0, qB = 3.1623
nrVmin, HAx = 94.00 m, and HB, = 69.00 m.
Case 3. The maximum feasible value of qB is obtained when either
(1) HBl = 62.00 m and HA, < HAj (=92.00 m), for flow to continue
362 NODE FLOW ANALYSIS

at A or (2) HA] = H^ = 92.00 m, and HBl < 62.00 m, whichever

condition is critical. In both these situations qA — 0. From condition 1,
100.00 - 0.6ql - 2.5q2B = 62.00, giving qB = 3.5012 rrrVmin, and
HA, = 100.00 - 0.6 X 3.50122 = 92.64 m > H* (=92.00 m). From
condition (2), 100.00 - 0.6q% = 92.00 m, giving qB - 3.6515 m3/min,
and HB, = 100.00 - 0.6 x 3.65152 - 2.5 x 3.65152 = 58.67 m <
62.00 m, which is not permissible. Therefore, condition 1 is critical and
gives: qA = 0; maximum feasible value of qB = 3.5012 nrVmin; HA, =
92.64 m > HAi (=92.00 m); and HB, = 62.00 m. Here node B is critical
but the HGL at A is less than 94.00 m, the value required for outflow to
occur at A. However, as HAl > HAn the flow continuity is maintained.
If HB had been less than 62.00 m, qB could have been increased and
would have attained a maximum feasible value of 3.6515 nxVmin when
HB = 58.67 m. Thereafter for any value of HB < 58.67 m, qB would re-
main 3.6515 m3/min.

Example 13.4

In the network shown in Figure: Example 13.4, node 1 is the source

node with HGL = 100.00 m, and nodes 2 . . . 6 are demand nodes. The

network has seven pipes, 1 . . . 7, with R values as shown (underlined) in

the figure. The head loss in pipes is given by h = RQ18S in which h is in
meters and Q in cubic meters per minute. The nodal demands and the min-
imum required nodal HGL values are as given in Table: Example 13.4.
(1) Carry out the NHA of the network
(2) What is the maximum possible outflow at node 3, without affecting
the outflows at all the other nodes?

Solution

1. The NHA of the network can be obtained by any one of the methods
described earlier. The NHA solution, thus obtained, is shown in columns
 TABLE: EXAMPLE 13.4.
(Values Used for Carrying Out Analysis Are Shown in Parentheses.)

NHA Solution for

NHA Solution Available Increased oh

<7, r e q . Hf, qr H,. Lj L/min

q.t H,,
Node m3/min m m3/min m m m3/min m
(1) (2) (3) (4) (5) (6) (7) (8)

1 100.00 -5.00 a (100.00) -5.153 a (100.00)

2 1.10 90.00 (1.10) 92.32 2.32 (1.100) 91.87
3 1.20 87.00 (1.20) 89.96 2.96 1.353 89.20
4 0.80 89.40 (0.80) 92.69 3.29 (0.800) 92.28
5 1.00 88.80 (1.00) 89.35 0.55 (1.000) (88.80)
6 0.90 88.00 (0.90) 88.86 0.86 (0.900) 88.22
a
Supply at source node.
364 NODE FLOW ANALYSIS

4 and 5 of Table: Example 13.4. The values used for carrying out the NHA
are shown within parentheses.
2. The available excess heads (//,• - Hf'n) at the demand nodes for the
NHA solution obtained earlier are shown in column 6. The least excess
head is 0.55 m, available at node 5.
When the outflow at node 3 increases, the available excess heads at all
nodes will decrease. As node 5 has the minimum available excess head, let
us assume that it would govern the maximum possible outflow at node 3
without affecting the outflows at all other nodes. Therefore, we take Hs =
H%m = 88.80 m as a known value and q3 as unknown. Thus, the un-
knowns are qu H2, q3, H3, H4, and H6. (Note that all network-solvability
rules are satisfied and the NHA solution is feasible.) The NHA solution is
obtained and shown in columns 7 and 8. As //, = Hf" at node 5, and
Hj > Hfm at nodes 2, 3, 4, and 6, the obtained solution is admissible.
Thus, the maximum possible outflow at node 3, without affecting the out-
flows at all other nodes, is 1.353 m3/min.

Example 13.5

Obtain the node-flow analysis of the serial network shown in Figure: Ex-
ample 13.5. The pipe head loss relationship is h = RQl 85 in which h is in

Node.O 1 2 3 4
a > • • • » m * *
Pipe, 1 2 3 4

meters and Q in cubic meters per minute. The source HGL is 100.00 m,
and the data for pipes and demand nodes are given in rows 1-3 of Table:
Example 13.5.

Solution

The complete NFA solution is given in Table: Example 13.5. Rows 1-3
show the given data. The corresponding NHA solution is given in rows
4-7 in which assumed values are shown in parentheses. As //, < Hfm at
nodes 2, 3, and 4, the solution is inadmissible and the NFA solution is nec-
essary.
After several trial-and-error iterations, the finally obtained admissible
solution is shown in rows 8-11^ in which the assumed values are shown in
parentheses.
Herein, the NFA solution is obtained by trial and error. The systematic
procedure for carrying out NFA is developed in the following pages.
 TABLE: EXAMPLE 13.5.
(Values Used in Carrying Out Analysis Are Shown in Parentheses.)

Row Node Pipe Node Pipe Node Pipe Node Pipe Node
No. Parameter 0 1 1 2 2 3 3 4 4

1 "a. 0.05 0.10 0.20 0.30

2 q; , m3/min
eq
2 4 2 3
3 H m "\ m 100.00a 92.00 94.00 91.00 88.00

4 qr m3/min -11.00 (2) (4) (2) (3)

5 Qx, m3/min 11 9 5 3
6 hx, m 4.222 5.826 3.928 2.290
7 Hr m (100.00) 95.778 89.952b 86.024° 83.734b

8 q, m3/min (2) 2.305 1.322 (3)

9 Q,, m3/min 8.627 6.627 4.322 3
10 hK, m 2.693 3.307 3.000 2.290
11 H., m (100.00) 97.307 (94.000) (91.000) 88.710
a
Available HGL at source.
"Solution inadmissible as H < H!™n.
366 NODE FLOW ANALYSIS

13.3 NODE CLASSIFICATION

In the previous section we considered a two-node serial network and dis-

cussed the effect of an increase in demand at one node on the demand at
the other. In Examples 13.1-13.3 the networks were adequate for design
demands, and we determined excess flow capacity available at a node,
without affecting the demand at the other node or disregarding the effect
on demand at the other node. Later, we extended this to a looped network
in Example 13.4. We could solve all these examples directly, without any
trial-and-error procedure. In Example 13.5 we considered a serial network
that was deficient for the specified demands, and we wanted to find as to
how the network would behave and to what extent it would cater to the
nodal requirements. Therefore, we obtained the NFA of the network. For
such a simple network, we could not obtain the NFA solution directly, and
we had to resort to a trial-and-error procedure. Herein, we shall now de-
velop a systematic iterative procedure to obtain the NFA of a distribution
network.
Before the general NFA theory is developed, it is necessary to under-
stand different categories of nodes. We had broadly classified the nodes as
source nodes and demand nodes in Chapter 6. We shall now further clas-
sify the demand nodes according to their HGL values and flows.

13.3.1 CLASSIFICATION ACCORDING TO HGL

1. Critical Node (CRIT). When the available HGL at a node, //, is equal
to the minimum required one, Hf" (//, = Hfn), the node is a critical
node.
2. Supercritical Node (SUPC). When the available HGL at a node is
more than the minimum required one (//,• > H]"n), the node is a super-
critical node.
3. Subcritical Node (SUBC). When the available HGL at a node is less
than the minimum required one (//,• < Hjm), the node is a subcritical
node.

13.3.2 CLASSIFICATION ACCORDING TO FLOW

1. Zero-Flow Node (ZERF). When the required demand at a node, qTp

is zero (q?q = 0), the node is a zero-flow node.
2. No-Flow Node (NOFL). When a demand node is unable to provide
water (q} = 0), even though there is a demand (qp > 0), the node is a
no-flow node.
3. Partial-Flow Node (PRTF). When a demand node partially fulfills the
demand (0 < qj < qp), the node is a partial-flow node.
 Node-Category Compatibility 367

4. Adequate-Flow Node (ADQF). When a demand node provides the re-

quired demand {qs = qp), the node is an adequate-flow node.
5. Surplus-Flow Node (SURF). When a demand node provides outflow
more than the required demand (qj > qp), the node is a surplus-flow
node.
6. Negative-Flow Node (NEGF). When the outflow at a demand node is
negative, (#, < 0, qp ~> 0), the node is a negative-flow node.
As stated earlier in section 13.1, the withdrawal at a demand node does
not exceed the specified required demand, qp in the NFA of a network.
Therefore, a demand node does hot behave like a surplus-flow node. Simi-
larly, a demand node cannot behave like a negative-flow node. However,
the categories of surplus-flow and negative-flow nodes have been defined
herein, because these categories transitionally occur in the node-flow anal-
ysis procedure described later. The flow-dependent node categories which
occur in practice are zero-flow, no-flow, partial-flow, and adequate-flow
shown by the symbols ® , o , o , and • , respectively.

13.4 NODE-CATEGORY COMPATIBILITY

The category of a node depends upon its HGL value and the flow. There-
fore, the nature of a node can be completely described by combining both
these categories. However, in practice, some HGL-dependent and flow-
dependent categories are not compatible with each other. For example, a
supercritical node cannot be a partial-flow node or a no-flow node. Simi-
larly, an adequate-flow node cannot be a subcritical node. Thus, the two
categories of a node can be combined only if they are compatible with
each other. Such compatibility of node categories is given in Table 13.1.
Note that a zero-flow node is compatible with any of the HGL-dependent

TABLE 13.1. Compatibility of Node Categories.

Node Category Compatible with Node Category

(1) (2)

Subcritical Zero-flow, No-flow

Critical Zero-flow, Partial flow
(exceptionally No-flow or Adequate-flow)
Supercritical Zero-flow, Adequate-flow
Zero-flow Subcritical, Critical, Supercritical
No-flow Subcritical (exceptionally Critical)
Partial-flow Critical
Adequate-flow Supercritical (exceptionally Critical)
368 NODE FLOW ANALYSIS

categories: critical, supercritical, or subcritical. Because negative-flow

and surplus-flow nodes do not occur in practice, these categories are not
included in the table.

13.5 NFA THEORY

In Chapter 11, we considered the system-energy concept in which the

flow in a network gets adjusted such that the expenditure of the system en-
ergy is minimum. Conversely therefore, when a distribution network is
deficient and thus unable to satisfy the nodal demands, the network will try
to meet the demands, as far as possible, under the given conditions. In
other words, the flow in the network will adjust such that the total outflow
from the network is maximum under the given conditions. Thus, the node
flow analysis problem can be looked upon as a network outflow optimiza-
tion problem subject to certain constraints.

13.5.1 PROBLEM FORMULATION

Consider a multisource looped distribution network having M source

nodes denoted by m, m = 1 . . . M, N demand nodes denoted by j =
M + 1 . . . M + N and X pipes denoted by JC, x = 1 . . . X. Then, the
objective function is given by

Maximize: Total Outflow = £ qt (13.1)

I

The constraints are as follows:

At all source nodes, the HGL should be equal to the given HGL. There-
fore,

Hm = Hom, for all m (13.2)

in which Hom = specified HGL value at source node m.

At the demand nodes, the node category compatibility given in Table
13.1 should be satisfied. Thus, the outflow at a demand node must be con-
sistent with its HGL-dependent category. Therefore, for an adequate-flow
node,

qj = -qp, if //, > Hf (13.3a)

For a partial flow node,

0 < qj < q™, if Hj = //•?"" (13.3b)

 NFA Theory 369

And, for a no-flow node,

qj = 0, if Hj < Hf (13.3c)

in which qfq = required demand at node j ; and Hfn = minimum re-

quired HGL at nodej for outflow to occur.
Now, at any instant, a demand node can either be an adequate-flow, a
partial-flow, or a no-flow node. Therefore, only one of Equations (13.3a)
. . . (13.3c) will be applicable. Thus, Equations (13.3a). . . (13.3c) consti-
tute a set of alternate constraints.
The node-flow continuity relationship must be satisfied at all source and
demand nodes. Therefore,

2 Q* + 9m = 0, for all m (13.4a)

connected
to m

& + qj = 0, for ally (13.4b)

connected
toy

The loop-head loss relationship must be satisfied for all loops of the net-
work. Therefore,

K = 0, for all loops (13.5)

loop

The complete optimization problem can now be written as

Maximize: Total Outflow = 2^ <h (13.6a)

i

Subject to:

Hm = Hom, for all m (13.6b)

qj = qp, if Hj > Hf" (13.6c)

or

0 < «/ < qP, XHj = Hf" (13.6d)

370 NODE FLOW ANALYSIS

or

ft = 0, if//, < HT" (13.6e)

J ] & + ?„ = 0. for all m (I3.6f)

connected
to m

G, + ft = 0. for all j (13.6g)

connected
to;

/?t = 0. for all loops (13.6h)

loop

The optimization problem given by Equations (13.6a) . . . (13.6h) is non-

linear with constraints of Equation (13.6c) . . . (13.6e) as alternate con-
straints.

13.5.2 PROBLEM SOLUTION

In the usual node-head analysis, at all demand nodes we take ft =

and thus ft does not depend upon the relative values of //, and H f".
Therefore, in the usual NHA, the constraint (13.6c) without the HGL con-
dition is present and the alternate constraints (13.6d) and 13.6e) are absent.
As seen earlier, the NHA involves the simultaneous solution of Equations
(13.6b), (13.6c), (13.6f), (13.6g), and (13.6h) for which any one of the
methods described earlier can be employed.
In Chapter 11 we have seen that the NHA of a network can be achieved
by treating the problem as an optimization problem. Conversely therefore,
the solution of the network flow optimization problem can be achieved by
carrying out the NHA of the network. However, the network flow optimi-
zation problem is complicated as compared to the optimization problems
of Chapter 11, because of the presence of the alternate constraints. How-
ever, in the NFA, only one of the set of alternate constraints (13.6c) . . .
(13.6e) must prevail at a node. .Therefore, at each node, if one constraint
from the set of alternate constraints is properly chosen, the solution of the
optimization problem and therefore the NFA of the network can be ob-
tained by any one of the NHA methods described earlier.
According to the unknown-number rule of the network-solvability rules,
for the NHA of a network to be feasible, the total number of the unknowns
(nodal heads, nodal flows, and pipe resistance constants) must be equal to
 NFA Theory 371

the total number of the nodes. In the NFA problem described herein, the
resistance constants of all pipes are known. At all source nodes, the nodal
heads are known but the nodal flows are unknown, the number of these un-
knowns being M. In the case of demand nodes, except for zero-flow de-
mand nodes (qp = 0), the nodal flow will lie between qf* and 0 depend-
ing upon which of the alternate constraints prevails. The number of these
unknowns, therefore, must lie between 1 (the network must have at least
one nonzero-flow demand node) and N (all demand nodes are nonzero-
flow demand nodes). Further, at all the Ndemand nodes, the HGL values
are also unknown and, therefore, the number of these unknowns will be N.
Thus, the total number of the unknowns for the entire network lies be-
tween M + N + 1 and M + IN, while the total number of the nodes and
therefore the total number of the available equations is M + N.
In order to bring down the total number of unknowns to M + N and
make NHA feasible, we shall treat either the demand or the HGL as
known at each demand node. Thus, if we choose qj = q'p as given by
constraint (13.6c) or q} = 0 as given by constraint (13.6e), the nodal flow
will be treated as known and the nodal head as unknown. On the other
hand, if we choose the constraint (13.6d), we shall take //, = Hf" and
therefore the nodal head will be treated as known and the nodal flow as
unknown. Thus, the total number of unknowns at any instant will be M +
N, equal to the number of nodes in the network.

13.5.3 SOLUTION PROCEDURE

The solution procedure consists of the following steps.

1. For the first NFA iteration, treat all demand nodes as adequate-flow
nodes (zero-flow nodes will remain as zero-flow nodes) and obtain the
NHA solution for the network. Determine the available HGL values at all
demand nodes and decide on their HGL-dependent category (subcritical,
critical, or supercritical).
2. Check the compatibility of the obtained node category with the as-
sumed node category by referring to Table 13.1. If the compatibility is
achieved for all nodes, the selection of a constraint from the alternate con-
straints is proper, and the solution of the optimization problem and thereby
the NFA solution of the network is achieved.
[For the first NFA iteration, because the assumed node category for all
demand nodes is adequate-flow, the HGL category compatible with this is
supercritical and exceptionally critical. Thus, if all demand nodes, except
zero-flow nodes, are obtained as supercritical (or exceptionally critical),
the NFA solution is achieved. (Zero-flow nodes are always compatible
with their head-dependent category.) In this case, the network is adequate
for the specified demands, and the NEA and NHA are identical.]
372 NODE FLOW ANALYSIS

3. When the node category compatibility is not achieved for a demand

node, depending upon the assumed node category, convert the obtained
node category to the one shown in Table 13.2 [165,176]. When node cate-
gory compatibility is achieved for a node, retain the assumed (not ob-
tained) node category for that node.
4. Treat the converted or retained node category as the assumed node
category and carry out the next NFA iteration. Determine the available
flow or HGL, as the case may be, for each demand node, decide its ob-
tained node category, and then go to step 2.

Example 13.6

Solve Example 13.5 by using the NFA procedure outlined earlier.

Solution

The solution obtained by following the NFA procedure is shown in Ta-

ble: Example 13.6. Rows 1-3 show the given data.
Step 1. For the first NFA iteration (rows 4-10), all demand nodes are
treated as adequate-flow nodes (row 4), available HGL values are deter-
mined (row 8) and the HGL-dependent category is obtained (row 9).
Step 2. The assumed and obtained node categories (rows 4 and 9, re-
spectively) are checked for compatibility with the help of Table 13.1 (row
10). Herein the node category compatibility is achieved for node 1 but it
is not achieved for nodes 2-4.
Step 3. As the node category compatibility is achieved for node 1, the
assumed node category of adequate-flow in iteration 1 (row 4) is retained
for iteration 2 (row 11). However, node category compatibility is not

TABLE 13.2. Conversion of Node Categories.

Assumed Convert to
Node Category Obtained Node Category Node Category
0) (2) (3)

Adequate-flow Supercritical, Critical3 Adequate-flow13

Subcritical Critical
Critical Negative-ftow No-flow
No-flow,a Partial-flow, Adequate-flowa Critical"
Surplus-flow Adequate-flow
Zero-flow Supercritical, critical,8 subcritical Zero-flow"
No-flow Supercritical Critical
Critical,8 Subcritical No-flowb
a
May occur exceptionally.
"Assumed node category is retaned because node category compatibility is achieved.
 TABLE: EXAMPLE 13.6.
(Values Used in Carrying Out Analysis Are Shown in Parentheses.)

Row Node Pipe Node Pipe Node Pipe Node Pipe Node
No. Item 0 1 1 2 2 3 3 4 4

Data
0.05 0.10 0.20 0.30
m3/min 2 4 2 3
,m 100.00 92.00 94.00 91.00 88.00

NFA Iteration 1
4 Assumed node category — ADQF — ADQF ADQF ADQF
5 Qy, m3/min (2) - (4) (2) (3)
6 O,, m3/min 11 - 9 - 5 3
7 hx, m 4.222 — 5.826 — 3.928 2.290
8 Hr m (100.00) — 95.778 - 89.952 86.024 83.734
9 Obtained node category — SUPC — SUBC SUBC SUBC
10 Is node category
compatibility achieved? — Yes — No No No

(continued)
 TABLE: EXAMPLE 13.6. (continued).

Row Node Pipe Node Pipe Node Pipe Node Pipe Node
No. Item 0 1 1 2 2 3 3 4 4

NFA Iteration 2
11 Assumed node category — — ADQF — CRIT — CRIT — CRIT
12 qr m3/min — — (2) — 2.305 0.850 — 3.472
13 Qx, m.3/min — 8.627 — 6.627 — 4.322 — 3.472
14 h~, m — 2.693 — 3.307 — 3 — 3 —
15 Hr m (100.00) — 97.307 — (94.000) — (91.000) — (88.000)
16 Obtained node category — — SUPC — PRTF — PRTF — SURF
17 Is node category
compatibility achieved? — — Yes — Yes — Yes — No

NFA Iteration 3
18 Assumed node category ADQF — CRIT — CRIT — ADQF
19 qr m3/min — — (2) — 2.305 — 1 322 — (3)
20 Ox, m3/min — 8.627 — 6.627 — 4.322 — 3 —
21 hx, m — 2.693 — 3.307 — 3 — 2.290 —
22 H,, m (100.00) — 97.307 — (94.000) — (91.000) — 88.710
23 Obtained node category — — SUPC — PRTF — PRTF — SUPC
24 Is node category
compatibility achieved? — — Yes — Yes — Yes — Yes
 NFA Theory 375

achieved for nodes 2-4. Therefore, as the assumed node category is ade-
quate-flow (row 4) and the obtained node category is subcritical (row 9),
the node category is converted, by referring Table 13.2, to critical (row 11).
Step 4. The next NFA iteration is carried out (rows 11-17). Herein, the
node category compatibility is achieved for nodes 1-3, but it is not
achieved for node 4.
Steps 2-4 are repeated and iteration 3 is carried out. Herein, node cate-
gory compatibility is achieved at all nodes. Thus, the NFA solution is
reached, and it is as shown in rows 19-22.
It will be observed that nodes 1 and 4 can meet the demand completely,
while nodes 2 and 3 can meet the demand only partially.

Example 13.7

Carry out the NFA of the serial network shown in Figure 1: Example
13.7. The pipe head loss relationship is h = RQl8S in which h is in meters
2m)'min 3 4 3

0 M5 1
1 000
? j2 3 3t CK30
0£00
f t
0 3
100-00m 95-00m 96-00m 88-00m 89-00m

and Q in cubic meters per minute. The source HGL is 100.00 m, and data
for pipes and demand nodes are shown in Figure 1: Example 13.7 and also
given in rows 1-3 of Table: Example 13.7.

Solution

The NFA is carried out as shown in Table: Example 13.7. The solution
is obtained in 5 NFA iterations, and is shown in Figure 2: Example 13.7.
2 0 4 1-831
183

41 2
* t t3 1 1 4
10000m 97748m 95-138m 89-918 89- 00m

Herein, some demand nodes have been observed to be negative-flow and

surplus-flow transitionally. Nodes 1 and 3 completely satisfy the demand,
node 4 partially satisfies the demand, while node 2 is unable to provide
any water.

Example 13.8 [165]

A water distribution network consists of a source node (labelled 1) and

six demand nodes (labelled 2-7) as shown in Figure 1: Example 13.8. The
 TABLE: EXAMPLE 13.7.
(Values Used for NFA Iterations Are Shown in Parentheses.)

Row Node Pipe Node Pipe Node Pipe Node Pipe Node
No. Item 0 1 1 2 2 3 3 4 4

Data
1 0.05 0.10 — 0.20 — 0.30 —
2 t7;'?a, rrt3/min — — 2 — 3 — 4 — 3
3 H; ; ", m 100.00 — 95.00 — 96.00 — 88.00 — 89.00

NFA Iteration 1
4 Assumed node category — ADQF — ADQF — ADQF — ADQF
5 qr m3/min — — (2) _ (3) — (4) — (3)
6 Qx, m3/min — 12 — 10 — 7 — 3 —
7 hK, m 4.960 7.079 _ 7.319 — 2.290 _
8 Hr m (100.00) — 95.040 — 87.961 — 80.642 _ 78.352
9 Obtained node category — — SUPC — SUBC — SUBC — SUBC
10 Is node category
compatibility achieved? — — Yes — No — No No
 TABLE: EXAMPLE 13.7. (continued).

Row Node Pipe Node Pipe Node Pipe Node Pipe Node
No. Item 0 1 1 2 2 3 3 4 4

NFA Iteration 2
11 Assumed node category — — ADQF CRIT CRIT — CRIT
12 qr m3/min — — (2) — -2.176 — 9.262 _ -1.917
13 Ox, m3/min — 7.169 5.169 — 7.345 — -1.917 _
14 hK, m — 1.912 — 2.088 — 8.000 — -1.000 —
15 H,, m (100.000) — 98.088 — (96.000) — (88.000) — (89.000)
16 Obtained node category — — SUPC — NEGF — SURF — NEGF
17 Is node category
compatibility achieved? — — Yes — No — No — No

NFA Iteration 3
18 Assumed node category — — ADQF NOFL — ADQF — NOFL
19 q;, m3/min — — (2) — (0) — (4) — (0)
20 Ox, m3/min — 6 — 4 4 — 0 —
21 hx, m — 1.376 — 1.300 — 2.599 — 0 —
22 Hr m (100.00) — 98.624 — 97.324 — 94.725 — 94.725
23 Obtained node category — — SUPC — SUPC — SUPC — SUPC
24 Is node-category
compatibility achieved? — — Yes — No — Yes — No

(continued)
3
oo

T A B L E : E X A M P L E 13.7. ( c o n t i n u e d ) .

Row Node Pipe Node Pipe Node Pipe Node Pipe Node
No. Item 0 1 1 2 2 3 3 4 4

NFA Iteration 4

25 Assumed node category — ADQF — CRIT — ADQF — CRIT

26 qr m 3 /min — — (2) — -0.982 — (4) — 2.151
27 O x . m 3 /m!n — 7.169 — 5.169 — 6.151 — 2.151 —
28 hK. m - — 1.912 _.. 2.088 — 5.762 — 1.238 _
29 H~, m (100.00) — 98.088 — (96.00) -. 90.238 — (89.00)
30 Obtained node category — — SUPC — NEGF — SUPC - PRTF
31 Is node category
compatibility achieved? — — Yes — No — Yes — Yes

NFA Iteration 5

32 Assumed node category — _ ADQF __ NOFL — ADQF — CRIT

33 qr m 3 /min — ._ (2) (0) -- (4) — 1.831
34 Q,. m-Vmin — 7.831 5.831 5.831 — 1.831 —
35 ht, m 2.252 _ 2.610 — 5.220 ... 0.918 _
36 Hr m (100.00) — 97.748 95.138 89.918 — (89 000)
37 Obtained node category — — SUPC — SUBC — SUPC — PRTF
38 Is node category
compatibility achieved? — — Yes — Yes — Yes — Yes
 NFA Theory 379

4-16

head loss in pipes is given as h = RQ2 in which h is in meters and Q in

cubic meters per minute. The R values of the pipes are as shown in the
figure. The available HGL at the source node and the minimum required
HGL at the demand nodes are given in column 2 of Table: Example 13.8.
The supply at the source node and the design demands are given in Col-
umn 3.
Obtain the NFA of this network when pipe 1-2 is closed for repairs.

Solution

The NHA for the design demands for the entire network (pipe 1-2 not
closed) is carried out, and the available HGL values at all nodes for the
design demands are given in column 4. All nodes are supercritical and the
design is satisfactory.
Pipe 1-2 is closed for repairs and the NFA solution of the network for
this condition is required. The NFA iterations are given in Table: Example
13.8. The values assumed for the iteration are shown in parentheses. All
demand nodes are assumed as adequate-flow for iteration 1 (column 5),
but all of them are found to be subcritical (column 8), and therefore, they
are converted to critical (column 10). At the end of iteration 2, nodes 2 and
5 are negative-flow, nodes 6 and 7 are partial-flow, while nodes 3 and 4 are
surplus-flow as shown by column 13. The negative-flow nodes 2 and 5 are
converted to no-flow, and the surplus-flow nodes 3 and 4 are converted to
adequate-flow. Critical nodes 6 and 7 (obtained as partial-flow) are re-
tained as critical (column 15) because node-category compatibility is
achieved for them. At the end of iteration 3, the obtained node categories
(column 18) are compatible with the assumed node categories (column 15)
for all demand nodes (column 19) and thus the NFA solution is reached
(columns 16 and 17). The complete NFA solution is also shown in Figure
2: Example 13.8. Herein, nodes 3 and 4 completely satisfy the demand,
 TABLE: EXAMPLE 13.8.
(Values Used for NFA Iterations Are Shown in Parentheses.)
Pipe 1-2 Closed. NFA Iteration

Available HGL Is Node

Required Design for Design Category
HGL, Demand, Demand, Assumed Outflow, HGL, Obtained Compatibility
Node m m3/min m Node Category m3/min m Node Category Achieved?
(1) (2) (3) (4) (5) (6) (7) (8) (9)

1 100.00 -5.000 100.00 -5.000 (100.00)

2 87.00 0.700 89.92 ADQF (0.700) 63.51 SUBC No
3 85.00 0.800 87.30 ADQF (0.800) 63.01 SUBC No
4 90.00 0.400 93.20 ADQF (0.400) 80.00 SUBC No
5 88.00 0.800 89.69 ADQF (0.800) 66.66 SUBC No
6 85.50 1.400 87.01 ADQF (1.400) 63.47 SUBC No
7 86.00 0.900 86.98 ADQF (0.900) 64.39 SUBC No
 TABLE: EXAMPLE 13.8. (continued).

Pipe 1-2 Closed. NFA Iteration

Available HGL Is Node

Required Design for Design Category
HGL, Demand, Demand, Assumed Outflow, HGL, Obtained Compatibility
Node m m3/min m Node Category m3/min m Node Category Achieved?
(1) (2) (3) (4) (10) (11) (12) (13) (14)

1 100.00 -5.000 100.00 -3.535 (100.00)

2 87.00 0.700 89.92 CRIT - 0.272 (87.00) NEGF No
3 85.00 0.800 87.30 CRIT 1.275 (85.00) SURF No
4 90.00 0.400 93.20 CRIT 1.597 (90.00) SURF No
5 88.00 0.800 89.69 CRIT -0.549 (88.00) NEGF No
6 85.50 1.400 87.01 CRIT 1.126 (85.50) PRTF Yes
7 86.00 0.900 86.98 CRIT 0.358 (86.00) PRTF Yes

(continued)
 TABLE: EXAMPLE 13.8. (continued).

Pipe 1-2 Cbsed. NFA Iteration

Available HGL Is Node

Required Design for Design Category
HGL, Demand, Demand, Assumed Outflow, HGL, Obtained Compatibility
Node m m3/min m Node Category m3/min m Node Category Achieved?
(1) (2) (3) (4) (15) (16) (17) (18) (19)

1 100.00 -5.000 100.00 -3.108 (100.00)

2 87.00 0.700 89.92 NOFL (0.000) 86.70 SUBC Yes
3 85.00 0.800 87.30 ADQF (0.800) 85.48 SUPC Yes
4 90.00 0.400 93.20 ADQF (0.400) 92.27 SUPC Yes
5 88.00 0.800 89.69 NOFL (0.000) 87.93 SUBC Yes
6 85.50 1.400 87.01 CRIT 1.381 (85.50) PRTF Yes
7 86.00 0.900 86.98 CRIT 0.527 (86.00) PRTF Yes
 General Observations
383

(92-27)

1-381

nodes 6 and 7 partially satisfy the demand, while no water is available at

nodes 2 and 5.

13.6 GENERAL OBSERVATIONS

The following general observations can be made on the theory and pro-
cedure of the NFA.
1. Each NFA iteration forms one NHA solution under the assumed node
category combination. Therefore, the NFA solution can be obtained by any
one of the NHA methods described earlier. However, when some demand
nodes are assumed as critical nodes, they behave as source nodes (HGL
known). Therefore, it is preferable to use methods based on H or AH
equations for NHA solutions in the NFA iterative procedure. In methods
based on Q or AQ equations, the formation of pseudo loops is necessary.
As pseudo loops may change from iteration to iteration, and as some de-
mand nodes become source nodes, these methods are not well-suited for
carrying out NFA of distribution networks.
2. The NFA procedure starts (iteration 1) with all nodes as adequate-flow
nodes. The total network outflow is maximum but the solution may be in-
feasible because certain constraints from the set of alternate constraints
may not be properly satisfied. (The exception is when the NFA and NHA
are identical.)
3. At successive NFA iterations, the number of demand nodes at which
node category compatibility is achieved increases, and the solution,
though still infeasible, successively approaches feasibility.
4. When the node category compatibility is achieved for all demand
nodes, the procedure ends. The NHA solution obtained in the last iteration
under the assumed node category combinations, satisfies the alternate con-
straints properly at all demand nodes, and thus provides the NFA solution.
384 NODE FLOW ANALYSIS

The total network outflow is less compared to the earlier infeasible solu-
tions, but is maximum for a feasible solution and yields the optimal solu-
tion to the optimization problem of Equations (13.6a) . . . (13.6h).
5. In the usual optimization procedures, the constraints are satisfied and
the solution is feasible at each iteration. The objective function is op-
timized successively, and the procedure ends when no further optimization
is feasible. In the NFA procedure suggested herein, at least some con-
straints are initially violated (an exception occurs when the NHA and NFA
are identical) and the solution is infeasible. In the iterative procedure,
however, the number of violated constraints decreases, and the iterative
procedure ends when the constraints are properly satisfied at all nodes.
6. No proof is provided herein to show that the NFA solution is the opti-
mal solution. However, it can be observed that a slight variation in the
NFA solution results either in a feasible but suboptimal solution or in an
infeasible solution wherein one or more constraints are violated.

13.7 PROBLEMS

Problem 13.1: Solve Example 13.1 if Rol = 0.5 and Ro2 = 3.0.
Problem 13.2: In Example 13.1, nodes A and B are independent of each
other. Find (a) the limiting value of HA beyond which node A will start af-
fecting the design demand at B and (b) the limiting value of HB beyond
which node B will start aflecting the design demand at A.
Problem 13.3: Using the data of Example 13.1, determine the HGL
values of points fi, . . . fi6 described in Figures 13.2 and 13.3.
Problem 13.4: The minimum permissible HGL values at nodes 3 . . .
6 of the network of Example 8.3 are 94.20, 94.10, 90.00, and 92.50 m, re-
spectively. Find (a) the maximum possible outflow at node 5 without af-
fecting the outflows at all the other nodes and (b) the maximum possible
increase in the resistance constant of pipe 2 without affecting the outflow
at any demand node.
Problem 13.5: Obtain the NFA of the serial network of Example 13.5
when the nodal demands at nodes 1 . . . 4 change to 3, 2, 4, and 5 nrVmin,
respectively.
Problem 13.6: Obtain the NFA of the serial network of Example 13.6
when the minimum required nodal heads at nodes 1 . . . 4 are 93.00,
90.00, 92.00, and 91.00 m, respectively.
Problem 13.7: Carry out the NFA of the network of Example 13.8 when
(a) pipe 1-4 is closed for repairs and (b) pipes 1-2, 2-3, and 4-7 are simulta-
neously closed for repairs.
 Calibration

14.1 INTRODUCTION

N the different network analysis methods that we studied so far, we

I used known parameters. The values of these known parameters were
given to us, and we used them to determine the values of the unknown pa-
rameters. In practice, however, for analyzing an existing network, the nec-
essary data have to be either collected in the field or predicted reasonably
accurately to formulate mathematical models of the networks. These math-
ematical models can then be used to carry out operation and planning
studies [177].
The applications in operation studies are the following:
(1) Study of areas where enough residual pressures are not available
(2) Planning operation strategies to satisfactorily meet the consumer de-
mands under different loading conditions
(3) Determination of the capacity of the network to supply flows for fire-
fighting purposes at strategic locations
(4) Training of the operators

The applications in planning are the following:

(1) Developing criteria for the design of water distribution systems
(2) Sizing and scheduling of the suggested improvements in long-range
planning of the distribution systems
(3) Upgrading existing systems systematically through rehabilitation of
some mains, identification and prioritization of improvements and
adjustments of pressure zone boundaries

14.1.1 DEFINITION

Before a network model is used for operation and planning studies, it

must be ensured that the model would predict, with reasonable accuracy,
385
386 CALIBRATION

the behaviour of the network under different conditions. Such a process is

termed calibration of the model.
Shamir and Howard [90] state that "calibration consists of determining
the physical and operational characteristics of an existing system and de-
termining the data that when input to the computer model, will yield real-
istic results." Cesario and Davis [177] define calibration as "the process of
fine-tuning a model until it simulates field conditions for a specified time
horizon (such as maximum-hour conditions) to an established degree of
accuracy." Walski [178] precisely defines calibration as "a two step process
consisting of: (1) comparison of pressures and flows predicted with ob-
served pressures and flows for known operating conditions (i.e., pump op-
eration, tank levels, pressure reducing valve settings); and (2) adjustment
of the input data for the model to improve agreement between observed
and predicted values."
For proper calibration, as stated by the Research Committee on Distri-
bution Systems set up by the American Water Works Association [179], the
model should be checked and adjusted until computed and observed field
pressures are within reasonable agreement for various levels and extremes
of demand, pumping, and storage. Calibration at one set of operating con-
ditions and water use does not necessarily imply calibration in general,
and therefore, a wider range of operating conditions and water use should
be selected. This induces confidence in the formulated model and the re-
sults predicted therefrom.

14.1.2 REASONABLE AGREEMENT

Reasonable agreement between the computed and observed field pres-

sures depends upon the accuracy of the measurements of the water levels
in reservoirs, heads developed by pumps, and pressures at test nodes, the
accuracy with which the elevations of test nodes can be estimated from the
contour maps, and the amount of effort, time, and money the model user
is willing to spend in fine-tuning the model. Water levels in the reservoirs
and heads developed by pumps can be measured fairly accurately. A well-
calibrated pressure gauge can measure pressure within an accuracy of ± 10
kPa ( ± l m ) . However, the elevation data for the test nodes may be of vari-
able quality depending upon the interval on the contour maps. The accu-
racy of such elevation data is usually 1.5-3.0 m. Therefore, according to
Walski [178], an average difference of ± 1.5 m (15 kPa) with a maximum
value of ±5.0 m (50 kPa) for a good data set and the corresponding values
of ± 3 m (30 kPa) and ± 10 m (100 kPa), respectively, for a poor data set
would be a reasonable target for models of large distribution networks.
Cessario and Davis [177] recommended an accuracy of 3.5-7.0 m (35-70
kPa) as a reasonable target.
 Data Collection and Preparation 387

Models have been calibrated on the basis of differences between the

computed and observed field pressures [177,178,180-182]. Such differences
are useful for spotting inaccurate boundary heads and locating areas for
which the model is poorly calibrated. However, as stated by W&lski [183],
judging the accuracy of the calibration through differences between the ob-
served and computed field pressures may be misleading. For example, an
accuracy of ± 2 m in pressure heads in a system having a head loss of 4
m from the source to a node with lowest head would lead to a poorer cali-
bration than that obtained using an accuracy of ± 5 m in pressure heads in
a system having a head loss of 50 m. Therefore, instead of judging the cali-
bration through differences, Walski [183] recommended judging the cali-
bration through ratios of the observed to the predicted pressure or head-
loss differences between the test nodes and the source node or nearby
boundary locations having known pressures or heads. The ratio is superior
to the difference because it is independent of the units used and the nature
of the hydraulic gradient, whether steep or flat. An accuracy of ± 3 0 per-
cent in the pressure- or head-loss ratio under different loading conditions
would be a reasonable target. It is, however, preferable to use both criteria,
ratios and differences, in judging the quality of calibration. The ratio will
be the guiding criterion for calibration quality in flat terrains with a low
value of maximum head loss, while the difference will be the guiding crite-
rion for judging the calibration quality in undulating terrains with a large
value of maximum head loss.
In the end, it is emphasised that the accuracy of calibration depends
upon the accuracy of the data set and the range over which the flows are
measured. It is quite probable that when a model is calibrated for a single
loading pattern, two or more errors in input data may compensate each
other, and the model may appear well-calibrated. The possibility of such
errors is reduced when a model is calibrated over a wide range of flows.

14.2 DATA COLLECTION AND PREPARATION

Data collection and preparation forms an important aspect of network

calibration. These data include (1) network skeletonization, (2) heads at
source nodes and pumps, (3) supply at source nodes, (4) consumption at
demand nodes, and (5) pipe details.

14.2.1 SKELETONIZATION OF NETWORK

A water distribution network may consist of hundreds of pipes varying

in diameter from less than 50 mm to more than even 1000 mm. If all these
pipes are included in the mathematical model, the model would be too
388 CALIBRATION

large to tackle. Therefore, it is common in practice to consider only the

arterial piping system [179] and either ignore pipes smaller than 100 mm
or 150 mm in diameter [82], or group them together and replace them by
equivalent pipes [69,70]. This makes the model manageable and calibra-
tion easier. This process of eliminating smaller-diameter pipes or replac-
ing them by equivalent pipes and considering only important pipes in prep-
aration of the mathematical model of a network is termed skeletonization.
In addition, skeletonization also includes transferring the estimated de-
mands served by eliminated pipes to appropriate nodes on the model,
elimination of tree or branching pipe regions unless a region contains
large-demand nodes or test nodes, and developing separate models for in-
dependent simulation of isolated regions.
Eggener and Polkowski [180] studied the impact of skeletonization on
the accuracy of network results. They selected a part of the distribution
network of Menamonie, Wisconsin, primarily a residential community
with a total population of 11,275. The network consisted of pipes varying
in size from 50 mm to 500 mm. The elimination of pipe sizes during skele-
tonization occurred as follows: 50 mm—100 percent; 100 mm—77.3 per-
cent; 150 mm—20.1 percent; 200 mm—8.6 percent; 250 mm—6.1 per-
cent; and 300, 400, and 500 mm—0 percent. The total network length was
reduced from 43.5 km to 29.6 km, approximately. They compared the re-
sults of the skeletonized model with the full model and observed that, in
general, the performance of the skeletonized model was fairly close to that
of the full model under different loading conditions. As a typical example,
the fire flow available at a pressure of 137.8 kPa at a test node in the skele-
tonized model was 5,280 L/min as compared to 5,660 L/min in the full
model. They also observed that the accuracy of the results obtained from
the skeletonized model increased when all small diameter pipes in the vi-
cinity of source nodes and also in the vicinity of nodes with large concen-
trated demands were included in the skeletonized model.
In general, therefore, small diameter pipes can be omitted during skele-
tonization, especially when such pipes are perpendicular to the usual di-
rection of flow or are near large-diameter pipes. On the other hand, small-
diameter pipes lying near source nodes, large water users, or not in the
vicinity of large-diameter pipes should be included in skeletonization by
grouping and replacing them through equivalent pipes.
The geometric schematic showing the skeletonized model includes the
retained and equivalent pipes, nodes, pumps, storage reservoirs, check
valves, pressure reducing valves, control valves, and other accessories, all
properly identified and numbered. This should also include pipe lengths
and diameters as read from maps, elevation of nodes, pump capacities and
elevations, storage tank capacities and overflow elevations.
 Data Collection and Preparation 389

14.2.2 HEADS AT SOURCE NODES

The heads at source nodes can be determined fairly accurately. The wa-
ter elevations in the reservoirs and the heads supplied by the pumps can be
measured with appropriate gauges.

14.2.3 SUPPLY RATES AT SOURCE NODES

The rate of water supply at the source nodes can be either measured or
estimated with a reasonable degree of accuracy. The supply rate from a
reservoir can be estimated by observing the depletion in reservoir water
level and thereby in volume over a small time interval over which the sup-
ply rate can be assumed to be constant. Similarly, the supply rate from a
pumped source node can be measured with appropriate meters or esti-
mated from the head-discharge curves of the pumps, which should prefera-
bly be freshly determined prior to network observations.

14.2.4 CONSUMPTION AT DEMAND NODES

The correct estimation of the rate of water consumption at demand

nodes is rather difficult. The rate of water consumption at a node depends
upon the population served, the type of demand, i.e., domestic, public,
commercial, or industrial, and the time of the day. The domestic con-
sumption is estimated by multiplying the served population by the mean
per capita consumption. Industrial, public, and commercial demands, if
any, in the zone served by a node are estimated and included. If unac-
counted-for water losses are significant, the average daily volume of unac-
counted-for water is uniformly allocated to the nodes of the network,
unless areas of high unaccounted-for water losses are previously iden-
tified. The accuracy of the consumption data depends on proper categor-
ization and segregation of the users into usage classification, accurate geo-
graphic location of large industrial users, and proper geographic allocation
of residential usage. Thus, the estimation of nodal demands is compara-
tively less accurate and may need adjustment during calibration.

14.2.5 PIPE DETAILS

The length and diameter of the pipes and the location of the various
valves can be obtained from the maps. However, proper selection of the
pipe head loss coefficients such as the Hazen-Wiliams coefficients or the
Darcy-Weisbach friction factors is rather difficult.
390 CAUBRATION

For estimation of pipe head loss coefficients, it is preferable to have a

few observations on some pipes of the network, plot a curve of head loss
coefficient versus pipe age, and then use the curve as a guide for the esti-
mation of pipe head loss coefficients. However, as pointed out by Eggener
and Polkowski [180], it should be remembered that the head loss coeffi-
cient of a pipe in a network is a lumped parameter representing not only
the pipe friction but also other head loss causes such as fittings, valves,
service connections, and pipe-alignment deviations. The coefficient also
depends on the velocity of flow, which is usually kept high during tests to
obtain a significant head loss in the short pipe length used for testing and
also to minimize the impact of local demand during testing. The intersect-
ing mains are also valved off during testing and therefore the losses at
branches and intersections that are present during normal operation are
eliminated during testing. Furthermore, if the varying degree of protection
offered by different coatings and the difference between the nominal and
actual diameters, especially for small-diameter pipes, are considered, the
head loss coefficient versus the pipe age curve based on a few observations
becomes less reliable. However, such a curve can serve as a guide, and
therefore using this curve and the standard curve and charts described in
Chapter 2, and relying on experience and judgement, the model user can
predict the pipe head loss coefficients.
In general, therefore, the skeletonization of the network and the data re-
garding the heads and supplies at the source nodes are fairly reliable and
do not need any adjustment during calibration. However, the data regard-
ing the nodal consumptions (except such as those measured during fire
flow tests), and the pipe head loss coefficients are less reliable and there-
fore may need adjustment during calibration.
The Research Committee on Water Distribution Systems of the Ameri-
can Water Works Association [179] feels that with reasonable field sam-
pling of arterial pipe friction coefficients (CHW values) "the only major
source of error in simulation of contemporary performance will be in the
assumed loading distributions and their variations," thus emphasizing the
adjustment of the nodal consumptions. On the other hand, Eggener and
Polkowski [180] feel that if tests are performed on a typical fall or winter
evening in a predominantly residential community and the demand of the
entire network and heavy nonresidential users is measured during the test
and if judgement and experience are used in allocating unaccounted-for
water, the pipe friction coefficients are the weakest piece of input informa-
tion in network modeling. Thus, Eggener and Polkowski emphasize the
adjustment of pipe head loss coefficients. From these conflicting views it
is clear that the input information on the assumed nodal consumptions and
the pipe head loss coefficients is less reliable and therefore may need ad-
justment during calibration. In practice, however, in some cases the data
 Calibration Methods 391

on nodal consumptions may be more reliable than that on pipe head loss
coefficients, or vice versa. For example, in a study on the Boston water
distribution network, Shamir and Howard [90] preferred to adjust pipe
head loss coefficients, because the data on nodal consumptions were more
reliable since they were highly correlated with the basic flow pattern pro-
vided by approximately twenty supply points. On the other hand, in a
study on the Minneapolis water distribution network, many pipe resis-
tances were fairly accurately known and therefore fine tuning of the model
was carried out by adjusting the resistance of a few important pipes.
It would always be possible to carry out localized adjustment in the
nodal consumptions and/or the pipe head loss coefficients and thus force
the model to fit the observations under a single demand pattern. However,
such a procedure would be neither logical nor systematic and would give
widely different calibrated models with different users, and such models
would fail to simulate the field network under other demand patterns.
Therefore, for the reliable calibration of distribution networks:
(1) As wide a range as possible of nodal consumptions should be consid-
ered. This may include minimum-flow, average-flow, peak-flow con-
ditions and fire-flow tests.
(2) Nodal flow consumptions and pipe head loss coefficients should both
be considered for adjustment.
(3) Calibration should not be localized but preferably be spread over the
entire network.
A good calibration technique therefore should preferably be able to con-
sider both nodal consumptions and pipe head loss coefficients for adjust-
ment, systematically decide the extent to which adjustment is necessary,
provide adjustment to all nodal consumptions and/or pipe head loss coeffi-
cients thereby spreading the calibration over the entire network, should
preferably be simple, and, if iterative, should yield calibration in a few
iterations.

14.3 CALIBRATION METHODS

The nodal demands and the pipe head loss coefficients (or the pipe resis-
tance constants) are initially assumed during calibration. When these
"values are assumed, the number of the unknown parameters—supplies at
source nodes and heads at demand nodes—is the same as the nodes in the
network and therefore a unique analysis of the network can be obtained.
This yields the values of the unknown parameters. However, when any one
or more of these unknown parameters such as the flow from a reservoir or
a pumping facility in a multisource network, the discharge or head sup-
292 CALIBRATION

plied by a booster pump in the network, or the head at a demand node are
measured, the measured values are compared with the corresponding
values obtained from the model analysis. If the discrepancy between the
measured and obtained values is more than a predetermined permissible
value, calibration of the model becomes necessary.
The calibration of the model is carried out by following one of the fol-
lowing two approaches:
(1) In the first approach of model calibration, one or more of the as-
sumed, i.e., predicted parameters, are adjusted, model analysis is
carried out, the obtained values are compared with the measured
values, and the procedure is repeated until satisfactory calibration is
achieved. In this approach the number of the unknown parameters is
the same as the number of the nodes in the network and the parame-
ters measured in the field are used explicitly, only to compare them
with the corresponding obtained values. Herein the measured values
are not an integral part of the model analysis.
(2) In the second approach of model calibration, the measured parame-
ters are treated as known parameters and are directly used in the
model analysis. This approach provides an additional equation for
each measured value and thus permits the model user to introduce an
additional unknown parameter for each measured value. The addi-
tional unknown parameter can be the head loss coefficient of a par-
ticular pipe in the model or a multiplying factor to correct some or
all predicted pipe head loss coefficients. In this approach the mea-
sured values are used implicitly in the model calibration.
Several methods, based either on the implicit approach or the explicit
approach are available for calibration of network models. These methods
can be classified into two categories: (1) methods in which only the pipe
head loss coefficients are adjusted and (2) methods in which both parame-
ters, pipe head loss coefficients and nodal demands, are adjusted. These
methods are now described.

14.3.1 METHODS ADJUSTING PIPE COEFFICIENTS

Herein, only the pipe head loss coefficients are adjusted and it is pre-
sumed that the assumed nodal demands are fairly accurate and do not need
any adjustment.

14.3.1.1 Method Proposed by Rahal et al.

Rahal et al. [182] proposed a calibration technique for adjusting pipe re-
sistance constants when discrepancy exists between the measured and ob-
 Calibration Methods 393

tained heads at demand nodes. Using the Newton-Raphson method, they

developed an expression for the sensitivities of the nodal heads to changes
in pipe resistance constants. This expression is then used to derive another
expression for the adjustment of the pipe resistance constants so that the
calculated nodal heads approach the measured values. The sum of the
squares of observed head errors is then calculated. One way of selecting
the pipes for the adjustment of resistance constants would be to choose
those pipes which would give the largest reduction in the sum of the
squares of head errors. This criterion is mathematically satisfactory and
has fast convergence. However, this may result in a large change in the re-
sistance constant of a pipe and thereby significantly modify the character-
istics of the network. Therefore, they suggested another criterion: select-
ing the pipe which yields the largest improvement in the sum of the
squares of the head errors for the smallest change in pipe resistance con-
stant. Using this criterion, the pipes are successively selected and their re-
sistance constants adjusted, until satisfactory convergence is achieved. The
method is an explicit one and may require several iterations for obtaining
satisfactory calibration.

14.3.1.2 Method Proposed by Ormsbee and Wood

Ormsbee and Wood [184] proposed a method for calibrating network

models when the discharge in a pipe leading from a pumping facility is
known or when heads are measured simultaneously at one or more nodes.
Each measured value permits the model user to introduce one additional
unknown. When only one value such as a discharge in a pipe or head at
a node is measured, this known condition is used to determine the global
adjustment factor which is used to adjust the pipe head loss coefficients of
all pipes in the network. When more than one measurement is made, the
network is divided into zones, the number of zones being equal to the
number of measurements, and the known measurements are used to deter-
mine zonal adjustment factors. The zones or groups of pipes can be
formed depending upon the age of the pipes and, if so desired, some pipes
can be excluded from groups and therefore from adjustment. The method
is direct, implicit, calibrates the network in only one iteration, and can be
applied to different operating conditions. However, it can consider only
one loading condition at a time. This method is illustrated in Example
14.1.

Example 14.1

A skeletonized model of a water distribution network is shown in Fig-

ure: Example 14.1. Node 1 is a source node with HGL of 60.00 m. Nodes
394 CALIBRATION

4OL/S
80

2-7 are demand nodes with estimated water use as shown in the figure.
The length, diameter, and the actual and predicted HW coefficients for the
pipes are given in Table: Example 14.1. The heads are measured at nodes
4 and 7 and the values are 54.10 m and 48.88 m respectively. Calibrate the
model assuming that the predicted demands at all nodes and the HW coef-
ficient of pipe 7 are reasonably accurate.
Solution

Because the heads are measured at two nodes, two additional conditions
are available and therefore the network is divided into two zones. Let these
zones be zone 1 consisting of pipes 1-6 with adjustment factor A; and zone
2 consisting of pipes 8 and 9 with adjustment factor B. (Pipe 7 is excluded

TABLE: EXAMPLE 14.1.

Actual Predicted Predicted Adjusted

Length, Diameter, HW HW Resistance HW
Pipe m mm Coefficient Coefficient Constant Coefficient
(1) (2) (3) (4) (5) (6) (7)

1 1,000 250 100 115 3.873 x 10-3 128.90

2 800 400 130 115 3.141 x 10"* 128.90
3 150 200 120 115 1.722 x 10-3 128.90
4 300 300 110 115 4.781 x 10"" 128.90
5 500 300 120 115 7.969 x 10-" 128.90
6 600 300 110 115 9.563 x 10-i 128.90
7 600 300 110 115 9.563 x 10-< 115.00
8 600 300 110 115 9.563 x 10-t 93.15
9 300 200 90 115 3.444 x 10-3 93.15
 Calibration Methods 395

from calibration.) The model has three basic loops, loops I . . . III. Two
pseudo loops IV and V are introduced as shown in the figure. Let the dis-
charges in pipes 1 . . . 9 be Q1 . . . Q9 respectively, in the directions
shown in the figure. Let the estimated resistance constants of pipes 1 . . .
9 for the predicted HW values be Rol . . . Ro9, respectively. The Q equa-
tions can now be written as

(2. + & - Q* - 40 = 0 (1)

& - G 3 - Cs - 100 = 0 (2)

Q4 + Q6 - 20 = 0 (3)

Qs - Q* + Qi - 80 = 0 (4)

Q* ~ Qi ~ Q9 - 70 = 0 (5)

Qg - 40 = 0 (6)

ARo2Q\8S + ARo3Q\*5 - ARolQ'ol8S = 0 (7)

ARoSQlss + ARrtQi" - AR.tQlM - AR*Q\M = 0 (8)

BRoSQl85 + Ro1QlM - AR«QlM - ARBlQ\M =0 (9)

ARolQ\ss + AR^Qi'5 + 54.10 - 60.00 = 0 (10)

BR*QlM + BR*Q\M + 48.88 - 60.00 = 0 (11)

The Rol . . . Ro9 values are calculated as shown in Table: Example 14.1
and substituted in Equations (7-11). Equations (l)-ll) are simultaneously
solved using linear theory and the values of A and B are 0.8095 and 1.4776,
respectively. The adjusted HW coefficients of the pipes are shown in Ta-
ble: Example 14.1.
14.3.2 METHODS SIMULTANEOUSLY ADJUSTING PIPE
COEFFICIENTS AND NODAL DEMANDS

Herein both the pipe head loss coefficients and nodal demands are ad-
justed.
14.3.2.1 Trial-and-Error Method

One of the methods used for simultaneously adjusting the predicted pipe
head loss coefficients and nodal demands is to use the conventional trial-
396 CALIBRATION

and-error approach. The model user relies on his experience and judge-
ment for adjustment. Several runs of the model might be necessary to ar-
rive at a reasonably acceptable calibration.
When the model is calibrated under two loading conditions—high and
low flows—qualitative guidelines proposed by Walski [178] can be used.
These include the following:
(1) Adjust the predicted pipe coefficients if there is a similar error at
high and low flows.
(2) Adjust the predicted pipe coefficients and nodal demands by similar
amounts if the model is accurate at low flow but inaccurate at high
flow.
(3) Adjust the predicted nodal flows if the model is accurate at high flow
but inaccurate at low flow.
14.3.2.2 Method Proposed by Walski

Walski [178,183] proposed a method that simultaneously adjusts the pipe

coefficients and the nodal demands. The method requires the following in-
formation that can be obtained during routine fire flow testing [185]:
(1) The heads at a test node t (a hydrant) corresponding to some lower
flow rate q, (hydrant closed) and higher flow rate q, + F (hydrant
open). Let these heads be H, and H,f, respectively.
(2) The flow F at the hydrant during the fire flow test
(3) The heads at nearly constant head locations, such as a reservoir,
pump, or a pressure reducing valve. Let these heads be Hs and Hsf
corresponding to flow rates q, and q, + F, respectively.
Let the demand node labels bej,j=l...N(tEN) and pipe labels
be x, x = 1 . . . X. Let the user's prediction of the nodal demands be qjp,
j = 1 . . . N; and of the HW coefficients be C,p, x = 1 . . . X. Let 5 be
the sum of the demands at nodes significantly affecting the hydrant test.
Let the user's prediction of the heads at the test node during the low and
high flows, as obtained by running the model asing the predicted qjp and
Cxp values be H,p and H,fp, respectively.
The section of the network from the node with known head to the test
node is represented by a single equivalent pipe. Using the Hazen-Williams
head loss equation, Walski [178] derived adjustment factors A and B as

A °L fl4n
A U4
~ b(S + F) - aS ' '

B = (14-2)
b(S + F) - aS ~ ^
 Calibration Methods 397

in which

ft - H, \"
a = (14.3)
ft - H,F
and

b =
ft/ - ft, \°- (14.4)

Using the adjustment factors A and 5, the adjusted nodal demands qja
and pipe coefficients CIU are obtained from

qJa = Aqjp (14.5)

= BCr, (14.6)

Walski's method is explicit and can be used for multiple loadings. It can
be used iteratively to improve calibration results. The method is illustrated
in Example 14.2.

Example 14.2

A skeletonized model of a water distribution network (same as in Fig-

ure: Example 14.1) is shown in Figure: Example 14.2. Node 1 is a source

Test node
5

node with HGL of 60.00 m and nodes 2 . . . 7 are demand nodes. Nodes
4 and 7 are used as test nodes with fire flows of 160 L/s and 75 L/s, respec-
tively. The observed and predicted HGL values at test nodes are given in
Table 1: Example 14.2. The length, diameter, and actual and predicted
398 CALIBRATION

TABLE 1: EXAMPLE 14.2.

HGL at Node 4 in Meters HGL at Node 7 in Meters

High Flow High Flow

Condition Low Flow (160 Us) Low Flow (75 L/s)

Actual 54.10 40.47 48.88 7.15

Predicted 53.93 40.41 50.54 19.87
Adjusted 53.58 38.80 49.57 8.61

HW coefficients for the pipes are the same as in Table: Example 14.1. The
actual and predicted nodal demands are given in Table 2: Example 14.2.
Calibrate the model using Walski's method.

Solution

As the fire flow tests are carried out at two test nodes, the network is
divided into two zones: zone 1 comprised of nodes 2, 3,4 and 5, and pipes
1 . . . 6, governed and therefore adjusted by the test results at node 4 and
zone 2 comprised of nodes 6 and 7, and pipes 8 and 9, adjusted by the test
results at node 7. The HW coefficient of pipe 7 is not considered for adjust-
ment as in Example 14.1.
As the test results for low and high flows at node 4 are in agreement, no
adjustment is proposed in zone 1. However, the test results at node 7 give
higher residual heads for predicted conditions and therefore nodal flows
and pipe resistance constants require adjustment in zone 2. Considering
test node 7,

60 - 48.88 0.54
= 1.091
60 - 50.54

60 — 7 15 \ 0 5 4
(
S7 = qo6 + qol = 70 + 40 = 110 L/s

F7 = 75 L/s

1.091 X 75
A7 = = 0.865
1.160(110 + 75) - 1.091 X 110

B .
1.091 ~ °793
 Calibration Methods 399

Using the value of B7 the CHW values for pipes 8 and 9 are adjusted as
91.2. The demands at nodes 6 and 7 are also adjusted and these values are
shown in Table 2: Example 14.2.
The adjusted values are used and the pressure heads at nodes 4 and 7 are
obtained as shown in Table 1: Example 14.2. The predicted HGL values at
node 4 have slightly deteriorated, but those at node 7 have considerably
improved and, in general, the calibration is satisfactory.

14.3.2.3 Method Proposed by Bhave

Bhave [186] proposed a method that also simultaneously adjusts the pipe
coefficients and the nodal demands, and is thus similar to Walski's method.
However, in Walski's method, the nodal demands and therefore the inflow
at the source node increases or decreases depending upon the value of the
adjustment factor A. In Bhave's method it is presumed that the inflow at the
source node can be measured or estimated fairly accurately, and therefore
it remains fixed during calibration. When the nodal demands are initially
assumed, they are adjusted, if necessary, so that their sum equals the esti-
mated inflow. (This adjustment, incidentally, would take care of the water
losses in the network.) In the calibration, however, redistribution of nodal
demands may take place so that some predicted nodal demands increase
and others decrease, the total increase being equal to the total decrease for
the entire network.
Consider a skeletonized network shown in Figure 14.1, in which s is a
source node and ty and t2 are demand nodes where heads are measured si-
multaneously under a particular demand pattern. Divide the network into
two zones, one for each test node as shown in the figure. Let AFi and AF 2
be the demand adjustments for zones 1 and 2, respectively. Naturally,
AF 2 = —AFi so that the demand adjustment for the entire network is
zero.

TABLE 2: EXAMPLE 14.2.

Demand in Liters per Minute

Node Actual Predicted Adjusted

2 30 40 40.0
3 120 100 100.0
4a 30 20 20.0
5 100 80 80.0
6 50 70 60.6
7b 20 40 34.6
Total 350 350 335.2
a
Rreflowof 160 L/s.
b
Fire flow of 75 L/s.
400 CALIBRATION

Zone 2
Figure 14.1 Skeletonized network.

Consider a path from 5 to f, as shown by thick lines in the figure. Let the
pipes on this path be labelled 1, 3, . . . , x, . . . . For this path,

n / s~\ \n _i D (i~\ \n _L_ _L_ J? (f~\ \n _1_ — TT J1

lp\^^Ip/ "l 3p\3^.3p/ *T" . . . " T " l Y r p ' Kf.r.p f I . . . •—• «* A - **• t\p

(14.7)
in which Rip, R3p, . . . Rxp, . . . - predicted resistance constants of pipes
1, 3, . . . , x, . . . ; Qlp, Q3p, . . . . Qxp, . . . = discharges in pipes 1, 3,
. . . , JC, . . . for the predicted nodal demands and pipe resistance con-
stants; Hs = head at source node 5; and H,,p = predicted head at test
node ti.
Assume that the entire correction AF, for zone 1 is applied to pipe 1 on
path s-tu and also that this correction reduces along the path s-tt in pro-
portion to the discharge carried by a pipe so that the adjustment for pipe
x, i.e., AQX = (Qxp/QiP)AFj. If Bis the global adjustment factor for pipe
resistance constants, then for path s-t,,

BRlp(Qlp BR3P(Q3P

BRXP Qx ', + . .. = H,~ H,, (14.8)

Expanding Equation (14.8), neglecting higher order terms of AF,, com-

bining with Equation (14.7) and simplifying, we get

B(HS - -J AFl = //, _ Hu (14.9)

 Calibration Methods 401

As B = 1, taking B = 1 for simplicity in the second term, Equation

(14.9) reduces to
n{Hs H p)
B(H, - H.J + ~ " AF, = ft, - ft, (14.10)

Similarly, taking path s-t2 and taking AF2 = —AF,, we get

n{Hs Htv)
B(HS - Hup) - ~ AF, = ft - ft, (14.11)
Sl

Simultaneously solving linear equations (14.10) and (14.11), the values of

B and AF, are obtained.
The adjusted pipe resistance constant Rxa and the adjusted HW coeffi-
cient Cxa are given by, respectively,

Rxa = BRxp (14.12)

=
C'° D0.54 C*P (14.13)

The zonal demand adjustments AF, and AF2( = - AF,) are distributed
in proportion to the predicted nodal demands in zones 1 and 2, respec-
tively. Therefore, the adjusted nodal demand q]a is given by

AFZ
1+ (14.14)

in which qjp = predicted demand at node;'; AFZ = total nodal flow adjust-
ment for zone z\ and Nz = set of demand nodes in zone z.
This method is explicit and can be used iteratively so that B — 1 and
AF —• 0, and is stopped when reasonable agreement is reached between
the observed and adjusted results.
As the nodal demands and pipe coefficients are simultaneously adjusted,
at least two observations are necessary. These can be obtained by measur-
ing heads at two demand nodes under a single loading condition or at a de-
mand node under normal and fire flow conditions or at a demand node
under two different loading conditions. When more than two observations
are taken, the network is divided into several zones and the number of Bs
and AFs is taken such that their sum exceeds the number of available ob-
servations by one. (This is because the last AF value equals the negative
sum of the other AF values.) The paths are selected such that all the paths
402 CALIBRATION

(1) together cover all zones, (2) lie in the general direction of flow, and (3)
contain pipes that carry the majority of flow towards the test nodes as seen
from the model run under predicted condition.

Example 14.3

Calibrate the skeletonized network of Figure: Example 14.2, reproduced

herein as Figure: Example 14.3, for the test observations as given in Exam-
ple 14.2, by adopting Bhave's method.

Zone,3

Zone.i

Solution

The model is run for the predicted condition and the analysis results are
given in Tables 1, 2 and 3: Example 14.3. As the heads are measured at two
test nodes under normal and fire flow conditions, four additional equations
are available. The network is divided into three zones as shown in Figure:
Example 14.3. These zones are zone 1 comprising pipes 1, 2 and 3; zone
2 comprising pipes 4, 5 and 6; and zone 3 comprising pipes 8 and 9. Let
the pipe adjustment factors for these zones be Bu B2, and B3, respectively.
Pipe 7 is excluded from adjustment as in Example 14.2. Nodes 2, 3, 4, and
5 are affected by flow at test node 4, and therefore they are grouped to-
gether. Similarly, nodes 6 and 7 are grouped together. Let the total adjust-
ment for nodes 2 . . . 5 be AF so that adjustment for nodes 6 and 7 is
— AF. Thus, we have now four unknowns—B t , B2, B3, and AF—for the
four known conditions.
We shall now consider path 1-3-2-4, comprising pipes 2 and 3 (zone 1)
and 4 (zone 2) and path 1-6-7 comprising pipes 8 and 9 (zone 3). Note that
the paths are selected such that all the three zones are covered, they lie in
the general direction of flow and contain pipes that carry the majority of
flow towards the test nodes as seen from model run results for the pre-
dicted condition as given in Table 1: Example 14.3.
 TABLE 1: EXAMPLE 14.3.

Predicted Condition HW Coefficient

Pipe Flow, L/s Pipe Head Loss, m

After
Actual HW HW Low High Low High Iteration Final
Pipe Coefficient Coefficient Flow Flow Flow Flow I Adjustment

1 100 115 51.17 90.23 5.66 16.19 139.4 132.8

2 130 115 183.46 293.99 4.89 11.71 139.4 • 132.8
3 120 115 27.10 69.89 0.78 4.49 139.4 132.8
4 110 115 38.32 120.13 0.41 3.40 87.2 111.9
5 120 115 56.39 124.12 1.40 6.01 87.2 111.9
6 110 115 -18.37 59.84 -0.21 1 87 87.2 111.9
7 110 115 5.07 15.69 0.02 0.16 115.0 115.0
8 110 115 114.98 200 64 6.26 17.57 92.3 90.8
9 90 115 40.00 115.00 3.19 22.57 92.3 90.8
 TABLE 2: EXAMPLE 14.3.

HGL at Node 4 in Meters HGL at Node 7 in Meters

High Flow High Flow

Condition Low Flow (160 L/s) Low Flow (75 L/s)

Actual 54.10 40.47 48.88 7.15

Predicted 53.93 40.41 50.54 19.87
Adjusted 54.21 38.94 50.31 10.65
(after iteration I)
Adjusted 54.09 40.45 48.89 7.18
(final)

TABLE 3: EXAMPLE 14.3.

Demand in Liters per Second

Predicted

After
Node Actual Predicted Iteration I Final

2 30 40 43.4 40.8
3 120 100 108.4 101.6
4a 30 20 21.7 20.4
5 100 80 86.8 81.2
6 50 70 57.1 67.5
7b 20 40 32.6 38.5
Total 350 350 350 350
a
Fire flow o1 160 L/s.
D
Fire flow o' 75 L/s.
 Practical Considerations 405

Considering path 1-3-2-4 and using the pipe flows and head losses from
Table 1: Example 14.3 for the predicted low and high flow conditions, we
get, respectively,

(4.89 + 0.78)3. +

= (60 - 54.10) (1)

+ 4.49,6, + 3.40ft + '•lg2 " ^ "° M >- AF

= (60 - 40.47) (2)

Similarly, for path 1-6-7 comprising pipes 8 and 9, we get for the low and
high flows, respectively,

(6.26 + 3.I9W, - ' - ^ x ( 6 C > - 50.54)

= (60 - 48.88) (3)

(.7.57 + 22.57)6, -

= (60 - 7.15) (4)

Simplifying and simultaneously solving Equations (1-4), B, = 0.701,
B2 = 1.669, B3 = 1.504, and AF = 20.3 L/s. The adjusted HW coeffi-
cients for zones 1, 2, and 3 are 139.4, 87.2, and 92.3, respectively. The
total flow adjustment of 20.3 L/s is distributed to nodes 2 . . . 5 in propor-
tion to the predicted flows. Similarly, the flow adjustment of —20.3 L/s is
distributed to nodes 6 and 7. The adjusted nodal flows at the end of itera-
tion 1 are shown in Table 3: Example 14.3. The HGL values obtained at
nodes 4 and 7 for the adjusted nodal flows and pipe HW coefficients are
shown in Table 2: Example 14.3. The final adjusted values obtained after
several iterations are also shown in the tables.

14.4 PRACTICAL CONSIDERATIONS

14.4.1 FINE-TUNING THE MODEL

Four network calibration methods proposed by Rahal et al. [182],

Ormsbee and Wood [184], Walski [178], and Bhave [186] are described ear-
406 CALIBRATION

Her. The method of Ormsbee and Wood calibrates the model implicitly in
only one iteration. Other methods consider the test results explicitly. If
necessary, several iterations can be carried out to finetune the model so
that the predicted results satisfactorily match the observed ones. However,
it should be remembered that the calibration methods provide adjustments
for groups of pipes and clusters of nodes, and therefore the adjusted and
observed results would not match for water use at individual nodes and for
the resistance constant of individual pipes. Therefore, in practice, a few
iterations would be sufficient to obtain calibration commensurate with the
accuracy of the collected data.
14.4.2 SOURCES OF ERROR IN FIELD DATA COLLECTION

Calibration methods, described earlier, adjust the predicted pipe head

loss coefficients or the water use or both. However, when the corrections
are large, the mode! user should not blindly apply the corrections and
change the pipe coefficients and/or the water use without trying to under-
stand why the initial predictions of these values were poor. It is worthwhile
to spend some time and investigate the sources of errors and gain valuable
insight into the behaviour of the water distribution system. Valve status
(fully open or fully closed) and the settings of the pressure reducing valves
have been reported as possible sources of error [158,180]. Walski [187J has
described in detail several sources of error encountered by him while cali-
brating the distribution system of Austin, Texas.
When large discrepancies are obtained between predicted and observed
values, thus necessitating wide changes in pipe coefficients and/or water
use. it is worthwhile to look for the following possible sources of errors
[178,187]:

(1) If the adjusted pipe coefficient is much lower than the initially pre-
dicted value (Cxa < C p ), it is implied that the water carrying capac-
ity of the pipes is less than anticipated. While this may be due to an
increase in pipe roughness, it may also be due to a closed or a partly
closed valve or a pipe that is installed and shown on the maps but not
actually commissioned.
(2) If the corrected pipe coefficient is much greater than the initial pre-
diction (Cxa > Cxp), it is implied that the water carrying capacity of
the pipes is more than anticipated. While this may be due to a lesser
increase in pipe roughness than anticipated, it may be also due to the
presence of a rich mesh of smaller diameter pipelines excluded dur-
ing skeletonization, or the presence of new large pipelines placed in
service in the area but not shown on the maps and therefore excluded
in the model.
 Practical Considerations 407

(3) If the adjusted water use is much less than the initial prediction
(qja -4 qjp), the model user should attempt to identify water uses not
in operation when the data were collected. For example, the field
data might have been obtained on a school holiday, or on a day when
the workers near the test node may be on strike. It may also be due
to some water users getting water from neighbouring pressure zones.
(4) If the adjusted water use is much greater than the initial prediction
(qJa > qjp) the model user should look for unusual water uses such
as the filling up of a municipal swimming pool, excessive lawn
watering, or new industries excluded during the survey. The error
may also be due to an illegal connection or a large main break. A
pressure reducing valve in operation but excluded in the model may
also show higher head loss indicating more water use than antici-
pated.
(5) If the pressure drops considerably during a fire flow test, nearby
water uses would get reduced during testing, especially in low pres-
sure zones. In such cases it would not be proper to simply add higher
discharge to the normal water use, but the predicted normal water
use should be lowered, in the fire flow test.
(6) When a model consistently predicts much higher heads than those
observed in the field, it is worthwhile to run fire hydrant flow tests
several times with several residual pressure gauges and reconstruct
the hydraulic grade line in the test area. This may reveal that the
water reaches the hydrant through a path not thought to be significant
earlier.
(7) A remote part of a distribution system may also be supplied by an-
other system. If the two systems are modeled as separate, the pre-
dicted heads might be too low, and if they are connected, the heads
may be too high. The results may agree better if the systems are con-
nected at some places in the model, or the water uses are redistrib-
uted between the two systems. In large zones, field testing may con-
tinue for a whole day. Because of the diurnal variation in water use,
it might be necessary to adopt different water use multipliers during
the testing to match the predicted and observed results.
(8) In dynamic model calibration, the reservoir water levels may not
match well. This may be because the pumps are turned on and off at
the beginning of a time step in the model while in reality they may
turn on and off during the time period. It might be necessary to
reduce the time step.

It is thus seen that there are several sources of error in field data collec-
tion and adjustment and therefore proper model calibration is not easy.
408 CALIBRATION

The methods described earlier help the model user in calibration but con-
siderable judgement and intuition is required on the part of the model user
for satisfactory calibration. Calibration involves description of the whole
from observations on parts and therefore as much data as possible should
be collected. Fire flow tests with several residual pressure gauges would be
useful in resolving the valve status and topology of the distribution system.
However, some discrepancies between model results and field observa-
tions are better resolved by posing specific questions to valve crew super-
visors, construction inspectors, and operations personnel. Many problems
arise in calibration due to the lag between installation of new pipes and
showing them on system maps. Accurate and updated maps of distribution
systems are a prerequisite for good calibration.

14.4.3 FIELD EQUIPMENT

Walski [188] has given an exhaustive list of the equipment and described
their utility in collection of data in the field. Such equipment includes the
following:
(1) Regularly calibrated pressure gauges, preferably of the same type
and size, ranging from 100 kPa to 2,000 kPa (To obtain good results,
the selected range should barely cover the range of the pressures ex-
pected in the system.)
(2) Pitot gauges for measuring hydrant discharges having a usual range
of 2,500-5,000 L/min, and vernier calipers to measure the inside di-
ameter of the hydrant outlets
(3) Hydrant-cap and pitot fittings
(4) Hydrant wrenches and pliers
(5) Safety equipment such as safety shoes, reflecting vests, hard hats,
traffic cones, and flags
(6) Recording equipment including clipboard, papers, data sheets, ordi-
nary and video cameras
(7) Street maps and distribution system maps at several scales
(8) Communication equipment
(9) Survey equipment for elevation data collection
(10) Equipment for differential pressure measurement for determination
of the head loss coefficients of pipes
 1

Miscellaneous Topics

15.1 COMPARISON OF COMMON ANALYSIS METHODS

EVERAL methods for analyzing water distribution networks have been

S described in earlier chapters. These methods included the Hardy Cross
method (Chapter 8), Newton-Raphson method (Chapter 9), and the linear
theory method (Chapter 10). Other methods such as the one using electri-
cal analyzers, that based on unsteady behaviour during start-up, and that
through optimization principles have also been described (Chapter 11).
The use of electrical analyzers is more or less obsolete, while network
analysis through unsteady behaviour during start-up or that based on op-
timization principles is not common. The dynamic analysis (Chapter 12)
and node flow analysis (Chapter 13) are achieved through repeatedly
carrying out node-head analysis of the network. Thus, the methods which
are commonly used in network analysis are the Hardy Cross, Newton-
Raphson, and the linear theory methods. These methods are compared in
Table 15.1 [95,128,189].

15.2 SENSITIVITY ANALYSIS [88]

During the analysis of water distribution networks, sometimes it is nec-

essary to investigate the behaviour of the network when some nodal heads,
nodal flows, or pipe resistances change. When many parameters change si-
multaneously, such as the demands at nodes and the heads at reservoirs
during different parts of the day, a fresh analysis of the network is neces-
sary.
However, there are situations in which it is necessary to know the impact
of change in a single parameter on the behaviour of the network. For ex-
ample, it may be necessary to know how the network would behave when
the consumption at a single node increases, resistance of a pipe changes,
409
 TABLE 15.1. Comparison of Common Network Analysis Methods.

Hardy Cross Method Newton-Raphson Method

Based on Based on Linear Theory Method Based on

Nodal Heads
Sr. Balancing Balancing Nodal Pipe (Finite Element .
No. Item Heads Rows Heads Loops Discharges Method)

1 Unknown parameters AQ AH H AQ 0 H
and equations used
2 Number of unknowns Minimum Medium Medium Minimum Maximum Medium

3 Initialization Necessary Not Necessary

4 Mode of analysis Analysis in parts, considers Analysis as a whole, all loops or nodes are considered simulta-
one loop or node at a time neously
5 Suitability for hand Suitable. Networks with Not suitable except for small networks
calculation several nodes can be ana-
lyzed if sufficient time and
manpower are available
 TABLE 15.1. (continued).

Hardy Cross Method Newton-Raphson Method

Based on Based on Linear Theory Method Based on

Nodal Heads
Sr. Balancing Balancing Nodal Pipe (Finite Element
No. Item Heads Rows Heads Loops Discharges Method)

6 Number of iterations Usually large Small and usually not more than 10
7 Effect of network size Increases as the size of the Generally independent of the size of the network
on the number of network increases
iterations
8 Time per iteration Small Large
9 Effect of network size Increases gradually as the Increases rapidly as the size of the network increases
on time per iteration size of the network in-
creases

10 Problem of conver-
May be present Reported to be absent
gence

11 Computer storage Small and increases grad- Large and increases rapidly with the size of the network
capacity requirement ually with the size of the
network
412 MISCELLANEOUS TOPICS

or the head available at a reservoir in a multisource network decreases.

Such situations arise when network models are calibrated by changing sin-
gle parameters, and also in evaluating proposed modifications. This type
of analysis which evaluates the effect on other parameters due to a change
in a single parameter is termed sensitivity analysis.
When a steady state NHA solution of a network is available, a sensitivity
analysis can be carried out without obtaining a fresh additional solution.
Let yk be the parameter which changes and we want the sensitivity analysis
of the network for the change in v*. Now, in a network of J nodes and X
pipes we have 2J + X parameters (J nodal heads, J nodal flows, and X
pipe resistances). Out of these we can select a set of any J parameters (this
set does not include yk and also satisfies all network solvability rules) on
which the effect of change in v* can be obtained. Let this set of variables,
termed free variables, be denoted by Kand the variables by vf. We are in-
terested in the derivatives dvf/dyk in which vf and \\ can each be nodal
head, nodal demand, or pipe resistance.
As the network with J nodes is balanced, the continuity equation at each
node yields

Fj = 0,j = 1 . . .J (15.1)

Taking the derivative of Equation (15.1) with yk,

-j± = 0,./ = 1 . . . J (15.2)

Expansion of Equation (15.2) yields

= 0,j = 1 . . .J (15.3)

which may be written in the matrix form

(15.4)
dv dv* dyk

in which [ ] denotes a matrix and j j a column vector. The set of J equa-

tions given by Equation (15.4) has J unknowns—the partial derivatives
dVf/dyk. The partial derivatives dFjldyk and dFj/dvf are obtained from the
balanced network. The solution of Equation (15.4) gives the rate of change
of J selected variables vf, v, £ V, with respect to change in the variable
>\, i.e., the variations in vf for a unit change in yk. The results from the
sensitivity analysis can also be used to evaluate the sensitivity of yk to
 Sensitivity Analysis 413

changes in J selected free variables vf. The results from sensitivity analy-
sis are approximate because they are derivatives at a point rather than
finite changes. However, sensitivity analysis is a convenient and rapid way
of comparing the effect of all possible changes.

Example 15.1

The network shown in Figure: Example 15.1 has node 1 as source node
with HGL = 100.00 m and nodes 2 . . . 6 as demand nodes with demands
WO-OOrn
19

and minimum required HGL values as given in Table: Example 15.1. The
head loss in pipes is given by h = RQl S5 in which /; = head loss in
meters, R = pipe resistance constant (shown underlined in the figure),
and Q = pipe discharge in cubic meters per minute. The NHA solution of
the network is given in the table. Using sensitivity analysis, determine the
maximum possible outflow at node 3 without affecting the outflows at all
other nodes.

Solution

When the outflow at node 3 increases, the residual heads at all nodes de-
crease. Therefore, for the maximum possible outflow at node 3 without af-
fecting the outflows at the other nodes, the available HGL values at all
nodes, including node 3, should not be less than the minimum permissible
values. Thus, we are interested in knowing the impact of an increase in q3
on the nodal HGL values. Therefore, qu H2 . . . H6 are selected as the six
free variables. The node-flow continuity equations at nodes 1 . . . 6 are,
respectively,

100 - / / , \ 0 5 4 /100 - //4\054

+ <?. = 0 (1)
2.0 1.0

100 - H2\0S4 [H, - H

F2 = - 1.10 = 0 (2)
2.0 2.5
 TABLE: EXAMPLE 15.1.

NHA Solution Available Permissible H ; for

Free dv, Increased
qr, Ljmin
«,, H,. Variable Sq3 Qi.
Node m3/min m m3/min m m v
i m3/min m
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

1 100.00 -5.00 100.00 <7i 1.000 100.00

2 1.10 90.00 1.10 92.32 2.32 H2 - 2.959 0.784 91.87
3 1.20 87.00 1.20 89.96 2.96 H3 -4.898 0.604 89.21
4 0.80 89.40 0.80 92.69 3.29 H4 - 2.634 1.249 92.29
5 1.00 88.80 1.00 89.35 0.55 H5 -3.598 0.153 88.80
6 0.90 88.00 0.90 88.86 0.86 H6 -4.178 0.206 88.22
 Sensitivity Analysis 415

- // 3 \ 0
<?3 = 0
2.5 4.0 6.0
(3)

(4)

1.00 = 0 (5)
2.0 4.0

3X -°-non9 0 = (6)

Note that the selected variable q3 is also taken as an unknown parameter

in addition to the six free variables qx, H2 . . . H6.
Partially differentiating and writing Equation (15.4) in the matrix form,
we get

dF1
dqt 9// 2 u dH4 dq3

dF2 3F 2 9// 2
0 0
dH2 dH3 dq3
9^3
dF, 9F, dF
dH2 dH4 dH6 dq3
_ _ (7)
9^3
dF4
0U o
U ^ ^ ^ 0
9// 3 dH4 dHs dq3
9tf4
1FS dFs dFs dFs
0 0 0
oH4 9<?3 dq3
9// 5
dFb dF6 BHb
0 0
dH3 dHs dH6 dq3 dq3
416 MISCELLANEOUS TOPICS

Evaluating and substituting the values of the partial derivatives. Equa-

tion (7) becomes

1 0.1454 0 0.2163 0

0 -0.3672 0.2218 0

0 0.2218 -0.6683 0.1609 0 0.2856

0 0 0.1609 -0.5905 0.2133 0

0 0 0.2133 -0.5679 0.3546

0 0 0.2856 0 0.3546 -0.6402

dH.,
-1

x (8)
8HA
0
dq3

dHs
0
dq3

dH,
0
dq3

The solution of Equation (8) yields the values given in column 8 of Ta-
ble: Example 15.1. These values are the changes in the values of the free
 Sensitivity Analysis 417

variables for an increase of 1 nvVmin in the consumption at node 3. As a

result, qt would increase by 1 m3/min as expected, H1 would decrease by
2.959 m, H3 by 4.898 m, and so on. Dividing the available excess head in
column 6 by the decrease in head in column 8, we get the permissible ex-
cess value of q3 as governed by the free variable under consideration.
Thus, q3 can increase by 0.784 nWmin so that H2 is just 90.00 m, the mini-
mum required value. Therefore, q3 can increase by 0.784 mVmin, without
affecting the outflow at node 2, and so on. As the least permissible in-
crease in q3 is 0.153 nrVmin, governed by node 5, the maximum possible
outflow at node 3 without affecting the outflows at the other nodes is
1.20 + 0.153 = 1.353 nrVmin.
This example is the same as Example 13.4. Therein, we compared the
available excess heads at the demand nodes. Because the available excess
head was the least at node 5, we presumed that the maximum possible out-
flow at node 3 would be governed by node 5. We, therefore, considered // 5
as a known parameter with the value of 88.80 m, the minimum permissible
value, and q3 as an unknown parameter. Taking q,, H2. q.i, H3, H4, and H6
as the unknown parameters, we carried out a fresh network analysis and
obtained the value of q3. As is seen, the values of q3 are identical by both
methods for a small increase in q3.
For the increased value of q3, the //, values at the nodes are also calcu-
lates.' "nd shown herein in column 10 of the table. These values are nearly
the same as those obtained by exact analysis in Example 13.4.

Example 15.2

Find the available heads at all demand nodes if the resistance constant
of pipe 2 in Figure: Example 15.1 is 1.5 instead of 1. Use the NHA solution
given in Table: Example 15.1 and adopt the sensitivity analysis.

Solution

Herein we would like to know the impact of the change in R2 on the

available heads at the demand nodes. Therefore herein yk is R2 and the free
variables vsare <?,, // 2 , . . . H6 as in Example 15.1. Therefore Equation (7)
of Example 15.1 will also serve herein except that q3 is replaced by R2 in
the two column vectors. Similarly, corresponding to Equation (8) of Ex-
ample 15.1, we herein get an equation in which the matrix is the same as
that in Equation (8) of Example 15.1; R2 replaces q3 in the first column
vector, and the entries in the column vector on the right-hand side are
- 1.5809, 0, 0, 1.5809, 0, and 0, respectively. The solution of the equation
giving the values of the free variables is given in Table: Example 15.2. The
values of the free variables as obtained from the actual analysis are also
418 MISCELLANEOUS TOPICS

TABLE: EXAMPLE 15.2.

Value of v, from

Free Variable dv, Sensitivity Actual

vf dR2 Afl2 Av, Analysis Analysis

3
<7i. m /min 0 0.5 0 5.00 5.00
H2, m -2.516 0.5 -1.258 91.06 91.15
H3, m -4.165 0.5 -2.082 87.88 88.01
HA. m -5.618 0.5 -2.809 89.88 90.11
H5, m -4.999 0.5 -2.500 86.85 87.04
H6, m -4.627 0.5 -2.313 86.55 86.70

given in the table. Some difference is observed between the two results be-
cause the result from the sensitivity analysis is approximate, as stated
earlier.

15.3 PITFALLS IN NETWORK ANALYSIS

Several networks were considered in the earlier chapters to illustrate the

different network analysis techniques. However, these networks were sim-
ple and did not pose any problem. While applying the analysis techniques
to practical networks, however, one must guard against several pitfalls
described by Collins [190]. Some of these pitfalls are briefly described
here.

15.3.1 ERROR CRITERIA AND CONVERGENCE

Convergence problems are encountered in network analysis. Conver-

gence depends upon the initial values of the decision parameters and the
criteria used for the termination of the iterative procedure. Generally, the
iterative procedure is terminated when the absolute value of the difference
between the initially selected value and the value obtained at the end of the
iteration is less than some prespecifled value for all the parameters. As is
shown in Chapter 11, the network co-content is a minimum at the true solu-
tion. Therefore, even though the error criterion is satisfied and the iterative
procedure is terminated, sizable errors may occur, as observed by Collins,
in the co-content values of the solutions obtained from different starting
values. Therefore, satisfying the error criterion does not necessarily mean
that a good solution is reached.
 Reliability Analysis 419

15.3.2 FINDING UNIQUE SOLUTIONS

As stated in Chapter 5, certain pumps may have more than one flow for
the same head. Therefore, in practical networks having several pumps, the
possibility of numerous solutions cannot be ruled out.
For control elements such as check valves and pressure reducing valves,
the head loss characteristics are different depending upon their mode of
operation. The common approach for considering such elements in the
analysis is to assume a particular operating mode for each element, solve
the network, and check whether the assumed operating modes of different
elements are correct. If they are correct, the network is presumed to be
solved; if not, the operating modes are changed, the network is resolved,
and the process is terminated when the assumptions for the operating
modes are correct for all elements. Even though this procedure is straight-
forward, a problem arises in deciding how accurate the solution must be
before one uses the solution to test the validity of the assumption, regard-
ing the operating mode of the control elements. Naturally, if the solution
being tested is not sufficiently accurate, a wrong conclusion regarding the
validity of the operating modes may be drawn. It is thus seen that the crite-
ria used for terminating the iterative procedure could be critical.

15.4 RELIABILITY ANALYSIS

A water distribution system should be able to provide, during its entire

life, the required quality of water for the expected loading conditions with
the desired residual pressures at all nodes. The expected loading condi-
tions may be the maximum-day, maximum-hourly, average-day, average-
hourly, or maximum-instantaneous demands. In addition to these de-
mands, fire flow demands should also be considered. The system must also
be able to accommodate abnormal conditions such as breaks in pipes and
appurtenances; mechanical failure of pumps, valves, and control systems;
power outages; malfunctions of storage facilities; system contamination;
and inaccurate demand projections. The possibility of occurrence of each
of these events should be examined to determine the overall performance
and thereby the reliability of the distribution system.
Reliability is usually defined as the probability that a system performs its
mission within specified limits for a given period of time in a specified en-
vironment. For water distribution systems, however, at present time there
is no universally accepted definition or quantitative measure to express
their reliability, de Neufville et al. [191] were perhaps the first to introduce
426 MISCELLANEOUS TOPICS

the quantitative criterion for the overall performance of the New York City
water supply tunnel system. They expressed the overall performance as

Overall performance = - ^ (15.5)

in which pj = average pressure above the minimum standard at nodey,

and qj = demand at nodey. They used this criterion for evaluating the
overall performance of different alternatives and to provide guidelines for
the selection of the best alternative.
Considerable research work has been done recently on the reliability of
water distribution systems. Some of these are the works of Darnel in et al.
[192], Shamir and Howard [193,194], Hobbs [195,196], Mays [197], Cul-
linane [198], Tung [199], Mays and Cullinane [200], and Wagner et al.
[201-203]. Mays and Cullinane [200] and Wagner et al. [201] have pro-
vided a good review of the literature on various aspects of water supply
reliability.
Because the reliability concept as applied to water distribution systems
is still in a stage of development and because no universally acceptable
definition for the reliability of water distribution systems is available, a
brief review dealing with qualitative aspects of reliability is presented
herein, and quantitative approaches are not considered.
For a general study of reliability the system may be viewed as composed
of a single supply area connected to a single demand area. The available
methods provide reliability measures such as the expected percentage of
time during which the load will exceed the capacity and the expected num-
ber of shortfall events per unit time.
For a somewhat more detailed analysis, the system's supply is modeled
with more components while the demand is considered as a lumped one.
The probability functions of time to failure and time to repair are used in
evaluating the reliability of each component. Fault-tree analysis that con-
siders the different ways in which component failures lead to supply short-
fall and computes the associated probabilities can also be used.
For a more detailed analysis, the demand is no longer a lumped one. For
a preliminary study, the concept of reachability of a specified node when
the node is connected to at least one source and the concept of connectivity
in which every demand node is connected to at least one source can be
used. However, connectivity of a demand node to a source does not neces-
sarily mean the supply of water if sufficient pressure is not developed at
the node. However, the concepts of reachability and connectivity help in
the preliminary analysis of the system, in identifying zones with serious
 Unsteady Flow Analysis 421

problems due to insufficient redundancy, and in demarcating nodes with

serious problems of shortages. Thus, these concepts provide a fast, easy-
to-perform check in evaluating unreliability due to inadequate network
connections or extremely unreliable links.
A detailed reliability analysis should consider the probabilities of failure
of different elements such as pipes and pumps, and the actual performance
of the system during these events. Analytical and simulation methods have
been developed for such an analysis [201-203]. However, these methods
involve volumes of computational work and therefore are limited to small
networks.

15.5 UNSTEADY FLOW ANALYSIS [154]

The type of flow in water distribution systems, so far considered in the

text (except that in Chapter II), has been steady. However, unsteady flow
may develop in the distribution systems due to several causes such as the
fluctuations in demand, changes in reservoir water levels, starting and
stopping of pumps, and changes in valve settings. In the dynamic analysis
methods presented earlier in Chapter 12, it was presumed that the changes
in nodal demands would instantaneously cause the corresponding changes
in the reservoir outflows and pump discharges. However, in large net-
works, the changes would not be instantaneous; the unsteadiness caused by
the changes may persist for quite some time, and the establishment of
steady flow may be attained thereafter. Of the several methods which are
available for dealing with unsteady flow in pipes and their networks [154],
the characteristics method is briefly described herein.
The unsteady flow in a pipe is described by the partial differential equa-
tions [Equations (11.3) and (11.4)]. These equations contain the distance
along the pipe x and time t as independent variables and velocity V and
head H as dependent variables. These equations are transformed by the
characteristics method into the following four ordinary differential equa-
tions:
id// dV fV | V
= 0 (15.6)
c dt dt ID
c+
dx
(15.7)

fV\V
= 0 (15.8)
c dt dt ID

(15.9)
422 MISCELLANEOUS TOPICS

As V = V{x,t), the C* and C characteristics described by Equations

(15.7) and (15.9) are shown on the *r-plane in Figure 15.1.
Multiplying Equations (15.6) and (15.8) by (c/g) dt and Equations (15.7)
and (15.9) by dt, replacing Kby Q/A in Equations (15.6) and (15.8) (A being
the cross-sectional area of the pipeline), and using the first order approx-
imation in the integration, the finite difference forms of the equations (Fig-
ure 15.1) are

Hp {QP
~ HR ~ QR) QR
I QR - tR) = 0
(15.10)

xp - xR = (VR + cR)(tp - tR) (15.11)

Hs = Qs) Qs
~ ^ ~ ~
(15.12)

Xp - XS = (Vs ~ Cs)(tp - ts) (15.13)

These four equations containing four unknowns xp, tp, Hp, and Qp can be
solved by the method of specified time intervals [154]. The procedure is
continued for successive time intervals. Reservoirs, pumps, and valves can
also be included in unsteady flow analysis.

15.6 COMPUTER APPLICATION

The emphasis throughout the text has been on understanding the funda-
mentals of the different network analysis methods. Therefore, the networks
considered for the illustrative examples were small so that they could be
tackled by hand calculation. However, as the analytical methods require

t
^ A x * -Ax*

y
t+At

R
Figure 15.1 C* and C~ characteristics on jtf-plane.
 Problems 423

either a large number of simple iterations (Hardy Cross method) or simul-

taneous solution of a large number of linear and/or linearized equations
(Newton-Raphson, and linear theory methods), the node head analysis of
even small networks involves considerable computational work. When
some control elements such as check valves and pressure reducing valves
are present in a network, several NHA solutions may be necessary before
results consistent with the control-element settings are obtained. Repeated
NHA solutions are necessary in dynamic analysis when the iterative pro-
cedure is used and also in the node flow analysis, and therefore, the com-
putational work further increases. No wonder, therefore, that the electrical
analyzers flourished before the advent of digital computers.
As the computers developed and became readily available, their applica-
tion to obtain network analysis solutions has been common. Computer
programs based on the BASIC and FORTRAN languages can be easily de-
veloped for simple networks. They can also be found in books of Streeter
and Wylie [9], Jeppson [14], Wylie and Streeter [154], and Watters [204].
Borgohain et al. [104] developed computer programs for dynamic analysis
of networks by the iterative and direct methods. Computer programs for
unsteady-flow analysis can be obtained in Wylie and Streeter [154] and
Watters [204]. Coulbeck [205] provided a computer program for the cali-
bration of networks.
Computer programs are also commercially available to deal with practi-
cal networks involving a large number of nodes and pipes and also several
reservoirs, pumps, and control elements [206,207].

15.7 PROBLEMS

Problem 15.1: In Example 15.1, using sensitivity analysis, find the max-
imum possible outflow at node 6 without affecting the outflows at all other
nodes.
 Nomenclature

A = area; cross-sectional area of pipe, reservoir;

constant in head-discharge relationship of
pumps; convergence acceleration factor; up-
stream node in two-demand-node serial sys-
tem; adjustment factor for nodal flow in cali-
bration
Ac = area of flow at vena-contracta
Aj = street junction for node A
Ai, Ai = cross-sectional areas at sections 1-1, and 2-2,
respectively; convergence acceleration factors
ADQF = adequate flow node
a = constant in head-discharge relationship of
pumps; node connected to node j through pipe
aj
ai, aj, ak, am = adjacent loop connected through pipes i, j , k,
and m, respectively
atJ = element in row i and column j of coefficient
matrix
aj = pipe connecting nodes a and j
B = constant in head-discharge relationship of
pumps; set of boundary elements; downstream
node in two-demand-node serial system; ad-
justment fector for pipe coefficient in calibra-
tion
Bj = street junction for node B
b = constant in head-discharge relationship of
pumps; node connected to node j through pipe
bj; boundary element
C = constant in head-discharge relationship of
pumps; number of circuits (loops) in network;
content of a network model
Cc = coefficient of contraction
425
426 Nomenclature

CHW =Hazen-Williams coefficient

CHW, =Hazen-Williams coefficient of equivalent pipe
CHW, =Hazen-Williams coefficient of pipe JC
CR =coefficient of roughness in modified Hazen-
Williams formula
C , = adjusted Hazen-Williams coefficient of pipe x
during calibration
Cxp = predicted Hazen-Williams coefficient of pipe x
during calibration
C, = SLQ2/w2ghf
C2 = AQI-KV
C(t, t + At) = correction factor for time interval (/, t + At)
C*, C~ = names of characteristic equations in unsteady
flow analysis
C» = modified conductance of pipe x
Cx',o = modified conductance of pipe x obtained from
fh iteration
,C'X = modified conductance of pipe x assumed in f
iteration
CC = co-content of a network model
CI = cast iron
CRIT = critical node
CV = check valve
c = circuit (loop) in network; wave speed
D = diameter of pipe; constant in head-discharge
relationship of pumps
D, = diameter of equivalent pipe
Dx = diameter of pipe x
D(t, t + At) = total network demand in time interval (t,
t + At)
DW = Darcy-Weisbach
d = downstream end of pressure reducing valve
dp = change in pressure
dV = change in volume
dv = change in velocity
dx = change in ^-direction
dy = change in y-direction
E = energy; constant in head-discharge relationship
of pumps
Ec = corrected error for water volume balance for
entire network
Ee(t, t + At) = corrected error in water volume balance for the
entire network during time interval (/, t + At)
E, = externally supplied or extracted energy
EL = energy loss
Ep = energy supplied by pump; predicted error in
water volume balance for the entire network
 Nomenclature 427

Ep(t, t + At) = predicted error in water volume balance for the

entire network during time interval (/, t + At)
Er = energy extracted by turbine
Ex = loss of energy in pipe x
EGL = energy grade line
E.S.R. = elevated service reservoir
e = size of roughness projections; equivalent pipe
e' = spacing of roughness projections
e« = corrected error in volume of water for reser-
voir r
en(t, t + At) = corrected error in volume of water for reservoir
r for time interval (t, t + At)
erp = predicted error in volume of water for reser-
voir r
erp{t, t + At) = predicted error in volume of water for reservoir
r for time interval (t, t + At)
F = force; flow at hydrant during fire flow test
Fal = factor representing the influence of adjacent
node i
/"„ = factor representing the influence of adjacent
loop connected through pipe x
Fb = function, expressing head in terms of flow, for
boundary element b
Fj = residue of inflow over outflow at node j
Fx, Fy, F, = force components in x, y, and z directions, re-
spectively
F ( ) = function of ( )
Fi, F2, . . . , Fn = value of function 1, 2, . . . , n
,Fy = residue of inflow over outflow at node j for r*
iteration
Fi, ,Fi, . . . , ,Fn = value of function 1, 2, . . . , n for f* iteration
F',F", . . . ,F" = first, second, . . . , « ' * derivative of F
f = Darcy-Weisbach friction coefficient
fb = function, expressing flow in terms of head for
boundary element b
f. = Darcy-Weisbach coefficient for equivalent pipe
fr = function for reservoir r
fx = Darcy-Weisbach friction coefficient for pipe x
/ ( ) = function of ( )
fr(Hr) = first derivative representing cross-sectional
area of reservoir r
fr[Hr(t)] = first derivative representing cross-sectional
area of reservoir r at time t
G = given network; ground node
GP = QP + (BI2A)
G' = connected loopless network for given net-
work G
g = gravitational acceleration
428 Nomenclature

H - head; hydraulic gradient level (HGL)

HA, HB = levels of outlets A and B, respectively
HAi, HHi = levels of street junctions at nodes A and B, re-
spectively
HA,, HB, = heads for design demands at nodes A and B, re-
spectively
Hb = head at boundary element b
Hb{t) — head at boundary element at time t
Hd = head at downstream end d of pressure reducing
valve
Hiy H}, Hj = head at nodes i, J, andy, respectively
ff mm — m i n i m u m permissible head at node j
,Hj - assumed value of HGL at node,/ for fh iteration
HP = head at solution point P in unsteady flow anal-
ysis
Hp = shut-off head for pump
Hr = water level in reservoir r
Hr(t) — water level in reservoir r at time /
Hr{t + At) = water level in reservoir r at time t + At
H,c(t + At) = corrected water level in reservoir r at time
t + At
Hs, Hs = available head at source node S, s, respectively
HK, = preset head in pressure reducing valve
Hsf — head at source node s during fire flow test
H,fp = predicted head at test node t during high (fire)
flow
H,p = predicted head at test node t during low flow
H,,p, H,1P = predicted heads at test nodes r,, and t2, respec-
tively
Hu = head at upstream node u of pressure reducing
valve
Ho = constant in head-discharge relationship of
pumps
HoA, HoB — known HGL values at nodes A, and B, respec-
tively
Hoc = known HGL at ground node G
HOJ = known HGL at node j
Hom = known or specified HGL at source node m
HG = hydraulic gradient
HGL = hydraulic gradient level
HW = Hazen-Williams
hb = head loss in bends and elbows
he = externally supplied or extracted head
hen = head loss at entrance
hex = head loss at exit
hf = head loss due to friction
hu = head loss due to friction in equivalent pipe
 Nomenclature 429

h,m = head loss in flow meter

hh = head loss due to friction in pipe x
hgc = head loss due to gradual contraction
htt = head loss due to gradual enlargement
hL = head loss
hm = minor head loss
h0 = head loss due to obstruction
hp = head supplied by pump
,hp = head supplied by pump in fh iteration
h, = static head
hx = head loss due to sudden contraction
hs. = head loss due to sudden enlargement
hT = system head; head supplied to turbine
h, = head loss in tee
hv = head loss in valve
ht = head loss in pipe x
I = current in resistor of electrical network
i = node; pipe
if = pipe
J = junction point; total number of nodes (source
as well as demand) in network
JUH = number of nodes with unknown head
JUq = number of nodes with unknown flow
j = node; pipe
K = bulk modulus of elasticity; Karman number
( = Re%//); minor head loss coefficient; stiff-
ness of structural member; constant for conver-
sion of energy to Joules or kilowatt hours
Kb = coefficient for head loss in bends and elbows
Ktn = coefficient for head loss in entrance
Kex = coefficient for head loss in exit
Kfm = coefficient for head loss in flow meter
Kgc = coefficient for head loss due to gradual con-
traction
Kg, = coefficient for head loss due to gradual en-
largement
Ko = coefficient for head loss due to obstruction
,KpRv = coefficient for head loss in pressure reducing
valve
KK = coefficient for head loss due to sudden con-
traction
KM = coefficient for head loss due to sudden en-
largement
K, = coefficient for head loss in tees
Kv = coefficient for head loss in valves
k = node; pipe
L = length of pipe
430 Nomenclature

Le =
length of equivalent pipe
Lx —
length of pipe x
In =
natural logarithm
log =
logarithm to base 10
M =
mass flow rate, number of source nodes in net-
work
m = dimensionless form factor for pipe roughness;
node; pipe
N = Manning roughness coefficient; number of de-
mand nodes in network
N. = Manning roughness coefficient of equivalent
pipe
Nx = Manning roughness coefficient of pipe x
Nz = set of demand nodes in zone z
NFA = node flow analysis
NHA = node head analysis
NOFL = no-flow node
n = exponent of discharge in general pipe head loss
relationship
P = wetted perimeter; power delivered by pump;
solution point in jtf-plane of characteristic grid
in unsteady flow analysis
PRTF = partial-flow node
PRV = pressure reducing valve
p = pressure
PJ = average pressure above the minimum standard
at node j
Q = volumetric flow rate; pipe discharge
Qd = discharge at downstream end of pipe
Qe = discharge in equivalent pipe
Qij = discharge in pipe ij
Qp — discharge supplied by pump
Qu = discharge at upstream end of pipe
Qx = discharge in pipe at distance x; discharge in
pipe x
Qxp = predicted discharge in pipe x
QxU) = discharge in pipe x obtained after f* iteration
,QX = assumed discharge in pipe x for /'* iteration
Qo = constant discharge in head-discharge relation-
ship of pumps
Qox = known value of discharge in pipe x
QA, <IB = flow at upstream node A and downstream node
B, respectively
JS <7«es = design demands at nodes A and B, respectively
™\ ^B"" = maximum possible flows at nodes A and B, re-
spectively
qb(t) = flow at boundary element at time /
 Nomenclature 431

qb(t + At) = flow at boundary element at time t + At

qfr = rate of filling for reservoir r
qj = flow (supply or demand) at node j
qia = adjusted flow at nodey
<ljp = predicted flow- at node j
q'P = required flow at node j
qj(t) = demand for node j at time t
qj(t + At) = demand for node j at time t + At
,qj — assumed value of flow at node j for t'H iteration
q, = flow at test node t
qoj = known value of flow at node j
R = hydraulic radius ( = A/P); radius of bend; pipe
resistance constant
Re = Reynolds number
R, = resistance of equivalent pipe
RPRV = resistance of pressure reducing valve
Rp = resistance of pump
/?„ = resistance of pipe with unknown resistance
Rx = resistance of pipe x
Rxa = adjusted resistance of pipe x
Rxp — predicted resistance of pipe x
Rx' = modified resistance of pipe x
Rox = known resistance of pipe x
,RX = assumed resistance of pipe x for t'h iteration
,RX = modified resistance of pipe x for /•'• iteration
r = radius of pipe, reservoir
rx = correction to ARX
,rx = correction to ARX for t'h iteration
S = slope of hydraulic grade line; sum of demands
at nodes significantly affecting hydrant test;
source node
SE = expenditure of system energy
SUBC = subcritical node
SUPC = supercritical node
SURF = surplus-flow node
t — time
t, tu t2 = test nodes
(f) = at time t
(/, + At) = at time t + At
(f, t + At) = time interval between / and i + At
,( ) = f iteration
U = total number of pipes with unknown resistances
u = upstream end of pressure reducing valve; pipe
with unknown resistance; displacement of
structural member
V = volume; average velocity; voltage drop along
432 Nomenclature

resistor; set of free variables in sensitivity

analysis
V, = volume of water in reservoir r
Vr{t) = volume of water in reservoir r at time t
Vr{t + At) = volume of water in reservoir r at time t + At
Vrp(t + At) = predicted volume of water in reservoir r at time
t + At
Vo = volume at standard atmospheric pressure
K,, F2 = average velocities at sections 1-1, and 2-2, re-
spectively
v = velocity
vf = free variable in sensitivity analysis
w = half band width
W.L. = water level
X = number of pipes in network
XUR = number of pipes with unknown resistance
x = distance along pipe; pipe
y = distance in ^direction
yk = parameter that changes in sensitivity analysis
Z = datum head
ZERF = zero-flow node
Zc = correction to AQC
,Zc = correction to AQC for f * iteration
0 (zero) = source node label in single-source network
1 , 2 , . . . = labels for nodes, pipes
I, II, . . . = labels for loops
a = annual growth rate of pipe roughness
7 = specific weight of fluid
AFU AF2 . . = demand adjustments for zones 1, 2, . . .
AFZ — total nodal flow adjustment for zone z during
calibration
AHam, AHbm = modified correction to HGL at nodes a, and b,
respectively
AHj = correction to HGL at nodey
AHjm = modified correction to HGL at node./
AHj_LJ = correction to HGL at nodey according to linear
theory method
A// ; _ NR = correction to HGL at node j according to
Newton-Raphson method
,AHj = correction to HGL at node j for t'h iteration
AHkm, AHmm = modified correction to HGL at nodes k, and m,
respectively
AL = equivalent pipe length for minor head loss
element
AQaim, AQaJm, AQakm, AQamm = modified flow correction for adjacent loops ai,
aj, ak, and am, respectively
 Nomenclature 433

tAQax = flow correction for adjacent loop connected

through pipe x for fh iteration
AQC = flow correction for loop c
AQcm — modified flow correction for loop c
,AQC = flow correction for loop c for f * iteration
AQu — flow correction according to linear theory
method
AQm = flow correction according to Newton-Raphson
method
AQP = flow correction for pseudo loop
,AQP = flow correction for pseudo loop for f * iteration
AQ, = adjustment to discharge in pipe x during cali-
bration
ARX = correction to resistance constant of pipe x
,ARX = correction to resistance constant of pipe x for f
iteration
At = time interval
AVb = volume supply of water by boundary element b
AVb(t, t + At) = volume supply of water by boundary element b
in time interval (t, t + At)
AV, = change in volume of water (outflow volume) at
reservoir r
AVr(t, t + At) = change in volume of water (outflow volume) at
reservoir r in time interval (t, t + At)
5y = thickness of ring at distance y
17 = efficiency
0 = angle of bend
p. = coefficient of viscosity
v = kinematic viscosity
Q = mass density
T = shear stress
0 ( ) = function of ( )
 Answers

Problem 2.1: 8,414, 0.2257 m; Using Blasius" formula, 15.87 m.

Problem 2.2: DW formula: 110.3 m; Manning formula (N = 0.013):
185.9 m; HW formula (CHW = 126.7 by interpolation from Table 2.4):
97.11 m.
Problem 2.3: 0.1429 nvVs; 7.028% decrease.
Problem 2.4: DW formula: 266 mm; HW formula: 263 mm.
Problem 2.5: 5.66 m, 5.58 m, 1.4% decrease; 17.81 m, 17.74 m, 0.4% de-
crease; 0.339 m, 0.255 m, 25% decrease.
Problem 2.6: DW formula: 246 m, 330 m, 391 m, 456 m. HW formula:
151 m (CHW = 99.7), 272 m (CHW = 72.7 m), 319 m (CHW = 66.7), 379 m
(CHW = 60.7).
Problem 4.1: DW formula: 41.31 m, 13,600 m, 177.6 mm; HW formula:
36.22 m, 11,940 m, 178.5 mm; Manning formula: 70.75 m, 14,890 m,
177.3 mm.
Problem 4.2: DW formula: 0.0462 m3/s, 13,426 m, 177.6 mm; HW for-
mula: 0.0484 m 3 /s, 11.940 m, 178.5 mm; Manning formula: 0.0354 nWs,
14,890 m, 177.3 mm.
Problem 4.3: DW formula: 0.309 nrVs, 239.6 m, 331 mm; HW formula:
0.332 m3/s, 253.1 m, 329.6 mm.
Problem 4.4: DW formula: 0.06799 m3/s, 0.02591 m7s, 0.08610 mVs, 5.14
m; HW formula: 0.06859 m3/s, 0.02551 nvVs, 0.08590 m3/s, 4.83 m.
Problem 4.5: 285.9 mm, 114.6.
Problem 4.6: 263.3 mm, 92.25.
Problem 4.7: DW formula: 47.40 m, 173.0 mm; HW formula: 42.31 m,
173.0 mm.
Problem 5.1: (1) 0.2794 nWs; [2(a)] £>,., = 0.3706 rrvVs, Q2.j = 0.1883
m 3 /s. Qj..3 = 0.5589 mVs, H} = 86.26! m, // 3 = -5.973 m; [2(b)]
435
436 Answers

Qx_, = 0.2432 m3/s. g,- 2 = 0.1035 m 3 /s, Qj.3 = 0.1397 m 3 /s, ft =

104.54 m, ft = 97.46 m.
Problem 5.2: 0,_ y = 0.28352 rrrYs, &- 2 = 0.05603 m 3 /s, Q,-3 = 0.22749
m 3 /s; 104.456 m.
Problem 5.3: 0 W = 0.25602 m Vs, Qj-i = 0.04156 m 3 /s, Q,-3 = 0.09624
rrvVs, Qj-4 = 0.11821 nrVs: 121.498 m.
Problem 5.4: (?,_,, - 0.23000 m3/s, 07,-2 = 0.06078 m 3 /s, Qj,-J2 =
0.16922 mVs, Qj2.3 = 0.06817 m3/s, Q,^ = 0.10102 m 3 /s; ft, =
123.029 m, ft, = 116.070 m.
Problem 5.5: 0.1195 m3/s.
Problem 5.6: 24.0 m, 0.107 m3/s.
Problem 5.7: 0 , . , = 0.0880 m3/s, Q2-j = 0.0974 m 3 /s, Q,-3 = 0.1854
nrVs; ft = 98.95 m.
Problem 5.8: 0 , . , = 0.1356 mVs, 0,- 2 = 0.0983 m 3 /s, Q3-j = 0.0974
m3/s, Qj.A = 0.1347 m3/s, ft = 127.38 m.
Problem 5.9: (?,_.,, = 0.1375 m3/s, 0 y ,- 2 = 0.0978 m 3 /s, Q,,.^ = 0.0396
m3/s, ft,.., = 0.1183 m3/s, Q^j, = 0.0786 m 3 /s; //,, = 127.31 m, ft, =
126.84 m.
Problem 5.10: (1) 0.2794 m 3 /s; [2(a)] 1 (check valve permits flow from
J to 2): Q^j = 0.5589 m3/s, Q2 = 0, Q,.3 = 0.5589 m 3 /s, ft = 47.80 m,
H3 = —44.43 m; 2 (check valve permits flow from 2 to J): Qt-j =
0.3764 m 3 /s, Q2.j = 0.1825 m 3 /s, G7-3 = 0.5589 m3/s, ft = 85.277 m,
ft = - 6 . 9 5 8 m; [2(b)] 1 (check valve permits flow from J to 2):
<2 W = 0.2405 m°7s, Qj-2 = 0.1008 m 3 /s, Qj-3 = 0.1397 m 3 /s, ft =
104.852 m, ft = 100.986 m; 2 (check valve permits flow from 2 to J):
Qt-j = 0.1397 m3/s, Q2 = 0, <2,-3 = 0.1397 m 3 /s, ft = 114.461 m, ft =
107.385 m.
Problem 5.11: (1) Permissible flow direction is from J to 2: Q w = 0.2601
m3/s, Qj-2 = 0.0375 mVs, <2,-3 = 0.2226 m 3 /s, ft = 103.76 m. (2) Per-
missible flow direction is from 2 to J: Q^j = 0.2416 nrVs, Q2 — 0,
Qj-3 = 0.2416 m3/s, ft = 106.51 m.
Problem 5.12: (1) g , = 0.3126 m 3 /s, Q2 = 0.0874 m3/s, Q3 = 0.4
m 3 /s; //„, = 101.54 m, ft, = 96.16 m, ft2 = 98.67 m, ft, = 92.00 m,
ft = 90.00 m, ft = 40.35 m. (2) Qx = 0.0921 m3/s, Q2 = 0.2079 m3/s,
Q3 = 0.3 m 3 /s; Hu, = 118.08 m. ft, = 75.00 m, ft2 = 93.40 m, ft2 =
84.25 m, ft = 74.35 m, ft = 45.21 m. (3) Q, = 0.2 m 3 /s, Q2 = 0,
& = 0.2 m 3 /s; //„, = 111.93 m, ft, = 100.00 m, HU1 = 100.00 m,
ft2 = 97.31 m, ft = 97.31 m, ft = 83.56 m. (4) Q, = 0.1981 nrVs,
Q2 = 0.1019 m3/s, & = 0.3 m3/s, ft, = 112.07 m, ft, = 80.00 m,
ft2 = 98.24 m, Hdl = 80 m, ft = 77.36 m, ft = 48.22 m.
P r o b l e m 6.1: (1) . . . (4) N o : (5) Yes; (6) Yes.
 Answers 437

P r o b l e m 6 . 2 : (1) N o ; (2) Yes; (3) Yes; (4) N o ; (5) Yes.

P r o b l e m 6 . 3 : (1) N o ; ( 2 ) . . . ( 4 ) Yes; ( 5 ) . . . (7) N o ; ( 8 ) Yes;
(9) Yes.
Problem 8.1: (?,-2 = 00916 m 3 /s, Qt.4 = 0.1484 m 3 /s, Q4.2 = 0.0052
m3/s, 62-3 = 0.0168 m 3 /s, Q4.s = 0.0832 m 3 /s, 65-3 = 0.0432 m 3 /s, q> =
0.2400 nrVs (supply); ft = 95.384 m, ft = 63.062 m, ft = 95.452 m,
ft = 82.436 m.
Problem 8.2: <2i-3 = 7.6679 nWmin, g 2 . 5 = 5.6596 m 3 /min, Q3.s =
1.1473 m 3 /min, Q3.4 = 4.5212 m 3 /min, QS6 = 2.8064 m3/min, Q4.6 =
1.1198 m 3 /min, Q4.2 = 0.4004 m 3 /min, Q2.6 = 1.0710 m 3 /min, qx =
13.3275 m3/min (supply), q2 = 0.6706 m3/min (supply); H3 = 104.120 m,
H4 = 100.032 m, ft = 103.594 m, ft = 99.655 m.
Problem 8.3: g,_ 2 = 0.09335 m 3 /s, Qt.4 = 0.14665 m3/s, Q4.2 = 0.01235
m3/s, g 2 - 3 = 0.02570 m3/s, Q4.5 = 0.07430 m 3 /s, Q5.3 = 0.03430 m 3 /s,
ql - 0.2400 m3/s (supply); H2 = 95.217 m, H3 = 72.367 m, // 4 =
95.551 m, R4 = 20,120.
Problem SA: Qy.3 = 5.4772 m 3 /min, Qt.s = 7.7809 m 3 /min, Qs.3 =
1.4609 m 3 /min, Q3.4 = 4.9381 m 3 /min, 65-6 = 2.3200 m 3 /min, Q64 =
1.3200 m 3 /min, Q4.2 = 3.2581 m 3 /min, Q2.6 = 4.000 m3/min; qt =
13.2581 m3/min (supply), q2 = 0.7419 m3/min (supply); H4 = 102.123 m,
ft = 97.892 m; R3 = - 4 . 2 6 8 , R6 = -3.973 (both resistances repre-
sent pumps).
Problem 8.5: Q^3 = 4.0966 m 3 /min, <2,-5 = 3.3488 mVmin, Q3.s =
1.1885 m 3 /min, Q3.4 = 0.9072 m 3 /min, Q5.6 = 0.5373 m7min, Q4.6 =
1.3475 m 3 /min, Q2.4 = 3.4401 mVmin, Q26 = 3.1153 nWmin, Q7.2 =
6.5554 nWmin; q, = 7.4454 rrrVmin (supply), q7 = 6.5554 m3/min
(supply); H2 = 110.524 m, H3 = 108.322 m, H4 = 108.157 m, ft =
107.757 m, ft = 107.613 m, hp = 24.822 m.
Problem 8^: Qt.3 = 3.1623 m 3 /min, Q,.s = 4.0992 nvVmin, Q3.5 =
2.4292 m 3 /min, Q4.3 = 1.2668 mVmin, e 5 . 6 = 2.5251 m 3 /min, Q6.4 =
2.2504 nWmin, Q2.4 = 2.0165 nWmin, Q2_6 = 4.7224 m3/min, Q7.2 =
6.7387 m3/min; ^j = 7.2615 m3/min (supply), qn = 6.7385 m3/min (sup-
ply); H2 = 110.134 m, ft = 109.321 m, ft = 106.639 m, ft =
103.444, hp = 24.676 m, # 6 = -1.1605 (pump).
Problem 8.7: (1) Permissible flow direction from node 3 to node 5.
Qi-3 = 7.6431 nvVmin, <2,.5 = 5.6856 m 3 /min, Q3.s = 1.1021 irrVmin,
63-4 = 4.5413 m7min, QS6 = 2.7876 m3/min, Q4.6 = 1.1315 m 3 /min.
64-2 = 0.4098 nvVmin, Q2.6 = 1.0809 rrrVmin; qv = 13.3287 irrVmin
(supply), q2 = 0.6711 m3/min (supply); ft = 104.158 m, ft =
100.034 m, ft = 103.535 m, ft = 99.649 m. (2) Permissible flow di-
rection from node 5 to node 3. Qt.3 = 7.0078 nWmin, Q^5 = 6.2862 m 3 /
, <2s-3 = 0, £?3-4 = 5.0078 m 3 /min, Qs_6 = 2.2861 m 3 /min, Q46 =
438 Answers

1.4020 nWmin, £4-2 = 0.6056 m3/min, Q2_b = 1.3119 rrvVmin; qx =

13.2940 m3/min (supply), q2 = 0.7063 m3/min (supply); H3 = 105.089
m, H4 = 100.073 m, Hs = 102.097 m, H6 = 99.484 m.
Problem 8 3 : Q,.3 = 0.7548 m3/s, g 3 . 4 = 0.3212 m3/s, 03_6 = 0.2336
m 3 /s, Qs_4 = 0.0788 nrVs, Qs.6 = 0.0664 mVs, Q2.6 = 0.2452 m 3 /s;
<?, = 0.7548 m3/s (supply), q2 = 0.2452 nrVs (supply); H3 = 93.121 m,
H4 = 89.458 m, Hs = 90.000 m, Ht = 89.738 m, Hu = 97.039 m, //„ =
90.000 m (set pressure head).
Problems 9.1 . . . 9&: Same as Problems 8.1 . . . 8.8, respectively.
Problems 10.1 . . . 10.8: Same as Problems 8.1 . . . 8.8, respectively.
Problem 11.5: £>, = 0.283514 m3/s, Q2 = 0.056032 m3/s, Q3 = 0.227482
m 3 /s, C = -18.29226 m V - m .
Problem 11.6: C = -812.6463 m V m .
Problem 11.7: tf, = 104.4588 m, CC = 9.87703 m V ' m .
Problem US: CC = 149.594 m V - m .
Problem 12.1: As given in Table: Answer: Problem 12.1.
Problem 12.3: /,(4) = 0.01845, /?,(4) = 188.2, Qt(4) = 0.1848 m 3 /s;
/ 2 (4) = 0.02145, R2(4) = 2,216, Q2(4) = 0.0689 m 3 /s; / 3 (4) = 0.02315,
R3(4) = 10,075, <23(4) = 0.0358 m 3 /s; / 4 (4) = 0.02344, /?4(4) = 3,825,
ft(4) = 0.0252 m 3 /s; / s (4) = 0.02120, /?5(4) = 820.9, QS(A) = 0.1163
m 3 /s; /,(4) = 0.02390, rt«(4) = 3,901. Q6(4) = 0.0589 m 3 /s; / 7 (4) =
0.01776, R7(4) = 139.7, fi7(4) = 0.2752 m 3 /s; H1{4) = 95.955 m;
hp(4) = 11.196 m; Volume of water supplied by sump at node 1 = 1,320
m3; Volume of water supplied by reservoir at node 2 = 1,860 m3.
Problem 12.4: /,(4) = 0.01845, /?,(4) = 188.2, Q,(4) = 0.1850 m 3 /s;
/ 2 (4) = 0.02145, R2(4) = 2,216, Q2(4) = 0.0691 m 3 /s; / 3 (4) = 0.02314,
R3(4) = 10,073, Q3(4) = 0.0360 m 3 /s; / 4 (4) = 0.02344, /?4(4) = 3,825,
g 4 (4) = 0.0252 m 3 /s; / 5 (4) = 0.02120, fls(4) = 820.9, Q5(4) = 0.1161
m^/s; / 6 (4) = 0.02390, fl6(4) = 3.901, Q6(4) = 0.0588 m 3 /s; / 7 (4) =
0.01777, j?7(4) = 139.7, jg,(4) = 0.2750 m 3 /s; H2{4) = 95.809 m;
hp{4) = 11.194 m; Volume of water supplied by sump at node 1 = 2,629
m3; Volume of water supplied by reservoir at node 2 = 4,191 m3.
Problem 12.5: As given in Table: Answer: Problem 12.5.
Problem 13.1: (1) qTx = 2.1705 nrVmin; (2) qW" = 1.3114 nfVmin.
Problem 13.2: (a) HA < 93.60 m; (b) HB < 89.36 m.
Problem 13.3: HB, = 91.00 m, HBl = 94.00 m, HB, = 91.40 m, HB. =
89.36 m, HBs = 69.00 m, HBt = 58.67 m.
Problem 134: (a) 0.5297 m 3 /s; (b) 573.5.
Problem 13.5: q, = 3.000 irrVmin, q2 = 1.907 rrrVmin, q3 --=• 0.850
m 3 /min, qA = 3.472 m3/min; H, = 96.95 m, H2 = 94.00 m, H3 =
91.00 m, HA = 88.00 m.
 Answers 439

Problem 13.6: qt = 2.000 nvVmin, q2 = 4.000 nrVmin, q3 = 1.043 m3/

min, <?4 = 1-917 nWmin; Hx = 97.111 m, H2 = 93.490 m, H3 = 92.000
m. H4 = 90.000 m.
Problem 13.7: (a) q, = 2.3342 nWmin (supply). q2 = 0.7000 m 3 /min,
q3 = 0.8000 nrVmin, qA = 0, qs = 0,q6 = 0.8342 m 3 /min, q7 = 0;
H2 = 87.360 m, // 3 = 85.531 m, H4 = 86.021 m, Hs = 86.067 m, H6 =
85.00 m, H7 = 85.691 m. (b) qx = 2.6606 m3/min (supply), q2 = 0.6565
m3/min, ^ 3 = 0.3467 nWmin, ^ 4 = 0.4000 nrVmin, ^ 5 = 0.4416 m3/min,
q6 = 0.8158 m3/min, q7 = 0 ; H2 = 87.000 in, // 3 = 85.000 m, // 4 =
94.337 m. « s = 88.000 m, H* = 85.500 m, H7 - 87.500 m.
Problem 15.1: 1.0111 m3/min.

TABLE: ANSWER: PROBLEM 12.1.

Dynamic Analysis for Time Interval 0-4 hr for case

Item (a) (b) (c)

Qi, m3/min 1.7500 0.0932 0.8998
Qi, m3/min 2.9193 1.4456 2.1508
Qi, nrVmin -0.2500 -1.9068 -1.1002
QA, m3/min 0.8344 1.9441 1.4337
Qi, mVmin 2.1536 1.7897 1.9845
Q6, m3/min 2.4465 2.8104 2.6157
Q7, m3/min 0.8465 1.2104 1.0156
qi, m3/min -4.6693 -1.5388 -3.0506
q2, m3/min -3.5309 -6.6613 -5.1496
Vu m3 3,876 1,878 4 ,084
V2, m3 4,036 5,554 5 ,748
Hum 102.326 101.127 102.450
H2, m 101.229 102.443 102.599
H3,m 101.198 101.122 102.121
H4, m 100.871 100.731 101.624
ft, m 99.215 99.555 100.201
ft, m 99.656 100.410 100.818
 TABLE: ANSWER: PROBLEM 12.5.

Static Analysis at Time

Item Ohr 4hr 8hr 12 hr 16 hr 20 hr 24 hr
Q,. m3/min -0.3853 2 0820 3.0504 2.7429 2.4166 1.1413 - 0 5559
£>2, mVmin -0.1996 3.2193 4.4563 3.9758 3.6881 2.0039 -0.4765
(2.i, nv'-'inin -1.3853 0.0820 0.0504 -0.2572 0.4165 -0.8587 -1.5559
Q<,, m3/min 1.2799 0.5830 0.2049 0.1649 0.4492 1.0164 1.4469
Qi. m'/min 0.2804 2.2023 2.6612 1.5407 2.9373 1.4203 0.1703
Qr., m'/min 1.3196 2.3977 2.3389 1.6593 3.0627 1.9797 1.4297
Q7, m 3 /min 0.7196 0.7977 1.3389 0.4593 1.0627 0.5797 0.8297
qx , m 3 /min 0.5849 -5.3013 -7.5066 -6.7186 -6.1046 -3.1452 1.0324
q-i , mVmin -3.9849 -2.8987 -2.4934 -2.0814 -3.0954 -3.8548 -4.4325
V, , m3 2,000 4,755 6,527 4,784 3,228 2,091 1,828
V1,m3 5,000 4,117 3,449 2,888 2,260 3,805 5,172
//,, m 101.200 102.853 103.916 102.870 101.936 101.254 101.096
H2, m 102.000 101.294 100.759 100.310 99.808 101.044 102.138
H3,m 101.268 101.297 100.761 100.278 99.887 100.742 101.231
H4,m 101.210 101.110 100.733 100.293 99.694 100.529 101.147
//5,m 101.172 99.383 98.282 99.402 96.752 99.763 101.132
// 6 , m 101.499 99.778 99.312 99.544 97.423 99.982 101.556
 References

1 Bhave, P. R. and A. G. Bhole. "Water Supply" in Civil Engineering Practice, Vol.

5. P. N. Cheremisinoff et al., eds. Lancaster, PA:Technomic Publishing Co.. Inc..
pp. 3-69 (1988).
2 Lai, B. B. and K. N. Dikshit. "A 2,000-Year-Old Feat of Hydraulic Engineering
in India," Archaeology, pp. 49-53 (1980).
3 Encyclopaedia Britannica, Vol. 19, 15th edition. Encyclopedia Britannica. Inc.,
p. 648 (1979).
4 Rouse, H. and S. Ince. History of Hydraulics. New York, NY:Dover Publications,
Inc. (1957).
5 Pierluigi, M. "A Review of the History of the Aqueducts of Rome,"/ Am. Water
Works Assoc, 68(11) :560-567 (1976).
6 LaNier, J. M. "Historical Developments of Municipal Water Systems in the
United States 1776-1976,"/ Am. Water Works Assoc, 68(4):173-180 (1976).
7 Bhave, P. R. "Historical Development of Transportation by Pipelines," Keynote
Address delivered at First National Seminar on Transportation by Pipelines. Oc-
tober 17-18, 1987, Nagpur, India, pp. 1-5 (1987).
8 Bhave, P. R. and C. F. Lam. "Transportation Planning and Network Geometry:
Optimal Layout for Branching Distribution Networks," in Civil Engineering Prac-
tice, Vol. 4. P. N. Cheremisinoff etal., eds. Lancaster, PA:Technomic Publishing
Co., Inc., pp. 473-497 (1988).
9 Streeter, V. L. and E. B. Wylie. Fluid Mechanics, Eighth edition. New York. NY:
McGraw Hill, Inc. (1984).
10 Weisbach, J. Die Experimental Hydraulik. Freiberg, Germany:Enge!hardt (1855).
11 Nikuradse, J. "Stromungsgesetze in Rauhen Rohren," Verein Deutscher Ingeni-
eure, Forschungsheft, p. 361 (1933).
12 Colebrook, C. F. "Turbulent Flow in Pipes with Particular Reference to the Tran-
sition Region between the Smooth and Rough Pipe Laws." J. Inst. of Civil Engrs.,
11:133-156(1938-1939).
13 King, H. W. and E. F. Brater. Handbook of Hydraulics, Fifth edition. New York.
NY:McGraw Hill Book Co., Inc. (1963).
14 Jeppson, R. W. Analysis of Flow in Pipe Networks. Ann Arbor, MI:Ann Arbor
Science Publishers, Inc. (1977).
15 Lamont, P. A. "Common Pipe Flow Formulas Compared with the Theory of
Roughness," 7. Am. Water Works Assoc., 73(5):274-280 (1981).
441
442 References

16 Johnson, S. P. "A Survey of Flow Calculation Methods," Preprinted Paper for

Summer Meeting, Am. Soc. of Mech. Engrs. (1934).
17 Rouse, H. "Evaluation of Boundary Roughness," Proc. Second Hydraul. Conf.,
Iowa Hydraulic Bulletin No. 27, Univ. of Iowa, Iowa City, IA (1943).
18 Rouse, H. "Some Paradoxes in the History of Hydraulics," J. Hydraul. Div., Am.
Soc. of Civil Engrs., 106(HY6):1077-1084 (1980).
19 Moody, L. F. "Friction Factors for Pipe Flow," Trans., Am. Soc of Mech. Engrs.,
66:671-684 (1944).
20 Powell, R. W. "Diagram Determines Pipe Sizes Directly," Civil Engineering, Am.
Soc. of Civ. Engrs., 20(9):45-46 (1950).
21 Ackers, P. "Resistance of Fluids Flowing in Channels and Pipes," Hydraulic Re-
search Paper, Her Majesty's Stationery Office, London, England (1958).
22 Thiruvengdam, A. "Direct Solution of Pipe Friction Problems from Colebrook-
White Formula," Water Power, 12:262-263 (1960).
23 Ramalho, R. S., F. M. Tiller and V. J. Berry. "New Friction Factor Chart for Pipe
Flow," Chemical Engineering, 71(15):153-154 (1964).
24 Bewtra, J. K. "Diagram for Resistance to Flow in Uniform Conduits," Environ-
mental Health, 7(4):256-261 (1965).
25 Barr, D. I. H. and A. A. Smith. "Application of Similitude to the Correlation of
Uniform Flow Data," Proc. Inst. of Civil Engrs., 37:487-509 (1967).
26 Asthana, K. C. "Transformation of Moody Diagram," J. Hydraul. Div., Am. Soc.
of Civil Engrs., 100(HY6):797-808 (1974).
27 Li. W. H. "Direct Determination of Pipe Size," Civil Engineering, Am. Soc. of
Civil Engrs., 44(6):74 (1974).
28 Bar, D. I. H. "Explicit Working for Turbulent Pipe Flow Problems," J. Hydraul.
Div., Am. Soc. of Civil Engrs., 102(HY5):667-673 (1976).
29 Blasius, H. "Das Ahnlichkeitsgesetz bei Reibungsvorgangen in Flussigkeiten,"
Verein Deutscher Ingenieure, Forschungsheft, 131 (1913).
30 Lamont, P. "The Choice of Pipe Flow Laws for Practical Use," Water and Water
Engineering, 73(875):9-17; and 73(876):55-63 (1969).
31 Ger, A. M. and E. R. Holley. "Comparison of Single Point Injections in Pipe
Flow,"/. Hydraul. Div, Am. Soc. of Civil Engrs., 102(HY6):731-746 (1976).
32 Te::ho, R., R. R. Tickner and R. E. James. "An Accurate Equation for the Com-
putation of Friction Factor for Smooth Pipes from the Reynolds Number," J. of
Applied Mechanics, 32:443 (1965).
33 Chen, J. J. J. "A Simple Explicit Formula for the Estimation of Pipe Friction Fac-
tor," Proc. Inst. of Civil Engrs., 77(2):49-55 (1984).
34 Chen, J. J. J. 'Systematic Explicit Solutions of the Prandtl and Colebrook-White
Equations for Pipe Flow," Proc. Inst. of Civil Engrs., 79(2):383-389 (1985).
35 Nahavandi. A. N. and G. V. Catanzaro. "Matrix Method for Analysis of Hydrau-
lic Networks,"/ Hydraul- Div., Am. Soc. of Civil Engrs., 99(HYl):47-63 (1973).
36 Wood, D. J. "An Explicit Friction Factor Relationship," Civil Engineering, Am.
Soc. of Civil Engrs., 36(12):60~61 (1966).
37 Barr, D. I. H. "New Forms of Equations for the Correlation of Pipe Resistance
Data," Proc. Inst. of Civil Engrs., 53(2):383-390 (1972).
38 Barr, D. I. H. "Two Additional Methods of Direct Solution of the Colebrook-
White Function," Proc. Inst. of Civil Engrs., 59(2):827-835 (1975).
 References 443

39 Churchill, S. W. "Friction Factor Equation Spans All Fluid Flow Regimes,"

Chemical Engineering, 84:91-92 (1977).
40 Jain, A. K. "Accurate Explicit Equation for Friction Factor" J. Hydraul. Div.,
Am. Soc. of Civil Engrs., 102(HY5):674-677 (1976).
41 Swamee, P. K. and A. K. Jain. "Explicit Equations for Pipe Flow Problems," J.
Hydraul. Div., Am. Soc. of Civil Engrs., K)2(HY5):657-664 (1976).
42 Collins, M. A. Discussion of "Explicit Equations for Pipe Flow Problems," by
P. K. Swamee and A. K. Jain, J. Hydraul. Div., Am. Soc. of Civil Engrs.,
102(HY11):1707-1709 (1976).
43 Zigrang, D. J. and N. D. Sylvester. "Explicit Approximations to the Solution of
Coiebrook's Friction Factor Equation," J. Am. Inst. of Chemical Engrs., 28(3):
514-515 (1982).
44 Haaland, S. E. "Simple and Explicit Formulas for the Friction Factor in Turbu-
lent Pipe Flow," / Fluids Engrg., 105:89-90 (1983).
45 Williams, G. S. and A. Hazen. Hydraulic Tables, 3rd edition. New York, NY:
John Wiley and Sons, Inc. (1933).
46 Scobey, F. C. "The Flow of Water in Commercially Smooth Pipes," Water Re-
sources Center Archives, Series Report 17, Univ. of California, Berkeley, CA
(1966).
47 Pomeroy, R. D. "Flow Velocities in Pipelines,"/ Hydraul.Engrg., Am. Soc. of
Civil Engrs., 109(8): 1108-1117 (1983).
48 White, W. R. "The Hydraulic Characteristics of Clayware Pipes," Hydraulic Re-
search Station, Report No. Int. 133, Wallingford, Berkshire, England (1974).
49 Jain, A. K., D. M. Mohan and P. Khanna. "Modified Hazen-Williams Formula,"
J. Environ. Engrg. Div., Am. Soc. of Civil Engrs., 1O4(EE1):137-146 (1978).
50 Powell, R. W. "The Origin of Manning Formula,"/ Hydraul. Div., Am. Soc. of
Civil Engrs., 94(HY4): 1179-1181 (1968).
51 Williams, G. P. "Manning Formula—A Misnomer?"/. Hydraul. Div., Am. Soc.
of Civil Engrs., 96(HY1): 193-200 (1970).
52 Hamberg, D. and U. Shamir. "Schematic Models for Distribution Systems De-
sign. I: Combination Concept,"/ Water Resour. Ping. Mgmt., Am. Soc. of Civil
Engrs., 114(2): 129-140 (1988). -
53 Colebrook, C. F. and C. M. White. "The Reduction of Carrying Capacity of Pipes
with Age,"/ Inst. of Civil Engrs., 7:99-118 (1937).
54 Committee. "Pipeline Friction Coefficients and the Effect of Age Thereon," J.
New England Witer Works Assoc., 49:235 (1935).
55 California Section Committee. "Loss in Capacity of Water Mains," / Am. Water
Works Assoc, 54(10):1293-1312 (1962).
56 Bhave, P. R. Unpublished Lecture Notes.
57 Ghatpande, M. M. and N. Y. Dhekne. "Pipe Friction Coefficient," Indian J. of
Environ. Hlth., 10:163-176 (1968).
58 Raman, V., S. K. Pathak, A. W. Deshpande, S. C. Dave and K. E. Rosario. "Sta-
tus of Distribution Systems in India with Reference to Leakage and Carrying Ca-
pacity," / Indian Whter Works Assoc, ll(l):l-10 (1979).
59 Dhabadgaonkar, S. M. and C. S. Thakre. "Determination of Carrying Capacity of
a Pipe Line,"/ Indian Water Works Assoc, 16(l):19-25 (1984).
60 Hudson, W.D. "Studies of Distribution System Capacity in Seven Cities," / Am.
Water Works Assoc, 58(2):157-164 (1966).
444 References

61 Archer, W. H. "Loss of Head Due to Enlargement in Pipes," Transactions, Am.

Soc. of Civil Engrs., 76:999-1026 (1913).
62 Gibson, A. H. "The Conversion of Kinetic to Pressure Energy in the Flow of
Water through Passages Having Divergent Boundaries," Engineering, 93:205
(1912).
63 Parker, P. M. The Control of Water. London:Routledge and Kegan Paul Ltd., pp.
796-800 (1925).
64 Ito, H. "Friction Factors for Turbulent Flow in Curved Pipes," J. Basic Engrg.,
81(D):123-134 (1959).
65 Ito, H. "Pressure Losses in Smooth Pipe Bends,"/ Basic Engrg., 82(D):131-143
(1960).
66 Amirtharajah, A. "Design of Granular Media Filters," Chapter 28 in Water Treat-
ment Plant Design. R. L. Sanks, ed. Ann Arbor, MI: Ann Arbor Science Publish-
ers, Inc., pp. 675-737 (1978).
67 Fietz, T. R. Discussion of "Pressure Reducing Valves in Pipe Network Analysis,"
by R. W. Jeppsonand A. L. Davis. J. Hydraul. Div., Am. Soc. of Civil Engrs.,
1O3(HY5):578-58O (1977).
68 Jeppson, R. W. and A. L. Davis. Closure to "Pressure Reducing Valves in Pipe
Network Analysis,"/ Hydraul. Div., Am. Soc. of Civil Engrs., 103(HYll):1362
(1977).
69 Jeppson, R. W. "Equivalent Hydraulic Pipe for Parallel Pipes," J. Hydraul. Div.,
Am. Soc. of Civil Engrs., 1O9(HY1):35^5 (1982).
70 Walski, T. M. "Using Water Distribution Network Models," / Am. Water Works
Assoc, 75(2):58-63 (1983).
71 Barnard, R. E. "Design Standards for Steel Water Pipe," J. Am. Water Works As-
soc. , 40(1): 24 (1948).
72 Zimke, P. C. "Pressure Drop in Pipe Bends and Elbows," / Am. Water Works As-
soc, 53(3):329-33O (1961).
73 Anonymous. "Design and Installation of Steel Water Pipe: Hydraulics of Pipe-
lines," / Am. Water Works Assoc, 53(6):757-782 (1961).
74 Warring, R. H. Handbook of Valves, Piping and Pipelines. Houston, TX:Gulf
Publishing Co. (1982).
75 Collins, M. A., L. Cooper, R. V. Helgason and J. L. Kennington. "Solution of
Large Scale Pipe Networks by Improved Mathematical Approaches," Technical
Report IEOR 77016-WR-77O01, School of Engrg. and Applied Science, South
Methodist Univ., Dallas TX (1978).
76 Collins, M. A., L. Cooper and J. L. Kennington. "Multiple Operating Points in
Complex Pump Networks," J. Hydraul. Div., Am. Soc. of Civil Engrs., Vol. 105
(HY3):229-244 (1979).
77 Jeppson, R. W. and A. Travallae. "Pumps and Reservoirs in Networks by Linear
Theory,"/ Hydraul. Div.,- Am. Soc. of Civil Engrs., 101(HY3): 576-580 (1975).
78 Donachie, R. P. "Digital Program for Water Network Analysis," / Hydraul. Div.,
Am. Soc. of Civil Engrs., 1OO(HY3):393-4O3 (1974).
79 Chandrashekar, M. "Extended Set of Components in Pipe Networks,"/ Hydraul.
Div., Am. Soc. of Civil Engrs., 1O6(HY1):133-149 (1980).
80 Fietz, T. R. "Improved Head-Discharge Relations for Pipe Network Analysis by
the Loop Method," / Hydraul. Research, 11(2):123-136 (1973).
 References 445

81 Hazen, R. "Pumps and Pumping Stations," J. New England Viaer Works Assoc.,
67:121 (1953).
82 Fair, G. ML, J. C. Geyer and D. A. Okun. Elements of Water Supply and Waste
Water Disposal. New York, NY:John Wiley and Sons, Inc. (1971).
83 Jeppson, R. W. and A. L. Davis. "Pressure Reducing Valves in Pipe Network
Analysis," J. Hydraul. Div., Am. Soc. of Civil Engrs., 102(HY7):987-1001
(1976).
84 Jacoby, S. L. S. "Design of Optimal Hydraulic Networks,"/ Hydraul. Div., Am.
Soc. of Civil Engrs., 94(HY3):641-661 (1968).
85 Bhave, P. R. "Noncomputer Optimization of Single-Source Networks" J. Envi-
ron. Engrg. Div., Am. Soc. of Civil Engrs., 104(EE4):799-814 (1978).
86 Bhave, P. R. "Optimization of Gravity-Fed Water Distribution Systems: Theory,"
J. Environ. Engrg., Am. Soc. of Civil Engrs., 109(1): 189-205 (1983).
87 Harary, F. Graph Theory. Reading, MA:Addison-Wesley Publishing Co. (1969).
88 Shamir, U. and C. D. D. Howard. "Water Distribution Systems Analysis," J. Hy-
draul. Div., Am. Soc. of Civil Engrs., 94(HYl):219-234 (1968).
89 Shamir, U. and C. D. D. Howard. Closure to "Water Distribution Systems Analy-
sis," J. Hydraul. Div., Am. Soc. of Civil Engrs., 96(HY2):577-578 (1970).
90 Shamir, U. and C. D. D. Howard. "Engineering Analysis of Water Distribution
Systems," J. Am. Water Works Assoc, 69(9):510-514 (1977).
91 Gofman, E. and M. Rodeh. "Loop Equations with Unknown Pipe Characteris-
tics," J. Hydraul. Div., Am. Soc. of Civil Engrs., 107(9): 1047-1060 (1981).
92 Bhave, P. R. "Rules for Solvability of Pipe Networks" / Indian Water Works
Assoc, 22(1):7-10 (1990).
93 Bhave, P. R. "Analysis of Water Distribution Networks: Part I," J. Indian Water
Works Assoc, 13(2): 149-154 (1981).
94 Bhave, P. R. "Analysis of Water Distribution Networks: Part II," J. Indian Water
Works Assoc, 13(3):245-252 (1981).
95 Bhave, P. R. "Analysis of Water Distribution Networks: Part HI,"/ Indian Water
Works Assoc, 13(4):305-314 (1981).
96 Bhave, P. R. "Unknown Pipe Characteristics in Hardy Cross Method of Network
Analysis," J. Indian Water Works Assoc, 18(2):133-135 (1986).
97 Bhave, P. R. "Unknown Pipe Characteristics in Network Analysis: Formulation
of Equations" (to be published).
98 Bhave, P. R. "Unknown Pipe Characteristics in Network Analysis: Solution of
Equations" (to be published).
99 Saaty, T. L. and J. Bram. Nonlinear Mathematics. New York, NY:McGraw Hill
Book Co. (1964).
100 Cross, Hardy. "Analysis of Flow in Networks of Conduits or Conductors," Bulle-
tin No. 286, Engrg. Experiment Station, Univ. of Illinois, Urbana, IL (1936).
101 Cornish, R. J. "The Analysis of Flow in Networks of Pipes," / Inst. of Civil En-
grs., 13:147-154(1939).
102 Liu, K. T. H. "The Numerical Analysis of Water Supply Networks," International
Assoc. of Hydraul. Research, 13th Congress, Kyoto, Japan, 1:35-42 (1969).
103 Bhave, P. R. "Rapid Convergence in Hardy Cross Method of Network Analysis,"
/ Indian Water Works Assoc, 16(l):l-5 (1985).
104 Borgohain, T. P., A. K. Khare, D. J. Desai, K. P. Singh and S. Banerjee. "Com-
446 References

puter-Aided Dynamic Analysis of Water Distribution Systems," B. E. Project Re-

port, Nagpur University (1988).
105 Conklin, J. "Flow in Pipe Networks for Direct Determination," Engrg. News
Record, 127:370 (1941).
106 Williams, G. N. "Enhancement of Convergence of Pipe Network Solutions," J.
Hydraul. Div., Am. Soc. of Civil Engrs., 99(HY7): 1057-1067 (1973).
107 Epp, R. and A. G. Fowler. "Efficient Code for Steady-State Flows in Networks,"
J. Hydraul. Div., Am. Soc. of Civil Engrs., 96(HYl):43-56 (1970).
108 Prim, R. C. "Shortest Connection Networks and Some Generalizations," The Bell
System Technical Journal, 36(6):1389- 1401 (1957).
109 Bhave, P. R. Closure to "Optimization of Gravity-Fed Water Distribution Sys-
tems: Theory,"/ Environ. Engrg., Am. Soc. of Civil Engrs., 110(2):509-512
(1984).
110 Voyles, C. F. and H. R. Wilke. "Selection of Circuit Arrangements for Distribu-
tion Network Analysis by the Hardy Cross Method," J. Am. Water Works Assoc.,
54(3):285-290 (1962).
111 Dillingham, J. H. "Computer Analysis of Water Distribution Systems—Part II,"
Water and Sewage Works, 43-45 (1967).
112 McCormick, M. and C. J. Bellamy. "A Computer Program for the Analysis of
Networks of Pipes and Pumps," J. Inst. of Engrs., Australia, 38(3):51-58 (1968).
113 Kootattep, S. and H. Aya. "Appropriate Method of Distribution Network Analysis
for Developing Countries," Aqua, 6:311-315 (1985).
114 Hildebrand, F. B. Introduction to Numerical Analysis. New York, NY:McGraw
Hill Book Co. Inc., pp. 443-451 (1956).
115 Scarborough, J. B. Numerical Mathematical Analysis, 6th ed.. Baltimore, MD:
The Johns Hopkins Press, pp. 201-207 (1966).
116 Martin, D. W. and G. Peters. "The Application of Newton's Method to Network
Analysis by Digital Computer," J. Inst. of Water Engrs., 17:115-129 (1963).
117 Pitchai, R. "A Model for Designing Water Distribution Pipe Networks," Ph.D.
diss., Harvard University (1966).
118 Lam, C. F. and M. L. Wolla. "Computer Analysis of Water Distribution Systems:
Part I—Formulation of Equations,'' J. Hydraul. Div., Am. Soc. of Civil Engrs.,
98(HY2):335-344 (1972).
119 Lam, C. F. and M. L. Wolla. "Computer Analysis of Water Distribution Systems:
Part II—Numerical Solution," J. Hydraul. Div., Am. Soc. of Civil Engrs.,
98(HY3):447-460 (1972).
120 Grooms, H. R. "Algorithm for Matrix Band Width Reduction," J. Struc. Div.,
Am. Soc. of Civil Engrs., 98(ST1):203-214 (1972).
121 Mcllroy, M. S. "Pipeline Network Flow Analysis Using Ordinary Algebra," J.
Am., Water Works Assoc, 41(6):422-428 (1949).
122 Marlow, T. A., R. L. Hardison, H. Jacobson and G. E. Biggs. "Improved Design
of Fluid Networks with Computers," J. Hydraul. Div., Am. Soc. of Civil Engrs.,
92(HY4):43-61 (1966).
123 Muir, J. F. Discussion of "Improved Design of Fluid Networks with Computers,"
by T. A. Marlow et al., J. Hydraul. Div., Am. Soc. of Civil Engrs., 93(HY2):
88-90 (1967).
124 Wood, D. J. and C. O. A. Charles. "Hydraulic Network Analysis Using Linear
Theory,"/ Hydraul. Div., Am. Soc. of Civil Engrs., 98(HY7): 1157-1170 (1972).
 References 447

125 Collins. A. G. and R. L Johnson. "Finite-Element Method for Water-Distribu-

tion Networks," J. Am. Water Works Assoc, 67(7): 385-389 (1975).
126 Issacs. L. T. and K. G. Mills. "Linear Theory Methods for Pipe Network Analy-
sis," J. Hydmul. Div., Am. Soc. of Civil Engrs., 106(HY7): 1191 -1201 (1980).
127 Issacs, L. T. and K. G. Mills. "Computer Method for Pipe Network Analysis,"
Queensland Division Technical Papers, Inst. of Engrs., Australia (Queensland
Division). 21(1O):9-13 (1980).
128 Vasudeo Rao, B. "Finite Element Analysis of Flow Networks," Engineering Anal-
ysis, Southampton, England:Computational Mechanics Publications, 4(l):35-39
(1987).
129 Vasudeo Rao, B. and Medha Dixit. "Design of Water Supply Networks by Finite
Element Method," First National Seminar on Transportation by Pipelines, Octo-
ber 17- 18, 1987, Nagpur, India, pp. 29-36 (1987).
130 Watanatada, T. Discussion of "Hydraulic Network Analysis Using Linear
Theory," by D. J. Wood and C. 0. A. Charles, J. Hydmul. Div., Am. Soc. of Civil
Engrs., 99(HY4):693-694 (1973).
131 Bhave, P. R. and V. V. Sonak. "Rapid Convergence in Network Analysis Using
Linear Theory" (to be published).
132 Greenlaw, R. and D. Clack. Discussion of "Hydraulic Network Analysis Using
Linear Theory," by D. J. Wood and C. O. A. Charles, J. Hydmul. Div., Am. Soc.
of Civil Engrs., 99(HY4):694-697 (1973).
133 McPherson. M. B. "Application of System Analyzers," Water and Sewage Works
Reference Number, November R53-R67 (1962).
134 Anonymous. Water Works Engineering in Disaster. Washington, DC:Office of
Civil Defense. Publication No. 2022 (1943).
135 Camp. T. R. "Hydraulics of Distribution Systems—Some Recent Developments
in Methods of Analysis," J. New England Water Wks. Assoc., 57:334-358 (1943).
136 Gavett, W. "Computation of Flows in Distribution Systems," J. Am. Water Works
Assoc, 35(3):267-287 (1943).
137 Corgill, W. E., L. Murray, J. Collins and R. Naugle. "Methods of Water Distribu-
tion Systems Analysis," Bucknell Engineer, Lewisburg, PA, l(2):6-7, 15 (1950).
138 Cobb, E. B. "The Analysis of Complex Waterworks Distribution Networks," J.
New England Water Works Assoc, 70(l):29 (1956).
139 Cobb, E. B. "Simplified Analysis of Complex Waterworks Distribution Net-
works," J. New England Water Works Assoc, 73:275-293 (1959).
140 Mcllroy, M. S. "Studies on Fluid Network Analysis and the Basic Design of an
Electric Analyzer for Fluid Distribution System," Ph.D. diss., Massachusetts In-
stitute of Technology (1946).
141 Mcllroy, M. S. "Direct-Reading Electric Analyzer for Pipeline Networks,"/ Am.
Water Works Assoc, 42(4):347-366 (1950).
142 Mcllroy, M. S. "Water Distribution Systems Studied by a Complete Electrical
Analogy,"/ New England Water Works Assoc, 65(4):299-318 (1951).
143 Camp, T. R. and H. L. Hazen. "Hydraulic Analysis of 'W&ter Distribution Sys-
tems by Means of an Electric Network Analyzer," / New England Water Works
Assoc, 48(4):383-407 (1934).
144 Reid, G. W. and L. B. Wolfenson. "How Electric Calculators Solve Water Distri-
bution Problems," Public Works, 84(10):78 (1953).
448 References

145 Perry, H. A., Jr., D. E. Vierling and R. W. Kohler. "Network-Flow Analysis

Speeded by Modified Electrical Analogy," Engrg. News Record, 143(12):19-23
(1949).
146 Suryaprakasam, M. V., G. W. Reid and J. C. Geyer. "Use of Alternating Current
Network Calculator in Distribution System Design," J. Am. Witer Works Assoc.,
42(12):1154-1164 (1950).
147 Clennon, J. P. "Calculators Solve Flow Problems," Am. Gas Assoc. Monthly,
33(10):19, 37(1951).
148 Raman, V. "Hydraulic Analysis of Pipeline Networks by Direct Electrical Anal-
ogy Method,"/ Institution of Engrs. (India), 42(9):467^87 (1962).
149 McPherson, M. B. and J. V. Radziul. "Water Distribution Design and the Mcllroy
Network Analyzer,'"/ Hydraul. Div., Am. Soc. of Civil Engrs., 84(HY2):1588-1
to 1588-15 (1958).
150 Graves, Q. B. and D. Branscome. "Digital Computers for Pipeline Network Anal-
ysis," / San. Engrg. Div., Am. Soc. of Civil Engrs., 84(SA2): 1608-1 to 1608-18
(1958).
151 Linaweaver, F. P., Jr. "Use of Network Analyzers at Baltimore," J. Am. Water
Works Assoc, 54(12): 1518-1530 (1962).
152 Wood, D. J. "Analog Analysis of Water Distribution Networks * Transp. Engrg. J.,
Am. Soc. of Civil Engrs., 97(TE2):281-290 (1971).
153 Fox, J. A. and A. E. Keech. "Pipe Network Analysis—A Novel Steady State
Technique," / Water Engrs. and Scientists, 29(4):183-194 (1975).
154 Wylie, E. B. and V. L. Streeter. Fluid Transients. New York, NY:McGraw Hill
International Book Co. Inc. (1978).
155 Demuren, A. O. and F. J. K. Ideriah. Closure to "Pipe Network Analysis by Par-
tial Pivoting Method," / Hydraul. Engrg., Am. Soc. of Civil Engrs., 113(12):
1569-1570 (1987).
156 Arora, M. L. "Flow Split in Closed Loops Expending Least Energy," / Hydraul.
Div., Am. Soc. of Civil Engrs., 102(HY3):455-458 (1976).
157 Hall, M. A. "Hydraulic Network Analysis Using (Generalized) Geometric Pro-
gramming," Networks, 6:105-130 (1976).
158 Rao, H. S., L. C. Markel and D. W. Bree, Jr. "Extended Period Simulation of
Water Systems—Part B , " / Hydraul. Div., Am. Soc. of Civil Engrs., 103(HY3):
281-294 (1977).
159 Tart, J. S. "Wue of Telemetering and Recording," J. Am. Water Works Assoc.,
65(2): 123-127 (1973).
160 Demoyer, R., H. D. Gilman and M. Y. Goodman. "Dynamic Computer Simula-
tion and Control Methods for Water Distribution Systems," Office of Water Re-
sources Research, U.S. Department of the Interior (1973).
161 Rao, H. S. and D. W. Bree, Jr. "Extended Period Simulation of Water Systems-
Part A , " / Hydraul. Div., Am. Soc. of Civil Engrs., 103(HY2):97-108 (1977).
162 Bhave, P. R. "Extended Period Simulation of Water Systems—Direct Solution,"/
Environ. Engrg., Am. Soc. of Civil Engrs., 114(5): 1146-1159 (1988).
163 Bhave, P. R. "Node Flow Analysis of Serial Water Distribution Systems," / In-
dian Water Works Assoc, 12(1): 17-23 (1980).
164 Bhave, P. R. "Node Flow Analysis of Branching Water Distribution Systems," /
Indian Water Works Assoc, 13(1):25-30 (1981).
 References 449

165 Bhave, P. R. "Node Flow Analysis of Water Distribution Systems," Transportation

Engrg. Jour., Am. Soc. of Civil Engrs., 107(TE4):457-467 (1981).
166 Cembrowicz, R. G. and J. J. Harrington. "Capital Cost Minimization of Hydrau-
lic Networks," J. Hydraul. Div., Am. Soc. of Civil Engrs., 99(HY3):431-440
(1973).
167 Lam, C. F. "Discrete Gradient Optimization of Water Systems," J. Hydraul. Div.,
Am. Soc. of Civil Engrs., 99(HY6):863-872 (1973).
168 Rasmusen, H. J. "Simplified Optimization of Water Supply Systems," J. Environ.
Engrg. Div., Am. Soc. of Civil Engrs., 102(EE2):313-327 (1976).
169 Deb, A. K. "Optimization of Water Distribution Network Systems," J. Environ.
Engrg. Div., Am. Soc. of Civil Engrs., 102(EE4):837-851 (1976).
170 Alperovits, G. and U. Shamir. "Design of Optimal Water Distribution Systems,"
Water Resources Research, 13(6):885-900 (1977).
171 Cenedese, A. and P. Mele. "Optimal Design of Water Distribution Networks," J.
Hydraul. Div., Am. Soc. of Civil Engrs., 104(HY2):237-247 (1978).
172 Quindry, G., E. D. Brill and J. Liebman. "Optimization of Looped Water Distri-
bution Systems," J. Environ. Engrg. Div., Am. Soc. of Civil Engrs., 107(EE4):
665-679 (1981).
173 Rowell, W. F. and W. J. Barnes. "Obtaining Layout of Water Distribution Sys-
tems," 7. Hydraul. Div., Am. Soc. of Civil Engrs., 108(HYl):137-148 (1982).
174 Bhave, P. R. "Optimization of Gravity-Fed Water Distribution Systems: Applica-
tion," J. Environ. Engrg., Am. Soc. of Civil Engrs., 109(2):383-395 (1983).
175 Bhave, P. R. "Optimal Expansion of Water Distribution Systems," J. Environ.
Engrg., Am. Soc. of Civil Engrs., 111(2):177-197 (1985).
176 Rajcoomar, R. C. K. "Node Flow Analysis of Water Distribution Systems," M.
Tech. diss., Nagpur University (1982).
177 Cesario, A. L. and J. Q Davis. "Calibrating Water System Models," J. Am. Water
Works Assoc, 76(7):66-69 (1984).
178 Walski, T. M. "Technique for Calibrating Network Models," J. Water Resour.
Planning and Mgmt., Am. Soc. of Civil Engrs., 109(4):360-372 (1983).
179 Research Committee on Distribution Systems. "Water Distribution Research and
Applied Development Needs,"/ Am. Water Works Assoc, 66(6):385-390 (1974).
180 Eggener, C. L. and L. B. Polkowski. "Network Models and the Impact of Model-
ing Assumptions," J. Am. Water Works Assoc, 68(4): 189-196 (1976).
181 Harris, R. U. "Computer Modeling in Water System Planning and Design," /.
Am. Water Works Assoc, 76(7):78-81 (1984).
182 Rahal, C. M., M. J. H. Sterling and B. Coulbeck. "Parameter Tuning for Simula-
tion Models of Water Distribution Networks," Proc. Inst. of Civil Engrs., 69(2):
751-772 (1980).
183 Walski, T. M. "Case Study: Pipe Network Model Calibration Issues," J. Water
Resour. Planning and Mgmt., Am. Soc. of Civil Engrs., 112(2):238-249 (1986).
184 Ormsbee, L. E. and D. J. Wood. "Explicit Pipe Network Calibration," J. Water
Resour. Planning and Mgmt., Am. Soc. of Civil Engrs., 112(2): 166-182 (1986).
185 Committee. "Form and Procedure for Fire Flow Tests," J. Am. Water Works
Assoc, 68(5):264-268 (1976).
186 Bhave, P. R. "Calibrating Water Distribution Network Models," J. Environ.
Engrg., Am. Soc. of Civil Engrs., 114(1): 120-136 (1988).
450 References

187 Walski, T. M. "Model Calibration in Austin, Texas," International Symposium on

Computer Modeling of Water Distribution Systems, Lexington, KY (1988).
188 Walski, T. M. "Equipment Needs for Field Data Collection in Support of Model-
ing," International Symposium on Modeling of Water Distribution Systems, Lex-
ington, KY (1988).
189 Wood, D. J. and A. G. Rayes. "Reliability of Algorithms for Pipe Network Analy-
sis,"/ Hydraul. Div., Am. Soc. of Civil Engrs., 107(HY10): 1145-1161 (1981).
190 Collins, M. A. "Pitfalls in Pipe Network Analysis Techniques," ASCE Convention
and Exposition, October 23-25, Atlanta, GA, preprint 3730. pp. 1-11 (1979).
191 de Neufville, R., J. Schaake, Jr. and J. M. Stafford. "Systems Analysis of Water
Distribution Networks,"/ San. Engrg. Div., Am. Soc. of Civil Engrs., 97(SA6):
825-842 (1971).
192 Demelin, E., U. Shamir and N. Arad. "Engineering and Economic Evaluation of
the Reliability of Water Supply," Water Resour. Res., 8(4):861-877 (1972).
193 Shamir, U. and C. D. D. Howard. "Water Supply Reliablity Theory," J. Am. Water
Works Assoc, 73(7):379-384 (1981).
194 Shamir, U. and C. D. D. Howard. "Reliability and Risk Assessment for Water
Supply Systems" in Computer Applications in Miter Resources. H. Torno, ed.
New York, NY:Am. Soc. of Civil Engrs. (1985).
195 Hobbs, B. F. "Reliability of Urban Water Supply" in Computer Applications in
Water Resources. H. Torno, ed. New York, NY:Am. Soc. of Civil Engrs. (1985).
196 Hobbs, B. F. "Reliability Analysis of Water System Capacity" in Hydrology and
Hydraulics in the Small Computer Age. W. Waldrop, ed. New York, NY:Am.
Soc. of Civil Engrs., pp. 341-346 (1985).
197 Mays, L. W. "Methods for Reliability Analysis of %ter Distribution Networks"
in Hydrology and Hydraulics in the Small Computer Age. W. Waldrop, ed. New
York, NY:Am. Soc. of Civil Engrs., pp. 347-352 (1985).
198 Cullinane, M. J. "Reliability Evaluation of Water Distribution System Compo-
nents" in Hydrology and Hydraulics in the Small Computer Age. W. Waldrop, ed.
New York, NY:Am. Soc. of Civil Engrs., pp. 353-358 (1985).
199 Tung, Y. K. "Evaluation of Water Distribution Network Reliability" in Hydrology
and Hydraulics in the Small ComputerAge. W. Waldrop, ed. New York, NY:Am.
Soc. of Civil Engrs., pp. 359-364 (1985).
200 Mays, L. W. and M. J. Cullinane. "A Review and Evaluation of Reliability Con-
cepts for Design of Water Distribution Systems," Misc. Paper EL-86-1, Vicks-
burg, MS, U.S. Engineers Waterways Experiment Station (1986).
201 Wagner, J. M., U. Shamir and D. H. Marks. "Reliability of Water Distribution
Systems," R86-18, Cambridge, MA, Ralph M. Parsons Laboratory, Massachusetts
Inst. of Tech. (1986).
202 Wagner, J. M., U. Shamir and D. H. Marks. "Water Distribution Reliability: Ana-
lytical Methods," J. Water Resour. Planning and Mgmt., Am. Soc. of Civil En-
grs., 114(3):253-275 (1988).'
203 Wagner, J. M., U. Shamir and'D. H. Marks. "Water Distribution Reliability: Sim-
ulation Methods," J. Water Resour. Planning and Mgmt., Am. Soc. of Civil En-
grs., U4(3):276-294 (1988).
204 Watters, G. Z. Modern Analysis and Control of Unsteady Flow in Pipelines. Ann
Arbor, MI:Ann Arbor Science Publishers, Inc. (1979).
 References 451

205 Coulbeck, B. "A Computer Program for Calibration of Water Distribution Sys-
tems," J. Inst. ofmter Engrs. and Scientists, 38(l):79-80 (1984).
206 Wood, D. J. Computer Analysis of Flow in Pipe Networks Including Extended Pe-
riod Simulations (KYPIPE). Lexington, KY:Office of Continuing Education, Uni-
versity of Kentucky (1980).
207 Charles Howard and Associates, Ltd. Water Distribution Network Analysis: SPP8
User's Manual. Victoria, B.C., Canada:Charles Howard and Associates, Ltd.
(1984).
 Index

ACKERS, P., 29 Anonymous, 83, 303

Adjacent loop influence factor, 221 ARAD, N., 420
Adjacent node influence factor, 223 ARCHER, W. H., 60
Adjustment factor, 393-401 ARORA, M. L., 312-313
for nodal flows, 397, 401 Artificially roughened pipes, 27
for pipe resistances, 397, 401 Assignment rule, 130-137
global, 393 ASTHANA, K. C , 29
zonal. 39.5 AWWA California Section Committee, 54
ALPEROVITS, G., 354 AWWA Committee, 396
Alternate constraints, 369 AWWA Research Committee, 386, 388,
AMIRTHARAJAH, A., 67-68 390
Analogy between hydraulic and electrical AYA, H., 227-228
networks, 303 305
Analysis Bandwidth of a matrix, 257
dynamic (see dynamic analysis) BANNERJKE, S., 220-224, 423
node flow (see node flow analysis) BARNARD, R. E., 83
node head (see node head analysis) BARNES, W. 1 , 354
reliability, 419-421 BARR, D. I. H., 29, 34, 36
sensitivity, 409-418 Basic unknown parameters, 142, 270. 319
static. 327 BEU, K. H., 67
unsteady flow, 421-422 BELLAMY, C. J., 227-228, 235
Analysis methods Bends, 67
comparison of, 409 Bernoulli's theorem, 16-19
finite element, 281-294 BERRY, V. J., 29
Hardy Cross (see Hardy Cross BEWTRA, J. K., 29
method) BHAVE, P. R., 1, 3, 56, 124, 125, 129-131.
linear theory (see linear theory 142, 156-168, 185, 195-202,
method) 220-224, 226-227, 295-301, 328,
Newton-Cross, 187 337-350, 353-384, 3 9 9 ^ 0 5 , 409
Newton-Raphson (see BHOLE, A. G., 1, 124, 185, 226-227
Newton-Raphson method) BIGGS, G. E., 263
through optimization, 312-325 BLASIUS, H., 31
through unsteady behaviour during BORGOHAIN, T. P., 220-224. 423
start-up, 308-312 BRAM, J., 185
using electrical analyzers, 303-308 Branching networks (see Network,
ANDERSON, A. G., 67 branching)
Annual growth rate of roughness, 53 BRANSCOMK, D . , 308

453
454 Index

BRATKR. E. R, 29, 42, 60-67 Configuration, 123

BRE-:I-:. D. W., J R . . 327 337,406 CONKLIN, J., 225
BRICiHTMORK, A. W., 60, 66 Connectivity, 420
BRH.I., H. D.. 354 Content model, 318-322
BRUGES, C. E., 43 Continuity equation, 8-14, 128, 310
Bulk modulus of elasticity, 5 Convergence, 233, 418
COOPER, L., 99, 312, 318-325
Calibration, 385-408 CORGILL, W. E., 303

accuracy, 387 CORNISH, R. J., 185,209

data collection and preparation, COULBECK, B., 387, 392-393, 405, 423
387 391 CRIMP, W. S., 43

definition, 385 CROSS, HARDY, 185

field equipment, 408 CULLINANE, M. J., 420-421

finetuning, 405
methods. 391-405 DAMELIN, E., 420
necessity, 385 DARCY, H. P. G., 26
reasonable agreement, 386 Darcy-Weisbach formula, 26-27, 50, 70,
sources of error, 406-408 263, 309
CAMP, T. R., 303, 305-306 Darcy-Weisbach friction coefficient, 26,
CATANZARO, G. V , 34 123
CEMBROWICZ, R. G., 354 DAVE, S. C , 54
CHNHDESE, A., 354 DAVIS, A. L., 68, 101, 112, 258-261
CHSARIO, A. I.., 385-387 DAVIS, J. O., 385-387
CHANDRASHEKAR. M., 100, 112 DEB, A. K., 354
Characteristic equations, 309 Degree of a node, 257
CHARU.S, C. O. A., 263-265, 300 Demand node, 119, 139
Check valves, 110-112, 172-175, 206-209, Demand pattern, 124, 327
217-219, 244-246, 254-256, 276-278, DEMOYER, R., 328
293-295 DEMUREN, A. O., 312
in external pipes, 172-173 DE NEUFVILLE, R., 419
in internal pipes, 173- 175 Density, 3
CHEN, J. J. J.. 33-36 DESAI, D. J., 220-224, 423
CHURCHILL, S. W., 34 36 DESHPANDE, A. W., 54
Circuits, 120 DHABADGAONKAR, S. M., 54
CLACK, D., 300 DHEKNE, N. Y., 54
CLENNON, J. P.. 306 DIKSHIT, K. N., 2
COBB, E. B., 303 DILLINGHAM, J. H., 227-228
Cocoment model, 322-325 Distribution network (see network)
Coefficient matrix, 235 DIXIT, MEDHA, 263, 281
Coefficient of friction (see friction DONACHIE, R. P., 100
coefficient) Dynamic analysis (Extended period
Coefficient of roughness, 41 analysis, Extended period
COLEBROOK. C. R, 28. 32, 53 simulation), 327-352
Colebrook-White equation, 32 direct method, 337-350
COLLINS, A. G., 263, 279, 280, 291, 301 iterative method, 328-337
COLLINS, J., 303
COLLINS, M. A., 34, 99. 312, 318-325, Effect of age on pipe carrying capacity,
418-419 52-57
Compressibility. 5 EGGENER, C. L., 387-388, 390, 406
Computer application, 422-423 Elbows, 67
Computer programs, 423 Electrical analyzers, 303-308
 Index 455

with linear resistors, 305-306 laminarflow,31

with nonlinear resistors, 306-308 smooth turbulentflow,31-32
Elevation head, 17 rough turbulent flow, 32-33
Energy, 14-16, 313 transitionalflow,32
flow (pressure), 15 transitional turbulentflow,32
forms of, 16-17 FULLER, W. E., 67
kinetic, 16 GAUCKLER, P. G., 42
potential, 15 GAVETT, W., 303
system, 313 GER, A. M., 32
Energy grade line, 20 GEYER, J. C , 104, 306, 388
Energy principle, 14-20 GHATPANDE, M. M. 54
EPP, R., 225, 228, 235, 257-258, 261 GIBSON, A. H., 65
Equations, 139-183, 309 GILMAN, H. D., 328
characteristic, 309 GOFMAN, E., 129-130, 235, 248
formulation of, 139-183 GOODMAN, M. Y., 328
Equivalent pipes (see pipes, equivalent) GRAVES, Q. B., 308
Equivalent pipe lengths, 83 GREENLAW, R., 300
Equivalent surface roughness, 28-29 GROOMS, H. R., 257
Explicit friction coefficient relationships, Ground node, 318-325
33-36 //equations, 145-146, 152-154, 161-164,
all flow regimes, 36 169-170, 173, 174, 177, 179-180,
smooth turbulentflow,33 235-246, 257, 263, 278-295,
transitional turbulent flow, 33-35 297-301, 343-344
Extended period analysis (see dynamic H-q equations, 152-154, 164, 284-290
analysis) H-q-R equations, 161-164, 239-241,
Extended period simulation (see dynamic 286-290
analysis) H-R equations, 161-164,241-242,
FAIR, G. M., 104, 388 286-290
FIETZ, T. R., 68, 101, 112 AH equations, 147-148, 155-156, 166-168,
Finite element method, 281-295 171, 173, 175, 178-179, 347
Fire flow testing, 396 AH-Aq-AR equations, 166-168
FLAMONT, A. A., 43 AH-AR equations, 166-168
Flow, 28 HAALAND, S. E., 35-36
laminar, 28 HAGEN, G. H. L., 43
rough turbulent, 28 HALL, M. A., 312
smooth turbulent, 28 HAMBERG, D., 52
transitional, 28 HARARY, F., 126
transitional turbulent, 28 HARDISON, R. L., 263
turbulent, 28 Hardy Cross method, 185-229
Flow energy, 15 convergence problems, 227-229
Fluid properties, 3-8 modified, 220-224
Fluisters, 306-308 of balancing flows, 185, 209-219,
Foss, W. E., 43 222-224
FOWLER, A. G., 225-228, 235, 257-258, basic concept, 209-211
261 networks with known pipe
Fox, J. A., 308-312 resistances, 211
Free variables, 412 networks with pumps, 214-217
Friction coefficient, 31-36 networks with unknown pipe
explicit relationships (see explicit resistances, 211-214
friction coefficient networks with valves, 217-219
relationships) of balancing heads, 185-209, 220-222
456 Index

Hardy Cross method, of balancing due to valves, 68, 112

heads (continued) major, 25-58
basic concept, 185-189 minor, 25, 59-68
networks with known pipe Head loss formulas, 26-27, 36-45
resistances, 189-195 comparison, 43-44
networks with pumps, 202-205 Darcy-Weisbach, 26-27
networks with unknown pipe empirical, 36-43
resistances, 195-202 general. 44-45
networks with valves, 206-209 Ha/en-Williams, 38-41
selection of initial values, 224-227 Manning, 42-43
HARRINGTON, J. J., 354 modified Hazen-Williams, 41-42
HARRIS, R. U.. 387 HELGASON, R. V., 99, 312, 318-325
HAZEN, A., 38, 54 HILDEBRAND, F. B., 231-235
HAZEN, H. L., 305-306 Historical development of distribution
HAZEN, R., 104 systems, 2-3
Hazen-Williams coefficient, 38-41, 53-57, general, 2-3
123 in U.S.A., 3
effect of age on, 53-57 HOBBS, B. R, 420
Hazen-Williams formula, 38-42, 70, 263 HOLLEY, E. R., 32
Head HOWARD, C. D. D.,129, 235, 239, 328,
operating, 98 386, 409-413, 420
piezometric (hydraulic), 17 HUDSON, W. D., 54
potential (position, elevation), 17 Hydraulic grade line, 20
pressure, 17 Hydraulic gradient level, 124-125
static, 97
shut-off, 100 IDERIAH, F. J. K., 312
system, 97 INCE, S., 2, 60
velocity, 17 ISSACS, L. T , 263,279
total, 17 ITO, H., 67
Head-discharge curves, 97-99
pump, 97-99 Jacobian, 235
stable, 98 JACOBSON, H., 263
system, 97 JACOBY, S. L. S., 124
unstable, 98 JAIN, A. K., 34-36,41
discharge versus head, 99-102 JAMES, R. E., 33
head versus discharge, 100-102 JKPPSON, R. W., 29, 67, 68, 76, 99, 101,
Head equations (see H equations) 112, 185, 258-261, 273, 388, 423
Head loss JOHNSON, R. L., 263, 279, 280, 291, 301
due to bends and elbows, 67 JOHNSON, S. P., 29
due to entrance, 66
due to exit, 65 KARMAN, 32-33
due to flow meters, 68 Karman number, 29
due to friction, 25-58 Karman-Prandtl equation, 32-33
due to gradual contraction, 66 KEECH, A. E., 308-312
due to gradual enlargement, 60 KENNINGTON, J. L., 99, 312, 318-325
due to obstruction, 68 KHANNA, P., 41
due to sudden contraction, 65 KHARE, A. K., 220-224, 423
due to sudden enlargement, 60 Kinetic energy, 16
due to tees, 67 KING, H. W., 2 9 , 4 2 , 60-67
due to uniformly decreasing KOHI.ER, R. W., 306
discharge, 51-53 KOOTATTEP, S., 227-228
 Index 457

Labeling network elements, 125, 256-261 MCCORMICK, M., 227-228, 235

LAL, B. B., 2 MCILROY, M. S., 263, 303-308
LAM, C. R, 3, 235, 354 MCPHF.RSON, M. B., 303, 310
LAMONT, P. A., 29, 32, 38, 42, 53, 54 MELE, P., 354
LANIER, J. M., 3 MILLS, K. G., 263, 279
Li, W. H., 29 Minimum path length tree, 226-227
LIEBMAN, J., 354 Minimum spanning tree, 225
LlNAWEAVKR, F. P . , J R . , 3 0 8 Minimum number rule, 130-137
Linear theory method, 263 301 Minor appurtenances, 59-68, 124
node head equations, 278-295, 297-301 Minor head loss (see head loss, minor),
networks with known pipe Modified Hazen-Williams formula, 41-42
resistances, 278-286 MOHAN, D. M., 41
networks with pumps, 290-293 Momentum principle, 20-23
networks with unknown pipe Moody diagram, 28-31
resistances, 286-290 MOODY, L. E, 28, 33
networks with valves, 293-295 MUIR, J. R, 263, 264
pipe discharge equations, 264-278, Multiple reservoir system, 94-96
296-297 MURRAY, L., 303
networks with known pipe
resistances, 264-267 NAHAVANDI, A. N.. 34
networks with pumps, 273-276 NAUGLE, R., 303
networks with unknown pipe Network
resistances, 267-273 branching (dead-end), 121-122
networks with valves, 276-278 looped, definition of, 122-123
Link, 119 multisource with known pipe
distribution, 119
resistances, 140, 148-156, 193-195,
p-pipe parallel, 119
211, 235-239, 246-248, 264-267,
p-pipe series, 119
278-286
pseudo, 120
serial, 121, 355-359
real, 120
single source with known pipe
supply, 119
resistances, 140, 142-148, 189-191,
Liu, K. T. H., 187
211, 235-239, 246-248, 264-267,
Loop, 120
278-286
basic (independent, natural,
skeletonization of, 387-388
non-overlapping), 120 with pumps, 168-171, 202-205, 214-217,
overlapping (nonbasic), 120 242-244, 253-254, 273-276, 290-293
pseudo, 120, 261 with unknown pipe resistances, 140,
Loop equations (see AQ equations) 156-168, 195-202, 211-214, 239-242,
Loop head loss relationship, 128-129 248-253, 267-273, 286-290
Loss of energy, 16-17 with valves, 172-181, 206-209, 217-219,
Loss of head, 16-17, 25 244-246, 254-256, 276-278, 293-295
Network analysis
MANNING, R., 43 methods (see analysis methods)
Manning's formula, 42-43, 70, 263 necessity, 2
origin of, 43 Network elements, labeling of (see
Manning's roughness coefficient, 42, 123 labeling network elements)
MARKEL, L. C , 327-337, 406 Network solvability rules, 129-137
MARKS, D. H., 420-421 assignment rule, 130-137
MARLOW, T. A., 263 minimum number rule, 130-137
MARTIN, D. W., 235, 257 path rule, 130-137
MAYS, L. W., 420 unknown number rule, 130-137
458 Index

New England WWA Committee, 54 problem formulation, 368-370

Newton-Cross method, 187 problem solution, 370-383
Newton-Raphson method, 231-261, theory, 368-383
295-301 Node flow continuity relationship, 128,
basic concepts, 231-235 310
multiple variable function, 233-235 Node head analysis, 354, 362-364, 370,
single variable function, 232-233 383
head equations, 235-246 pitfalls in, 418^*19
networks with known pipe
resistances, 235-239 OKUN, D. A., 104, 388
networks with pumps, 242-244 ORBECK, M. J., 43
networks with unknown pipe ORMSBEE, L. J., 393, 405
resistances, 239-242
networks with valves, 244-246 Parameters, 123-129, 141-142
Loop equations, 246-256 basic unknown, 142
networks with known pipe interrelationship, 125-129
resistances, 246-248 states of, 141 -142
networks with pumps, 253-254 PARKER, P. M., 65-66
networks with unknown pipe Path, 120
resistances, 248-253 PATHAK, S. K., 54
networks with valves, 254-256 Path rule, 130-137
NIKURADSE, J., 27 PERRY, H. A., JR., 306
Nodalflow,125 PETERS, G., 235, 257
Nodal head, 125 PlERLUlGI, M . , 2
Node, 119, 366-367 Piezometric head, 17
adequate-flow, 367 Pipe diameter, 123
classification, 366-367 Pipe flow problems, 45-51
flow dependent, 366-367 determination of diameter, 49-51
head dependent, 366 determination of discharge, 48-49
category, 366-367 determination of head loss, 45-47
category compatibility, 367-368 Pipe head loss relationships, 126-127
critical, 366 Pipe length, 123
degree of, 257 Pipe resistance constant, 44, 125
demand (consumption, distribution), Pipe roughness coefficient, 123-124
119 Pipes
end, 119 artificially roughened. 27
ground, 318-325 discharge in, 125
intermediate, 119 equivalent, 69-87
junction, 119 for minor loss elements, 81-86
negative-flow, 367 lengths, 83
no-flow, 366 in parallel, 75-80
partial-flow, 366 head loss known, 77-78
sink, 119 total discharge known, 78-80
source, 119 in series, 69-75
starting, 119 discharge known, 70-73
subcritical, 366 total head loss known, 73-75
supercritical, 366 in series-parallel, 80-81
surplus-flow, 366 modified conductance of, 127, 279
zero-flow, 366 PITCHAI, R., 235
Node flow analysis, 353-387 POLKOWSKI, L. B., 387-388, 390, 406
general observations, 387 POMEROY, R. D., 38
 Index 459

Position head, 17 REID, G. W., 306

Potential energy, 15 Relationship between Newton-Raphson
POWELL, R. W.. 29,42 and linear theory methods, 295-301
PRANOTL, L., 32-33 Reliability analysis. 419-421
Pressure energy, 15 Reservoirs, 89-96
Pressure head, 17 area of, 338
Pressure reducing valve, 112-117, 175-181, balancing (equalizing), 1, 89-90
206-209, 217-219, 244-246, 254-256, distribution (service, storage), 1, 89-90
276-278, 293-295 floating on system, 90, 327
in external pipe, 175-179 impounding, 1, 89
in internal pipe, 179-181 Residue, 235
in several pipes, 181 Reynolds number, 26-36
PRIM, R. C , 225 RODEH. M., 129-130, 235, 248
Principle of continuity, 8-14 ROSARIO, K. E., 54
Principle of energy, 14-20 Roughness, 26, 53, 123-124
Principle of momentum, 20-23 annual growth rate of, 53
Properties of water, 8 coefficient of, 41, 123-124
Pseudo loop, 120, 261 equivalent surface, 26. 28
Pseudo pipe, 120 height, 26
Pumps, 1, 96-110 relative, 26
booster, 96, 171 spacing, 26
characteristic curves, 102-103 ROUSE, H., 2, 29. 60
combination, 103-104 ROWELL, W. F., 354
efficiency, 17-19, 102-103
parallel combination, 103-104
SAATY, T. L., 185
power supplied by, 17, 102-103
SCARBOROUGH, J. B., 231-235
series combination, 103-104
SCHAAKE, J., JR., 419
supply, 96, 168-171
SCOBEY, F. C , 38
Selection of initial values, 224-227, 265
Q equations, 142, 143-145, 150-152, nodal heads, 227
158-161, 169, 172, 174, 177, 179, pipe flows, 225-227, 265
256-261, 263, 264-278, 296-297, pipe resistances, 227
344-346 Sensitivity analysis, 409^118
Q-R equations, 158-161, 267-273 SHAMIR, U , 52, 129, 235, 239, 328, 354,
AQ equations, 146-147, 154-155, 164-166, 386, 391, 409-413, 420
171, 173, 174, 177-178, 180, 235, Shear stress, 6
246-261, 263, 346-347 Similarity in structural and hydraulic
AQ-&R equations, 164-166, 250-253 networks, 280-281
QUINDRY, G., 354 SINGH, K. P., 220-224, 423
SMITH, A. A., 29
RAD/ILL, J. V.. 308 SONAK, V. V., 295-301
RAIIAL, C M . , 387, 392-393, 405 Source node, 119
RAJCOOMAR, R. C. K., 372 Specific gravity, 4
RAMALHO, R. S., 29 Specific mass, 3
RAMAN, V., 54. 307 Specific weight, 3
RAO. H. S., 327-337,406 STAFFORD, J. M., 419
RASMUSF.N, H. J., 354 Standpipe, 90
RAYES, A. G., 409 Stanton diagram (Stanton-Panell diagram),
Reachability, 420 28
Reduction of carrying capacity with age, Static analysis, 327
53-57 STERLING, M. J. H , 387, 392-393, 405
460 Index

STRAUB, L. G., 67 VASUDEO RAO, B., 263, 281, 409

STREETER, V. L., 26, 309, 421-423 Velocity gradient, 6
STRICKLER, A., 43 Velocity head, 17
SURYAPRAKASAM, M. V., 306 VIERUNG, D. E., 306
SWAMEE, P. K., 34-36 Viscosity, 6-7
Swamee-Jain explicit relationship, 34 absolute (coefficient of, dynamic), 6
SYLVESTER, N. D., 35 kinematic, 7
System energy, 313-318 VOYLES, C. F., 227-228

TART, J. S., 328 WAGNER, J. M., 420-421

TECHO, R., 33 WALSKI, T. M., 76, 386-387, 396-399,
THAKRE, C. S., 54 405-408
THIRUVENGDAM, A., 29 WARRING, R. H., 83
Three reservoir system, 92-94 WATANATADA, T., 263
THRUPP, E. C , 43 Water, properties of, 8
TICKNER, R. R., 33 Water supply system, 1
TILLER, F. M., 29 Water uses, 1
Total head, 17 WAITERS, G. Z., 204
Total head line, 20 WEISBACH, J., 26
TRAVALLAE, A., 99 WHITE, C. M., 32, 53
TUNG, Y. K., 199 WHITE, W. R., 38
TUTTON, C. H., 43 WILKE, H. R., 227-228
WILLIAMS, G. N., 235, 228
Unknown number rule, 130-137 WILLIAMS, G. P., 42
Unknown parameters, basic (see basic WILLIAMS, G. S., 38, 54
unknown parameters) WOLFENSON, L. B., 306
Unsteady behaviour during start-up, WOLLA, M. L., 235
308-312 WOOD, D. J., 34, 263-265, 300, 308,
Unsteady flow, 308-312, 421-422 393, 405, 406, 409, 423
Unsteady flow analysis, 421-422 WOODWARD, S. M.,43
WYLIE, E. B., 26, 309, 421-423
VALOT, H. 43
Valves, 110-117 YARNELL, D. L., 43
check (see check valves)
foot, 110 ZlGRANG, D. J., 35
pressure reducing (see pressure Zigrang-Sylvester explicit relationship,
reducing valves) 35,71
VAN DER BERG, 228 ZIMKE, P. C , 83
 About the Author

RAMOD R. Bhave obtained his B.E. (Civil) from Poona University,

P Pune, Maharashtra in 1953, M.E. (Civil) from Maharaja Sayajirao
University of Baroda, Vadodara, Gujarat in 1961 and Ph.D. from Nagpur
University, Maharashtra in 1978. After serving with the state government
for a couple of years, Bhave joined the faculty of technology and engineer-
ing at the Maharaja Savajirao University of Baroda as a lecturer in 1956
and served there until 1962 when he joined Visvesvaraya Regional College
of Engineering, Nagpur, as Assistant Professor of Civil Engineering. He
became a Professor of Civil Engineering in 1974. He is a member of the
Institution of Engineers (India) and the Indian Society for Technical Edu-
cation, and a Fellow of the Indian Water Works Association.
He has published papers in several journals, including those of Indian
Waterworks Association, Institution of Engineers (India), European Jour-
nal of Operations Research, and the American Society of Civil Engineers.
He has contributed two chapters to the five-volume Civil Engineering
Practice published by Technomic Publishing Co. He has written a book on
technical writing and coauthored, Fluid Mechanics and Hydraulics, both
written in Marathi, the state language of Maharashtra. He has received the
Maharashtra State Award, best-paper awards of the Indian Water Works
Association in 1978, 1984, and 1985; the Sir Arthur Cotton Memorial Gold
Medal and the Shrimati Saroma Sanyal Memorial Prizes of the Institution
of Engineers, India. His biographical sketch is included in the 9th edition
(1989-90) of Marquis' Who's Who in the World.

461