This action might not be possible to undo. Are you sure you want to continue?
:
Seismic Analysis, Design and Retriﬁtting
M. Ghaemian
April 2, 2000
2
Contents
1 INTRODUCTION 9
1.1 TYPES OF DAMS . . . . . . . . . . . . . . . . . . . . . . . . 10
1.1.1 Embankment Dams . . . . . . . . . . . . . . . . . . . . 11
1.1.2 Concrete Arch and Dome Dams . . . . . . . . . . . . . 13
1.1.3 Concrete Gravity and GravityArch Dams . . . . . . . 14
1.1.4 Concrete Slab and Buttress Dams . . . . . . . . . . . . 16
1.2 APPURTENANT FEATURES OF DAMS . . . . . . . . . . . 16
1.3 SAFETY OF DAMS AND RESERVOIRS . . . . . . . . . . . 18
1.4 HOW DAMS ARE BUILT . . . . . . . . . . . . . . . . . . . . 20
1.5 FAMOUS DAMS OF THE WORLD . . . . . . . . . . . . . . 21
1.6 POWERGENERATOR, FLOODCONTROL ANDIRRIGA
TION DAMS . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.1 Power Generator Dams . . . . . . . . . . . . . . . . . . 25
1.6.2 Flood Control Dams . . . . . . . . . . . . . . . . . . . 26
1.6.3 Irrigation Dams . . . . . . . . . . . . . . . . . . . . . . 28
1.7 INSTRUMENTATIONS AND SURVEILLANCE OF DAMS . 28
1.7.1 Surveillance . . . . . . . . . . . . . . . . . . . . . . . . 28
1.7.2 Instrumentation . . . . . . . . . . . . . . . . . . . . . . 28
1.7.3 Instruments . . . . . . . . . . . . . . . . . . . . . . . . 30
1.8 ECOLOGICAL/ENVIRONMENTAL CONSIDERATIONOF
DAM OPERATION . . . . . . . . . . . . . . . . . . . . . . . 33
1.9 THE HISTORY OF DAMS DESIGN . . . . . . . . . . . . . . 34
1.9.1 Irrigation Dams . . . . . . . . . . . . . . . . . . . . . . 35
1.9.2 Dams Designed for Water Supply . . . . . . . . . . . . 36
1.9.3 Flood Control Dams . . . . . . . . . . . . . . . . . . . 38
1.9.4 Power Dams . . . . . . . . . . . . . . . . . . . . . . . . 39
1.9.5 The Moslem World . . . . . . . . . . . . . . . . . . . . 40
3
4 CONTENTS
1.9.6 Development of the Modern Dams . . . . . . . . . . . 42
1.10 BEAVERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2 RESERVOIR 49
2.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . 49
2.2 GENERAL FORM OF RESERVOIR’S EQUATION OF MO
TION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.2.1 Velocity Field . . . . . . . . . . . . . . . . . . . . . . . 50
2.2.2 System and Control Volume . . . . . . . . . . . . . . . 52
2.2.3 Reynold’s Transport Equation . . . . . . . . . . . . . . 53
2.2.4 Continuity Equation . . . . . . . . . . . . . . . . . . . 56
2.2.5 Linear Momentum Equation . . . . . . . . . . . . . . . 57
2.2.6 The Equation of the Motion . . . . . . . . . . . . . . . 59
2.3 VISCOSITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.4 NAVIERSTOKES AND EULER EQUATIONS . . . . . . . . 65
2.5 COMPRESSIBLE FLUID . . . . . . . . . . . . . . . . . . . . 67
2.6 BOUNDARYLAYER THEORY . . . . . . . . . . . . . . . . . 71
2.7 IRROTATIONAL FLOW . . . . . . . . . . . . . . . . . . . . 74
2.8 RESERVOIR’S EQUATION OF MOTION . . . . . . . . . . . 78
2.9 RESERVOIR BOUNDARY CONDITIONS . . . . . . . . . . . 79
2.9.1 DamReservoir Boundary Condition . . . . . . . . . . . 80
2.9.2 ReservoirFoundation Boundary Condition . . . . . . . 81
2.9.3 Free Surface Boundary Condition . . . . . . . . . . . . 84
2.10 SOLUTION OF THE RESERVOIR EQUATION . . . . . . . 86
2.11 RESERVOIR FAREND TRUNCATED BOUNDARY CON
DITION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3 FINITEELEMENTMODELLINGOF THEDAMRESERVOIR
SYSTEM 99
3.1 FINITE ELEMENT MODELLING OF THE STRUCTURE . 99
3.1.1 SingleDegreeOfFreedom Systems . . . . . . . . . . . 99
3.1.2 MultiDegreeOfFreedom System . . . . . . . . . . . . 101
3.2 COUPLING MATRIX OF THE DAMRESERVOIR . . . . . 106
3.3 FINITE ELEMENT MODELLING OF THE RESERVOIR . . 108
3.3.1 Truncated Boundary of the Reservoir’s FarEnd . . . . 110
3.4 EQUATION OF THE COUPLED DAMRESERVOIR SYS
TEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
CONTENTS 5
4 DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM 113
4.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . 113
4.2 THE COUPLED DAMRESERVOIR PROBLEM . . . . . . . 115
4.3 DIRECT INTEGRATION OF THE EQUATION OF MOTION115
4.4 USINGNEWMARKβ METHODFORTHECOUPLEDEQUA
TIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.5 STAGGERED DISPLACEMENT METHOD . . . . . . . . . . 118
4.5.1 Stability of the Staggered Displacement Method . . . . 119
4.6 STAGGERED PRESSURE METHOD . . . . . . . . . . . . . 121
4.6.1 Stability of the Staggered Pressure Method . . . . . . . 122
4.7 MODIFIED STAGGERED PRESSURE METHOD . . . . . . 123
4.8 USING αMETHOD FOR THE COUPLED EQUATIONS . . 124
4.8.1 Staggered Displacement Method . . . . . . . . . . . . . 124
4.9 SEISMIC ENERGY BALANCE . . . . . . . . . . . . . . . . . 126
4.10 ACCURACY OF THE SOLUTION SCHEME . . . . . . . . . 127
5 NONLINEARFRACTUREMODELS OF CONCRETEGRAV
ITY DAMS 129
5.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . 129
5.2 A BRIEF STUDY OF NONLINEAR PARAMETERS . . . . 134
5.2.1 Finite element models of crack propagation . . . . . . . 134
5.2.2 Discrete crack propagation model ,DCPM,( variable
mesh) . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.2.3 Continuum crack propagation models (CCPM) . . . . . 135
5.3 Constitutive models for crack propagation . . . . . . . . . . . 136
5.3.1 Strengthbased criteria . . . . . . . . . . . . . . . . . . 136
5.3.2 Fracture mechanics criteria . . . . . . . . . . . . . . . . 137
5.3.3 Shear resistance of fractured concrete . . . . . . . . . . 148
5.4 Postfracture behaviour of concrete . . . . . . . . . . . . . . . 149
5.5 Material parameters for fracture propagation analysis . . . . . 150
5.5.1 Strengthofmaterial parameters . . . . . . . . . . . . . 151
5.5.2 Linear elastic fracture mechanics parameters . . . . . . 153
5.5.3 Nonlinear fracture mechanics parameters . . . . . . . . 154
5.5.4 Shear resistance of fractured concrete . . . . . . . . . . 155
5.6 NONLINEAR MODELLING OF CONCRETE DAMS US
ING DAMAGE MECHANICS . . . . . . . . . . . . . . . . . . 156
5.6.1 NUMERICAL PROBLEMS RELATED TO STRAIN
SOFTENING . . . . . . . . . . . . . . . . . . . . . . . 157
6 CONTENTS
5.6.2 FUNDAMENTAL EQUATIONS OF DAMAGE ME
CHANICS . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.6.3 ISOTROPIC DAMAGE MODEL FOR CONCRETE . 159
5.6.4 ANISOTROPIC DAMAGE MODEL FOR CONCRETE160
5.6.5 EVALUATION OF DAMAGE VARIABLE . . . . . . . 164
5.6.6 Damage evolution for concrete subjected to tensile strain166
5.6.7 Opening and closing of the crack and initial damage . . 168
5.6.8 ANALYTICAL PROCEDURES IN A FINITE ELE
MENT MODEL . . . . . . . . . . . . . . . . . . . . . 168
5.7 CONSTITUTIVEMODEL FORSMEAREDFRACTUREANALY
SIS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.7.1 Prefracture behaviour . . . . . . . . . . . . . . . . . . 170
5.7.2 Strain softening of concrete and the initiation criterion 170
5.7.3 Fracture energy conservation . . . . . . . . . . . . . . . 172
5.7.4 Constitutive relationships during softening . . . . . . . 173
5.7.5 Coaxial Rotating Crack Model (CRCM) . . . . . . . . 173
5.7.6 Fixed Crack Model With Variable Shear Resistance
Factor (FCMVSRF) . . . . . . . . . . . . . . . . . . . 174
5.7.7 Closing and reopening of cracks . . . . . . . . . . . . . 175
List of Figures
1.1 Idealized section of embankment dams a) Rockﬁll dam with
symmetrical clay core b) Rock and gravel dam with reinforced
concrete slab . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2 Crosssections of several arch dams . . . . . . . . . . . . . . . 14
1.3 Crosssection of typical concrete gravity dam . . . . . . . . . . 15
1.4 Crosssection of a concrete buttress dam . . . . . . . . . . . . 16
2.1 Fluid point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.2 a)Lagrangian viewpoint b)Eulerian viewpoint . . . . . . . . . 52
2.3 Moving system . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.4 Volume element . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.5 Rectangular Parallelpiped element . . . . . . . . . . . . . . . . 61
2.6 Wellordered parallel ﬂow . . . . . . . . . . . . . . . . . . . . 64
2.7 Wave front movement and ﬂuid movement . . . . . . . . . . . 68
2.8 Moving pressure disturbance in a motionless ﬂuid and ﬁxed
wave in a moving ﬂuid . . . . . . . . . . . . . . . . . . . . . . 69
2.9 Details of boundary layers . . . . . . . . . . . . . . . . . . . . 72
2.10 Displacement thickness in boundary layer . . . . . . . . . . . . 73
2.11 Three types of ﬂuid motion . . . . . . . . . . . . . . . . . . . 74
2.12 Fluid rotating like a rigid body . . . . . . . . . . . . . . . . . 75
2.13 Shearing ﬂow between two ﬂat plates . . . . . . . . . . . . . . 75
2.14 Change of relative positions in an arbitrary ﬂow ﬁeld . . . . . 76
2.15 Boundaries of the damreservoir system . . . . . . . . . . . . . 80
2.16 Damreservoir interface . . . . . . . . . . . . . . . . . . . . . . 81
2.17 Free surface wave . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.18 Rigid daminﬁnite reservoir system . . . . . . . . . . . . . . . 87
2.19 Added mass approach . . . . . . . . . . . . . . . . . . . . . . 89
3.1 Systems of single degree of freedom . . . . . . . . . . . . . . . 100
7
8 LIST OF FIGURES
3.2 Forces on a single degree of freedom . . . . . . . . . . . . . . . 101
3.3 An example of multidegreeoffreedom (MDF) sytem with de
grees of freedom in y direction . . . . . . . . . . . . . . . . . . 101
3.4 An example of MDF system with two degrees of freedom at
each mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.5 Interface element on the damreservoir interaction boundary . 108
5.1 Modes of failure: (a) mode I  Tensile fracture; (b) mode II 
planar shear fracture; (c) mode III  tearing fracture . . . . . . 137
5.2 Fracture process zone (FPZ); (a) LEFM; (b) NLFM . . . . . . 139
5.3 Nonlinear fracture mechanics models: (a,b) ﬁctitious crack
model,(c,d) crack band model . . . . . . . . . . . . . . . . . . 141
5.4 (a) average stressstrain curve for smeared crack element; (b)
characteristic dimension, l
c
= l
1
, l
2
; (c) characteristic dimen
sion. lc=
√
l
0
l
00
. . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.5 Nonlinear fracture mechanics in smeared crack propagation
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.6 Closing and reopening of partially formed cracks . . . . . . . . 147
5.7 Strengthofmaterialbased failure criterion . . . . . . . . . . . 152
5.8 Material model in the damage mechanics concept; A)eﬀective
areas for isotropic and anisotropic damages; B)characteristic
length; C)strain equivalence hypothesis; D)stressstrain curve
for equivaalence hypothesis; E)closingopening criterion; F)initial
damage formulation . . . . . . . . . . . . . . . . . . . . . . . . 161
5.9 Stressstrain curve for energy dissipation due to fracture . . . 165
5.10 Constitutive modelling for smeared fracture analysis; a)softening
initiation criterion; b)fracture energy conservation; e)local axis
system; d)closing and reopening of cracks . . . . . . . . . . . 171
Chapter 1
INTRODUCTION
People from the beginning of recorded history have constructed barriers
across rivers and other water courses to store or divert water. The earliest
of these dams were used to water farms. For example, the ancient Egyptians
built earth dams that raised the river level and diverted water into canals to
irrigate ﬁelds above the river. Behind the dam, waters pile up to form an
artiﬁcial lake which sometime can be very long. The artiﬁcial lake backed up
by a dam is called a reservoir.
Dams are built primarily for irrigation, water supply, ﬂood control, elec
tric power, recreation, and improvement of navigation. Many modern dams
are multipurpose. Irrigation dams store water to equalize the water supply
for crops throughout the year. Water supply dams collect water for domes
tic, industrial, and municipal uses for cities without suitable lakes or rivers
nearby for a water supply. Flood control dams impound ﬂoodwaters of rivers
and release them under control to the river below the dam. Hydroelectric
power dams are built to generate electric power by directing water in pen
stocks through turbines, wheels with curved blades as spokes. The falling
water spins the blades of the turbines connected to generators. Power dams
are expected to generate power to repay the cost of construction. The out
put depends, ﬁrst, upon the head of water, or height of stored water above
the turbines. The higher the water the more weight and pressure bear upon
the turbine blades. A second factor is the volume of water throughout the
year. The minimum ﬂow in dry months ﬁxes the amount of ﬁrm power
which customers can rely upon to receive regularly. Sometimes extra power,
or runofstream power, generated in ﬂood seasons can be sold, usually at
lower rates.
9
10 CHAPTER 1. INTRODUCTION
Dams also provide beneﬁts other than those mentioned above. Their
reservoirs provide recreation, such as ﬁshing and swimming. They become
refuges for ﬁsh and birds. Dams conserve soil by preventing erosion. They
slow down streams so that the water does not carry away soil. Dams can
also create problems. Their reservoirs may cover towns or historic and scenic
places. Dams may impair ﬁshing. Another problem of dams is silting. Some
rivers pick up clay and sand and deposit them behind the dam, thereby
lessening its usefulness.
1.1 TYPES OF DAMS
Dams range in size and complexity of construction from low earth embank
ment constructed to impound or divert water in small streams to massive
earth or concrete dams built across major rivers to store water. The type of
dam that is built and its size are a complex function of a demonstrated neces
sity for water storage or diversion, the amount of water available, topography,
geology, and kinds and amount of local materials for construction. Although
large embankment dams do not posses the graceful and architecturally at
tractive conﬁgurations of many concrete dams, they commonly require an
equal amount of engineering skill in planning, design, and construction. The
world’s largest dams, as measured by the volumes of materials used in their
construction, are embankment dams. In contrast, many of the world’s high
est dams are built of concrete, and many of them are 180 m (600 ft) or more
high.
There are several basic types of dams. Diﬀerences depend on their geo
metric conﬁgurations and the material of which they are constructed. Under
special circumstances, feature of the basic types are combined within a par
ticular dam to meet unusual design requirement. The followings are the main
types of dams:
1. Embankment dams.
a. Homogenous dams, constructed entirely from a more or less uniform
natural material.
b. Zoned dams, containing materials of distinctly diﬀerent properties
in various portions of dams.
2. Concrete arch and dome dams.
a. Single arch and dome dams.
b. Multiplearch and multipledome dams.
1.1. TYPES OF DAMS 11
3. Concrete gravity and gravityarch dams.
4. Concrete slab and buttress dams.
5. Dams combining two or more basic characteristics of the above basic
types.
1.1.1 Embankment Dams
A broad spectrum of natural and fabricated materials have been used in the
construction of embankment dams. Embankment dams are made by building
an embankment of gravel, sand, and clay across a river. To prevent leakage,
often a core, or inner wall, of concrete or other watertight materials is used.
In a rolledﬁll dam, earth is hauled by vehicles onto the dam and rolled tight
with heavy machinery. In the hydraulic ﬁll dam, earth is carried to the dam
by water in pipes or ﬂumes and also deposited by the water. The placing of
the earth is so controlled that the ﬁner, watertight materials form the core.
In the semihydraulic ﬁll dam, trucks bring the earth to the dam and jets
of water distribute the materials. Rock dams are made by dumping rocks
across the river. A wall of rocks is then laid on the upstream side and over
this is built a waterproof facing of reinforced concrete, timber, or steel.
Controlling factors in choosing this type of dam are the amounts and
types of materials locally available for construction and the size and con
ﬁguration of the dam. Many small embankment dams are built entirely of
a single type of material such as stream alluvium, weathered bedrock, or
glacial till. Larger embankment dams generally are zoned and constructed
of a variety of materials, either extracted from diﬀerent local sources, or
prepared by mechanical or hydraulic separation of a source material into
fractions with diﬀerent properties. Where rock is used extensively, it may be
obtained by separation from bowldery stream deposits, glacial till, siderock
accumulations, or by quarrying.
Construction of an embankment dam requires prior investigation of foun
dation geology and an inventory and soilmechanics study of materials avail
able for emplacement in the embankment. An important element in a zoned
dam is an impermeable blanket or core which usually consists of clayey ma
terials, obtained locally. In the absence of such materials, the dam is built of
quarried rock or unsorted pebbly or bowldery deposits, and the impermeable
core is constructed of ordinary concrete or asphaltic concrete. Alternatively,
in locations where natural impermeable materials are unavailable, embank
ment dams are built of rock or earthrock aggregates and impermeable layers
12 CHAPTER 1. INTRODUCTION
Figure 1.1: Idealized section of embankment dams a) Rockﬁll dam with
symmetrical clay core b) Rock and gravel dam with reinforced concrete slab
of reinforced concrete, asphaltic concrete, or riveted sheet steel are placed on
the upstream face of the dam. Control of seepage through the dam or under
it commonly requires installation of porous materials within or immediately
beneath the dam.
Embankment dams have been built on a great variety of foundations,
ranging fromweak, unconsolidated streamor glacial deposits to highstrength
sedimentary rocks and crystalline igneous and metamorphic rocks. A partic
ular advantage of an embankment dam, as compared with a concrete dam, is
that the bearingstrength requirements of the foundation are much less. Mi
nor settlement of an embankment dam owing to load stresses during and after
construction generally is not a serious matter because of the ability of the
embankment to adjust to small dislocations without failure. Crosssections
of selected examples of embankment dams are shown in ﬁgure 1.1.
1.1. TYPES OF DAMS 13
1.1.2 Concrete Arch and Dome Dams
The ultimate complexity of design and analysis of stresses is attained in
arch and dome dams. These dams are thin, curved structures commonly
containing reinforcing, either steel rods or prestressed steel cables. Volume
requirements for aggregate for manufacture of concrete are much less than in
gravity and gravityarch dams, but the competency of bedrock in foundations
and abutment to sustain or resist loads must be of a high order. Arch dams
usually are built in narrow, deep gorges in mountainous regions where access
and availability of construction materials pose especially acute problems. At
sites where abutments are not entirely satisfactory rock may be excavated
and replaced with concrete to form artiﬁcial abutments. Their height can be
as high as 272 m (905 ft).
Arch dams are of two kinds. Constantradius arch dams commonly have
a vertical upstream face with a constant radius of curvature. Variableradius
dams have upstream and downstream curves (extrados and intrados curves)
of systematically decreasing radii with depth below the crest When a dam
is also doubly curved, that is, it is curved in both horizontal and vertical
planes, it is sometimes called a ”dome” dam. Curves that have been used in
construction of arch or dome dams are arcs or sectors of circles, ellipses, or
parabola. Some dams are constructed with two or several contiguous arches
or domes, and are then described as multiple arch or multipledome dams.
Figure 1.2 shows Crosssection of several varieties of arch dams.
Engineering analysis of arch and dome dams assumes that two major
types of deﬂections or dislocations aﬀect the damand its abutments. Pressure
of water on the upstream face of the dam and, in some instances, uplift
pressures from seepage beneath the dam, tend to rotate the dam about its
base by cantilever action. In addition, the pressure of reservoir water tends
to ﬂatten the arch and push it downstream, so that stresses are created which
act horizontally within the dam toward the abutments. That portion of the
bedrock abutment which receives the thrust from the load of reservoir water
either by a tendency for downstream movement of the dam or ﬂattening of
the arch is called the thrust block and must be suﬃciently strong to resist the
forces acting on it without failure or appreciable, dislocation. Simply stated,
an arch dam utilizes the strength of an arch to resist the loads placed upon it
by the familiar ”arch action”. It is clear that the foundation and abutments
must be competent not only to support the dead weight of the dam on the
foundation but also the forces that are directed into the abutments because of
14 CHAPTER 1. INTRODUCTION
Figure 1.2: Crosssections of several arch dams
arch action in response to loads created by impounded reservoir water, and,
in areas of cold climates, pressures exerted by ice forming on the reservoir
surface. In regions of seismic activity consideration must also be given to the
interaction of the dam and pulses of energy associated with earthquakes.
1.1.3 Concrete Gravity and GravityArch Dams
A concrete gravity dam has a crosssection such that, with a ﬂat bottom,
the dam is freestanding; that is, the dam has a center of gravity low enough
that the dam will not topple if unsupported at the abutments. Gravity dams
require maximum amounts of concrete for their construction as compared
with other types of concrete dams, and resist dislocation by the hydrostatic
pressure of reservoir water by sheer weight. Concrete gravity dams have
been constructed up to 285 m (950 ft)high. Properly constructed gravity
dams with adequate foundation probably are among the safest of all dams
and least susceptible to failure with time. They withstand the pressure or
push of water by their weight. In cross section, they are like a triangle, broad
at the base and narrow at the crest. They are built in this shape because
water pressure becomes greater with the depth of water. A typical gravity
dam is shown in ﬁgure 1.3.
Final selection of the site for a gravity or gravityarch dam is made only
after comprehensive investigation of hydrologic, topographic, and, especially,
1.1. TYPES OF DAMS 15
Figure 1.3: Crosssection of typical concrete gravity dam
subsurface geologic conditions. A favorable site usually is one in a constric
tion in a valley where sound bedrock is reasonably close to the surface both
in the ﬂoor and abutments of the dam.
An important consideration in construction of a concrete gravity or gravity
arch dam is the availability, within a reasonable hauling distance, of adequate
deposit of aggregate suitable for manufacture of concrete, whether the ag
gregate is obtained from unconsolidated deposits or is quarried.
The simplest form of a gravity dam is one in which the top or crest
is straight. depending on the topographic conﬁguration of a valley and the
foundation geology, it may be possible to construct a gravityarch dam which
incorporates the advantages of mass weight and low center of gravity of a
gravity dam with those inherent in an arch dam. in gravityarch dams the
requirements for sound rock in abutments are somewhat more stringent than
in simple gravity dams.
16 CHAPTER 1. INTRODUCTION
Figure 1.4: Crosssection of a concrete buttress dam
1.1.4 Concrete Slab and Buttress Dams
In locations where aggregate for concrete or for earthﬁll is in limited supply
and the foundation rocks are moderately to highly competent, buttress dams
provide a possible alternative to other kinds of dams. They are up to 130
m (430 ft) high. In crosssection, buttress dams resemble gravity dams, but
with ﬂatter upstream slopes. In a buttress dam a slab of reinforced concrete
upstream rest on a succession of upright buttress which have thicknesses and
a spacing suﬃcient to support the concrete slab and the load of the water
in the reservoir that exert on the slab. A crosssection of a typical buttress
dam is shown in ﬁgure 1.4
1.2 APPURTENANTFEATURES OF DAMS
Appurtenances are structures and equipment on a project site, other than
the dam itself. They include, but are not limited to, such facilities as intake
towers, powerhouse structures, tunnels, canals, penstocks, lowlevel outlets,
surge tanks and towers, gate hoist mechanisms and their supporting struc
1.2. APPURTENANT FEATURES OF DAMS 17
tures, and all critical water control and release facilities. Also included are
mechanical and electrical control and standby power supply equipment lo
cated in the powerhouse or in the remote control centres.
In previous sections of this chapter the characteristics and conﬁgurations
of basic types of dams have been outlined, but no consideration was given to
the various appurtenant features that enable use of a dam and the reservoir
behind it for their intended purposes. In this section mention is made of
various features that are incorporated into the designs of dams for control
of ﬂow of water impounded in the reservoir through or outside of a dam.
Design of many of these features requires intensive prior investigations of
hydrology, topography, and subsurface geology of the dam site. Following is
a tabulation and description of several kinds of appurtenant features included
in the construction of dams.
Coﬀer dams usually are temporary structures built upstream from a
dam to divert stream ﬂow around the excavation for a dam. In valleys of
steep proﬁle diversion commonly is accomplished by a tunnel or tunnels in
the walls of the valley. Commonly the diversion tunnels are put to further
use to control ﬂow from reservoir either for drainage of the reservoir or for
ﬂow under pressure into a hydroelectric generating plant. In valleys of low
proﬁle diversion is by tunnels, canals, or by conduits which subsequently
are buried by the dam. It is not unusual in embankment dams to incorporate
the coﬀer dam into the larger embankment structure comprising the designed
dam.
Tunnels in bedrock outside of dams serve a variety of purposes. Flow
through them is controlled by valves external to the dam or in valve chambers
or vaults within the damor in bedrock outside of the dam. Tunnels for control
of the water level in the reservoir are commonly called gravity tunnel and
serve a principal function in diverting water to some point downstream from
the dam. Tunnels that transmit water under pressure to elevate the water to
a higher level than the intake of the tunnel or to generate hydroelectric power
are called pressure tunnel and usually require considerable competency in the
rock through which they are constructed . Valves control the ﬂow of water
through tunnels and penstocks. In many large dams the valves are installed
in underground vaults or chambers to which access is gained downstream
from the dam.
Many dams are constructed to generate hydroelectric power. The pow
erhouse is located at, or in the vicinity of, the toe of a dam or at some
distance downstream. Flow of water into the power house is controlled by
18 CHAPTER 1. INTRODUCTION
valves upstream from the dam, within the dam downstream, or in valve
vaults excavated in rock outside of the dam.
A sluice is a passage through the dam itself for lowering the water level
of the reservoir. Pipes for conducting water to the power turbines are called
penstocks. The ﬂow of water from intake towers through spillways, sluices,
and penstocks is regulated by control gates.
A Spillway is designed to contain and control overﬂow of reservoir water
when the reservoir is full. Spillways are, or should be, designed to accommo
date ﬂows during maximum ﬂood stage so as to prevent damage to the dam
and appurtenant features. Their size and location with respect to the dam
is determined by the size and type of dam, local topography geology and a
careful review of the history of stream ﬂow at the site of the dam. Water
may pass over the crest of the dam itself, or near the dam in chutes, tun
nels, or shafts. Overﬂow of embankment dams outside of a spillway can have
especially disastrous consequences so that safety usually requires a spillway
capable of containing at least a hundredyear ﬂood. Spillways are located
within or on the downstream face of a dam outside of the dam on one side or
the other, or within the reservoir, where water spills into a ”glory hole” and
passes through a shaft and tunnel or tunnels in the abutment of the dam.
Gates are devices installed in the tops of spillways to control the ﬂow of
water over the spillway.
Levees are artiﬁcial riverbanks constructed high enough to prevent ﬂood
ing. A dam across a river intended to permit ﬂow, once a certain depth of
water has been reached, may be called a barrage. A small dam that forms
a millpond or ﬁshpond is called a weir.
1.3 SAFETYOF DAMS ANDRESERVOIRS
There are Several types of reservoirs as deﬁned by their locations. The com
mon type is a reservoir behind a dam in a valley. Increasing use is being made
of tidalstorage reservoir for power generation along coast lines and excavated
reservoirs for water storage for municipal or other use. In the latter type the
excavated material commonly is employed to construct an embankment on
one or all sides of the reservoir.
Reservoirs and their associated dams serve many purposes including elec
tric power generation, storage and diversion of irrigation water, storage of
industrial and municipal water supplies, recreation, and ﬂood control. Less
1.3. SAFETY OF DAMS AND RESERVOIRS 19
frequent uses of reservoirs include storage and control of stream water for
navigation, and storage of sewage and waste products from mining or manu
facturing operations. In some stances fuel powered and thermonuclear power
plants require large volumes of cooling water, and reservoirs are constructed
for this purpose. Whether a dam backs up water for a long or a short dis
tance is not important from the structural point of view. Pressure depends
not upon how far water is backed upstream but upon its depth at the dam.
But the length of the reservoir and its capacity may change the risk associated
in the case of failure.
There are no accurate records of the number of dams that have failed
throughout the history of their construction. However, ruins of dams built
through a long span of history in both ancient and more recent times indicates
that the number must be high, probably in the thousands, if dams of all
heights are included in the count. The magnitude of ﬂoods generated by
dam failure or by collapse of the walls of a reservoir are not related to the
height of a dam. More pertinent is the volume of water stored in the reservoir
behind the dam, and the conﬁguration of the valley below it, wether the dam
is low or high.
The cost in human life, goods, and property damage of a ﬂood generated
by breaching of a dam or collapse of reservoir wall depends to a large extent
on the magnitude of ﬂood and what lies in the pathway of the ﬂood. With
an expanding population in many part of the world and an increasing occu
pation of ﬂoodplains by dwellings, commercials and industrial facilities , and
highways, dam built long ago and dam built in recent times present a grow
ing potential for massive destruction of life and property. the cause of ﬂoods
associated with failure of dams and reservoir are numerous. In embankment
dams a common cause of breaching is overtopping of existing spillways or
waterlevel control facilities, although some failures have been attributed to
slope failure, foundation subsidence, or earthquake damage. Failure of con
crete dams usually is attribute to imperfect design or construction, to use of
inferior materials in the dam, or to failure of foundation and/or abutment
rocks. Construction of a dam and reservoir imposes new loads on foundation
materials. Initial adjustment in the dam and foundation occurs as the dam is
being built and as the load on the foundation is increased to a ﬁnal load equal
to the weight of the dam. Filling of the reservoir impose additional loads not
only on the ﬂoor and wall of the reservoir but also on the upstream face of the
dam. As the reservoir level is decreased and increased these loads ﬂuctuate ,
and a cycle dynamic system of changing loads is superimposed on the static
20 CHAPTER 1. INTRODUCTION
load of the dam on its foundation. Seepage of water through or beneath the
dam may produce slow deterioration that may promote eventual failure. The
responsibility for the safety of dams and reservoirs no longer belongs only to
the designer and builder but must be shared by those who have knowledge
and understanding, however imperfect they may be, of the expected behav
ior of natural materials under the condition superimposed by the loads of
dams and reservoirs. Modern technology and social responsibility require
that safe construction and maintenance of dams and reservoir shall be the
shared responsibility of engineers, geologists, and rock mechanical expert.
Recognition of the need for worldwide surveillance of dams and reservoirs
with emphasis on their safety has resulted in the formation of the Inter
national Commission on Large Dams (lCOLD) a unit of the World Power
Conference within individual countries increasing eﬀorts are being made to
regulate and maintain continued safety of dams and reservoirs through the
close cooperation of engineering organizations and government agencies. In
spite of these good works, dams continue to fail, and intensiﬁcation of eﬀorts
to assure the safety of existing dams and reservoirs and those that will be in
the future is an increasingly urgent necessity.
1.4 HOW DAMS ARE BUILT
The methods of building dams can be envisioned by following the construc
tion of Hoover Dam, built between 1930 and 1935. The engineers constructed
a concrete archgravity dam at an approved cost of $174,000,000. It is as tall
as a 60story skyscraper. Its crest is 45 feet (14 meters) thick and its base,
660 feet (201 meters). It stores the entire ﬂow of the Colorado River for two
years. Much preliminary work had to be done. The engineers made geologic
and topographic surveys to select the site. They made maps of 70 locations,
bored holes to test the rock for a sound foundation, and studied the river’s
speed, high water level, and silting. Once the location was chosen, designers
made their plans. They then made models to test their design. Where once
had been burning desert, engineers built Boulder City to house about 5,000
workers. Construction gangs built railroads and highways for transporting
great quantities of equipment and materials. Workmen strung cables across
the canyon from pairs of towers, which travel on tracks along opposite sides
of the site. Each of the ﬁve cableways could carry 25 tons (22,680 kilograms).
Two of them had spans of nearly half a mile. Construction crews also built
1.5. FAMOUS DAMS OF THE WORLD 21
a great gravel screening plant and two huge concrete mixing plants.
On each side of the river two tunnels, each 50 feet (15 meters) in di
ameter, were drilled and blasted from the rock of the canyon walls. These
tunnels were used to divert the river around the site. When construction
was completed the tunnels served as spillway outlets and penstocks for the
power plant. Next, coﬀerdams of earth and rock were built upstream and
downstream from the dam site to block the river. ”High scalers” stripped
tons of loose and projecting rock from canyon walls. The overburden, or
loose rock and muck, was dug out to expose the bedrock. Grout, a thin mor
tar of cement and water, was next forced into the foundation to ﬁll seams
and holes. Forms were made. These would be used for building the dam in
enormous blocks. Concrete was poured into the forms from eightcubicyard
buckets traveling on the cableways. As each block of concrete dried, grout
was pumped between the blocks, making the dam into one solid piece. Al
lowing so gigantic a structure to cool naturally would have taken a century
because of the heat given oﬀ by the setting cement. In addition the concrete
would have shrunk and cracked. Cold water circulating through 528 miles
(850 kilometers) of oneinch (25mm) pipes embedded in the concrete carried
oﬀ the heat. Refrigerating pipes were also used to freeze landslides of wet
earth at Grand Coulee Dam. Another problem was the freezing of control
gates in winter. A huge electrical heating apparatus was installed in the
spillway and gates.
1.5 FAMOUS DAMS OF THE WORLD
Rogun and Nurek, in Tajikistan, are among the world’s highest dams. They
rise to a height of more than 1,000 feet (300 meters) each. Some of the
world’s other major dams are Grande Dixence and Mauvoisin, in Switzerland;
Chicoasen, in Mexico; Inguri, in Georgia; SayanoShushensk, in Russia; Mica,
in Canada; and Guavio, in Colombia. Three of the foreign dams that are
constructed of the largest volume of materials are Tarbela, in Pakistan; Lower
Usuma, in Nigeria; and Guri, in Venezuela. Dams that form the largest
reservoirs in capacity are Owen Falls (Lake Victoria), in Uganda; Bratsk, in
Russia; Kariba Gorge, bordering Zambia and Zimbabwe; and Aswan High, in
Egypt. One of the oldest great dams is Egypt’s Aswan for irrigation on the
Nile, ﬁnished in 1902. The Dnepr, completed in 1932, was one of the earliest
major Soviet dams. Another historical dam is Mettur, on the Cauvery River
22 CHAPTER 1. INTRODUCTION
in India. The world’s ﬁrst dam to furnish hydroelectric power from ocean
tides was completed in 1966. It is located on the Rance River, near StMalo,
in France.
The following tables showsome of the famous dams of the world with their
speciﬁcations. In the tables the followings notation was used for simplicity.
HHeight (ft);
LLength (ft);
Mmaterial in dam (cu yd);
RCReservoir Capacity (billions of gal)
A—Arch;
B—Slab and buttress;
C—Concrete;
E—Rolled earth ﬁll;
G—Gravity;
H—Hydraulic earth ﬁll;
M—Multiple arch;
R—Rock ﬁll;
S—Stone masonry;
FC—Flood control;
I—Irrigation;
M—Mining;
N—Navigation; P—Power;
RP—Recreation purposes;
RR—River regulation;
WS—Water supply;
UC—Under construction.
1.5. FAMOUS DAMS OF THE WORLD 23
Table 1.1 Highest dams of the world
Name Location Type Purpose Year H L M R
Rogun Ta jikistan E PI 1987 1,066 2,506 81,096,000 3,091
Nurek Turkmenistan E PI 1980 1,040 2,390 75,864,000 2,745
Grande Dixence Switzerland CG P 1962 935 2,280 7,792,000 106
Inguri Russia A PIFC 1980 892 2,513 4,967,000 261
Vaiont Italy A  1961 858 624 460,000 
Tehri India ER  UC 856   935,271
Guavio Colombia ER  1989 807 1,280 23,223,000 269
Mica Canada ER PFC 1974 794 2,600 42,000,000 6,517
SayanoShushenshaya Russia A PN 1980 794 3,504 11,916,000 8,261
Chicoasen Mexico ER P 1981 787 1,568 15,700,000 439
Chivor Colombia ER P 1977 778 919 14,126,000 215
Mauvoisin Switzerland CA P 1957 777 1,706 2,655,000 48
Oroville California E PIFCWS 1968 770 6,920 78,008,000 1,153
Chirkeyskaya Russia A PIFCWS 1975 764 1,109 1,602,000 734
Bhakra India CG PI 1963 742 1,700 5,400,000 2,607
Hoover USA CAG PIFCRR 1936 726 1,244 4,400,000 9,696
Contra Switzerland A P 1965 722 1,246 861,000 23
Mratinje Yugoslavia A P 1975 722 879 971,000 232
Dworshak Idaho G PFCRP 1972 717 3,287 6,500,000 1,125
Glen canyon USA CAG P 1964 710 1,560 4,901,000 8,798
Toktogul Kyrgyzstan A PI 1977 705 1,476 4,186,000 5,148
Daniel Johnson Canada CM P 1968 703 4,311 2,950,000 37,473
Luzzone Switzerland CA P 1963 682 1,738 1,739,000 23
Keban Turkey ERG P 1974 679 3,881 20,900,000 8,182
Dez Iran CA PI 1963 666 696 647,000 882
Almendra Spain AG P 1970 662 1,860 2,188,000 700
Kolnbrein Austria A P 1978 656 2,054 1,995,000 53
Karun Iran A PI 1975 656 1,247 1,570,000 766
New Bullard’s Bar USA A PIFCRP 1970 637 2,200 2,700,000 557
New Melones USA ER PIFCRP 1975 625 1,600 15,970,000 782
24 CHAPTER 1. INTRODUCTION
Table 1.2 Larg est dams of the world
Name Lo cation Type Purpose Year H L M R.
New Cornelia Tailings USA E M 1973 98 36,500 274,026,000 6
TarbelaName Pakistan ER PI 1976 470 9,000 159,200,000 3,617
Fort Peck USA H PIFCN 1940 250 21,026 125,612,000 6,234
Lower Usuma Nigeria E  1990 528 14,531 121,640,000 26
Ataturk Turkey ER  1990 604 5,971 110,522,000 
Guri Venezuela ERG P 1986 531 30,853 101,819,000 36,720
YacyretaApipe ArgentinaParaguay EG PIWSRP UC 108 164,000 95,063,000 4,464
Oahe USA E PIFCN 1963 245 9,300 92,008,000 7,687
Mangla Pakistan E PI 1967 380 11,000 85,872,000 1,678
Gardiner Canada E PIWS 1968 223 16,700 85,743,000 2,607
Afsluitdijk The Netherland E FCWS 1932 62 10,500 82,927,000 1,585
Rogun Ta jikistan E PI 1987 1,066 2,506 81,096,000 3,091
Oroville USA E PIFCWS 1968 770 6,920 78,008,000 1,153
San Luis USA E PI 1967 382 18,600 77,666,000 664
Nurek Turkmenistan E PI 1980 1,040 2,390 75,864,000 2,745
Garrison USA E PIFCN 1956 203 11,300 66,506,000 7,925
Oosterschelde The Netherland EG FCWS 1986 164 29,528 65,397,000 
Co chiti USA E FCRP 1957 253 26,891 64,631,000 167
Tabka Syria E PIFC 1975 197 14,764 60,168,000 3,698
Kiev Ukraine EG PN 1964 72 177,448 57,552,000 984
Aswan High Egypt ER PIRP 1970 364 12,565 57,203,000 44,642
Bennett Canada E P 1967 600 6,700 57,203,000 18,575
Tucurui Brazil EG PNWS 1984 282 13,779 56,244,000 8,982
Mission Trailing #2 USA E  1973 128  52,435 15
Fort Randall USA E PIFCN 1956 165 10,700 50,205,000 1,858
Kanev Ukraine E P 1974 82 52,950 49,520,000 692
Itumbiara Brazil EG  1980 328 21,981 47,088,000 4,499
1.6. POWERGENERATOR, FLOODCONTROL ANDIRRIGATIONDAMS25
Table 1.3 Dams with greatest reservoir
Name Location Type Purpose Year H L M R
Owen Falls Uganda CG PI 1954 100 2,725  54,091
Bratsk Russia ECG PNWS 1964 410 16,864 18,283,000 44,713
Aswan High Egypt ER PIRP 1970 364 12,565 57,203,000 44,642
Kariba Gorge ZambiaZimbabwe CA P 1959 420 2,025 1,350,000 42,361
Akosombo Ghana R P 1965 463 2,100 10,400,000 39,102
Daniel Johnson Canada CM P 1968 703 4,311 2,950,000 37,473
Guri Venezuela E—RG P 1986 531 30,853 101,819,000 36,720
Krasnoyarsk Russia CG PFCN 1972 407 3,493 5,685,000 19,364
Bennet Canada E P 1967 600 6,700 57,203,000 18,575
Zeyskaya Russia B PIFC 1975 369 2,343 3,139,000 18,069
1.6 POWER GENERATOR, FLOOD CON
TROL AND IRRIGATION DAMS
1.6.1 Power Generator Dams
Most electric power is generated in large plants that use coal, gas, oil, or
nuclear energy. Electric energy may also be obtained from waterpower. The
roar of a waterfall suggests the power of water. Rampaging ﬂoodwaters can
uproot strong trees and twist railroad tracks. When the power of water is
harnessed, however, it can do useful work for man. Since ancient times man
has put to work the energy in the ﬂow of water. He ﬁrst made water work
for him with the waterwheel—a wheel with paddles around its rim. Flowing
water rotated the waterwheel, which in turn ran machinery that was linked to
it. Today, new kinds of waterwheels spin generators that produce electricity.
Electricity from waterturned generators is called hydroelectric power.
Waterpower produces about 8 percent of the electricity used in the United
States. It accounts, however, for only a fraction of the total energy used
for mechanical power, heat, light, and refrigeration. Coal, petroleum, and
natural gas supply most of the rest. Among water’s virtues as a source of
power is the fact that it can be used again after it supplies energy, while other
energy sources are destroyed when they are used. Hydroelectric installations
also supply added beneﬁts to a region. For example, a dam built to provide
a head for water turbines usually creates a reservoir that can supply water
for irrigation and drinking. Water passes downward through a hydraulic
26 CHAPTER 1. INTRODUCTION
turbine that is connected to a generator. Large plants, which depend on
large volumes of water dammed upstream, can generate more than 2,000
MW. There are many small plants on rivers, some generating only a few
hundred kilowatts. Among the largest plants in the United States are Hoover
Dam and the Grand Coulee Dam. Hoover Dam has an installed capacity of
1,244 MW. Grand Coulee has an installed capacity of 2,025 MW. It is part
of the Columbia River Basin project, which has a planned capacity of 9,770
MW.
The power produced by water depends upon the water’s weight and head
(height of fall). Each cubic meter of water weighs 1000 kilograms. For
example, a column of water that is one meter square and one meter high
would contain one cubic meter of water. It would press upon each square
meter of turbine blade with a force of 1000 (1 x 1000) kilograms. Engineers
measure waterpower in terms of watts (kilowatts or megawatts). One watt is
the force it takes to raise one kilogram one meter in one second. The power
potential of a waterfall is found by multiplying its ﬂow, measured in cubic
meter per second, by its height, measured in meter. Then the product is
multiplied by 980, which is 1000 (the number of kilogram in a cubic meter
of water) multiplied by 9.81(gravitational acceleration). For example, a 100
meter high waterfall with a ﬂow of 10 cubic meter per second would develop
100 x 10 x 980, or 980,000, watts. The output of a hydroelectric plant is
usually measured in kilowatts or megawatts of electricity.
1.6.2 Flood Control Dams
One way to avoid ﬂoods is to take the obvious precaution of living where
there is no danger of high waters. It has always been convenient and often
necessary to build homes and factories on the ﬂoodplains along rivers and
streams and on the seacoasts. American pioneer settlers depended upon the
streams for drinking water, transportation, and power to run their mills and
factories. Floodplains, deep with the silt laid down by overﬂowing rivers, are
fertile farmlands. The earliest towns and farms, therefore, were established
along the riverfronts, and large portions of them were built on land that was
subject to periodic ﬂooding. While the communities were small, the damages
suﬀered from ﬂoods were limited. With the great population and industrial
growth of the cities, ﬂood damage has become a serious national problem.
One of the basic approaches to ﬂood control is to minimize the extent of
ﬂooding by building dams, reservoirs, levees, and other engineering works.
1.6. POWERGENERATOR, FLOODCONTROL ANDIRRIGATIONDAMS27
Engineers in ancient times built earthen mounds to keep back ﬂoodwater.
Such artiﬁcial embankments, called levees, held Chinese rivers in check for
centuries. This method was followed in colonial America.
Because a levee at one point conﬁnes the water there and raises the peak
of ﬂood waters upstream and downstream, levees once started usually have
to be built at all the low points of a river system. Furthermore, a system of
levees is only as strong as its weakest spot. Thus uniform height and strength
are required. Only a government which controls the river from end to end
can safely supervise levee building. Floodways and spillways divert excess
water from the main river channel and carry it oﬀ by a diﬀerent route.
Dams and the reservoirs behind them help control ﬂoods. By emptying a
dam before a ﬂood is expected, storage space is obtained in which the ﬂood
waters can be impounded for gradual release later. Even if the reservoir is
nearly full, it acts like a safety valve. The amount of water which would add
3 meter to the height of a river 30 meter wide would add only one meter
to a reservoir or lake 90 meter wide. Moreover, evaporation from the broad
surface of a reservoir or lake is far greater than evaporation from the narrow
surface of a river. Thus less water ﬂows on to swell ﬂoods downstream.
Floodcontrol dams are built to create big storage capacity and are planned
for rapid ﬁlling and emptying. During excessive rains, water collects in the
storage reservoirs and is released in controlled amounts to the channel below
the dams. The water carries with it large quantities of the richest topsoil,
which muddies the rivers and is ultimately lost in the ocean. American
rivers, for example, carry oﬀ an estimated 280 million cubic meters of solid
matter each year. Agricultural experts propose to return the steepest hills
along the headwaters of American rivers to forest. By means of terracing,
contour plowing, and a wise choice of plants, runoﬀ and erosion are checked
on gentler slopes. By damming gullies, runoﬀ is slowed, and silt from above
slowly rebuilds the eroded spots. Thus ﬂood prevention and erosion control
go hand in hand. Preventing soil erosion also aids ﬂood control by slowing
the rate at which silt ﬁlls the reservoirs behind ﬂoodcontrol dams. Steps to
lessen the eﬀects of drought also aid in ﬂood control. Some lakes, swamps,
and marshes that were once drained to make farm lands are being restored
to preserve the level of underground water in time of drought. This program
also reduces ﬂoods by increasing evaporation and by the safetyvalve action
of wide lakes or swamps on narrow rivers. Thus the prevention and control of
ﬂoods are tied with drought measures as well as with waterpower, navigation,
soil conservation, and land use.
28 CHAPTER 1. INTRODUCTION
1.6.3 Irrigation Dams
Irrigation is the artiﬁcial supply of water to agricultural land. It is practiced
by more than half the farmers in the world because they need more water
for their crops than is available from rainfall. Irrigation projects must also
allow for removal of excess water.
Generally the need for irrigation water is highest in the dry season when
river ﬂows are lowest. To ensure continuous supply, water must be stored
on a seasonal and sometimes annual basis. By erecting a dam, an artiﬁcial
lake, or reservoir, is created from which water can be released as required.
Some reservoirs are capable of storing billions of gallons of water. Large dams
and the associated reservoirs are often built for multipurpose use—irrigation,
ﬂood control, hydroelectric power generation, municipal and industrial water
supply, and recreation. Systems of dams and reservoirs along a river and its
tributaries are developed with the purpose of providing comprehensive and
integrated water resources management in an entire basin.
1.7 INSTRUMENTATIONS ANDSURVEIL
LANCE OF DAMS
1.7.1 Surveillance
External and internal examination of dams are required during their life time.
Long term surveillance program can ensure dams owner of their safe opera
tion. Long term surveillance requires instrument installation and monitoring
to assess the performance of structures with respect to design parameters.
Design of monitoring installations is to measure displacement, settlement,
strain, stress, piezometric pressures, seepage, uplift, ﬂow velocity and water
levels, and alarm systems as required for a particular impoundment. Com
pilation of results on a computerized data base is done and results are inter
preted and reported in the form of recommendations outlining appropriate
action need to be taken, when it is necessary.
1.7.2 Instrumentation
Instruments are used to characterize dam site condition as well as dam struc
ture conditions. Instruments can be used for verifying design veriﬁcation,
1.7. INSTRUMENTATIONS AND SURVEILLANCE OF DAMS 29
safety assessment, performance assessment.
Design Veriﬁcation Instruments are used to verify design assumptions and
to check that performance is as predicted. Instrument data from the initial
phase of a project may reveal the need (or the opportunity) to modify the
design in later phases.
Safety Instruments can provide early warning of impending failures, al
lowing time for safe evacuation of the area and time to implement remedial
action. Safety monitoring requires quick retrieval, processing, and presenta
tion of data, so that decisions can be made promptly.
Performance Instruments are used to monitor the inservice performance
of a structure. For example, monitoring parameters such as leakage, pore
water pressure, and deformation can provide an indication of the performance
of a dam.
The Choice of Instruments depends on critical parameters of each project.
The designer must identify those parameters and then select instruments to
measure them. What information is required for the initial design? What
information is required for evaluating performance during and after construc
tion? When the parameters are identiﬁed, the speciﬁcation for instruments
should include the required range, resolution, and precision of measurements.
In some cases It may be necessary to have complementary parameters
and redundant measurements. In some cases, it may be suﬃcient to monitor
only one parameters but when the problem is more complex, it is useful to
measure a number of parameters and to look for correlation between the mea
surements. Thus it is common practice to choose instruments that provide
complementary measurements. For example, inclinometer data indicating
increased rate of movement may be correlated with piezometer data that
shows increased pore pressures. Another beneﬁt of selecting instruments to
monitor complementary parameters is that at least some data will always be
available, even if one instrument fails.
Instrument performance is speciﬁed by range, resolution, accuracy, and
precision. The economical designer will specify minimum performance re
quirements, since the cost of an instrument increases with resolution, accu
racy, and precision. Range is deﬁned by the highest and lowest readings
the instrument is expected to produce. The designer typically speciﬁes the
highest values required. Resolution is the smallest change that can be dis
played on a readout device. Resolution typically decreases as range increases.
Sometimes the term accuracy is mistakenly substituted for resolution. Res
olution is usually many times better than accuracy and is never expressed
30 CHAPTER 1. INTRODUCTION
as a ± value. Accuracy is the degree to which readings match an absolute
value. Accuracy is expressed as a ± value, such as ±1mm, ±1% of reading,
or ±1% of full scale. Precision or repeatability is often more important than
accuracy, since what is usually of interest is a change rather than an absolute
value. Every time a reading is repeated, the value returned by the instru
ment is slightly diﬀerent. Precision is expressed as a ± value representing
how close repeated readings approach a mean reading. The diﬀerence in cost
between a highquality instrument and a lesserquality instrument is gener
ally insigniﬁcant when compared to the total cost of installing and monitoring
an instrument. For example, the cost of drilling and backﬁlling a borehole
is typically 10 to 20 times greater than the cost of the piezometer that goes
in it. It is false economy to install a cheaper, less reliable instrument. It
is expensive and sometimes impossible to replace a failed instrument. Even
when it is possible to replace the instrument, the original baseline data is
no longer useful. Some instruments are excellent for shortterm applications,
but may exhibit excessive drift over the long term. Temperature and humid
ity also aﬀect instrument choice. Instruments such as hydraulic piezometers
and liquid settlement gauges have limited use in freezing weather. In tropi
cal heat and humidity, simple mechanical devices may be more reliable than
electrical instruments.
Consider the personnel and resources at the site when choosing instru
ments. Do technicians have the skills required to install and read a particular
type of instrument? Are adequate support facilities available for maintenance
and calibration of the instrument?
An automatic data acquisition system may be required when:
(1) there is a need for realtime monitoring and automatic alarms;
(2) sensors are located at a remote site or in a location that prevents easy
access;
(3) there are too many sensors for timely manual readings; or
(4) qualiﬁed technicians are not available.
If a data acquisition system is required, the choice of instruments should
be narrowed to those that can be connected to the system easily and inex
pensively.
1.7.3 Instruments
In the following section, Some instruments used in monitoring of dams are
brieﬂy described.
1.7. INSTRUMENTATIONS AND SURVEILLANCE OF DAMS 31
Piezometers measure porewater pressure and ground water levels. Piezome
ter measurements help engineers to Monitor water levels, Predict slope sta
bility, Design and build for lateral earth pressures, Design and build for uplift
pressures and buoyancy, Monitor seepage and verify models of ﬂow.
Inclinometers may be installed to check that actual movements of a
structure correlate to those predicted during the design phase. Inclinometers
are installed to monitor the magnitude, direction, and Corrective Measures
rate of movement. This information helps engineers determine the need for
corrective measures. Inclinometers are installed for long term monitoring
to detect LongTerm Performance changes in ground conditions or in the
structure itself. Inclinometers, particularly inplace inclinometers that are
monitored continuously, can provide early warning of catastrophic failure.
Horizontal inclinometers are used to monitor settlement in foundations and
embankments. Inclinometers monitor movement, a direct measure of stabil
ity, so they are often used in site investigations. Installed at the proposed
site for a dam, an inclinometer might detect movement at a subsurface shear
plane. The shear plane could cause problems later when the reservoir behind
the dam is ﬁlled and porewater pressure along the shear plane increases.
Beam sensors are used for monitoring settlement, heave, lateral defor
mation, or convergence. Two version of beam sensors are horizontal version
for monitoring settlement and heave and vertical version for monitoring lat
eral displacement and convergence.
Tiltmeter is used for monitoring changes in inclination. It is used for
monitoring the rotation of concrete dams.
Beam sensors diﬀer from tiltmeters in two important respects: First,
the beam sensor has a deﬁned gauge length, typically 1 to 3 meters, so
changes in tilt can be converted simply and accurately to millimeters of
movement (settlement, heave, convergence, or lateral displacement). Second,
beam sensors can be linked endto end to monitor diﬀerential movements
and provide absolute displacement and settlement proﬁles.
Tiltmeters typically have a more limited function, that of monitoring
rotation. A tiltmeter can be used with a large number of tilt plates to
detect diﬀerential movement in a structure, but the resulting data cannot
provide absolute displacement and settlement proﬁles. In general, however,
both types of sensors can be used to evaluate the performance of dams under
load. They Provide early warning of threatening deformations, allowing time
for corrective action to be taken or, if necessary, for safe evacuation of the
area.
32 CHAPTER 1. INTRODUCTION
Borehole Extensometers are used to monitor settlement, heave, con
vergence, and lateral deformation in soil and rock. Typical applications in
clude monitoring settlement or heave in excavations, foundations, and em
bankments. Data from the extensometer indicate the depths at which settle
ment has occurred as well as total amount of settlement.
Pneumatic settlement cells provide a singlepoint measurement of
settlement. They can be read from a central location and are particularly
useful where access is diﬃcult. They are used to monitor consolidation in
the foundation during construction and monitor longterm settlement in the
foundation and ﬁll.
Surface Extensometers (Tape Extensometer) is used to determine
changes in the distance between reference points anchored in walls or struc
tures of an excavation, precision measure of convergence.
Load cells are used to prooftest and measure loads in tie backs, rock
bolts, ground anchors, and struts. They are also used for monitoring the
performance of anchor systems.
Total Pressure Cells measure the combined pressure of eﬀective stress
and porewater pressure. In general, they are used to verify design assump
tions and to warn of soil pressures in excess of those a structure is designed
to withstand. Total pressure cells are installed within ﬁlls to determine the
distribution , magnitude and directions of total stresses. They can also be
installed with one surface against a structure to measure total stresses acting
on upstream face of the dam, retaining walls, against piles, and pipes. In Em
bankment Dam Total pressure cells used to conﬁrm design assumptions and
An array of cells provides data to determine distribution, size, and direction
of total stresses within the clay core.
Soil strainmeters are used for monitoring horizontal strain in embank
ments and monitoring tension cracks in earth structures.
Embeded jointmeters are used for monitoring movement at joints in
mass concrete structures. It is used in mass concrete structures such as dams,
abutments, foundations to monitor movement at joints.
Surfacemount jointmeters are used to monitor movement at joints
and cracks in concrete structures under harsh environments and submersion.
Typical application include monitoring movement at submerged construction
joints in concrete dams and monitoring joints or cracks in tunnels and tanks.
Strain gauge is used for monitoring strain in steel structural members,
reinforced and mass concrete.
1.8. ECOLOGICAL/ENVIRONMENTAL CONSIDERATIONOF DAMOPERATION33
1.8 ECOLOGICAL/ENVIRONMENTAL CON
SIDERATION OF DAM OPERATION
The following observations are generalized fromthe literature at large. Whether
any particular concern applies to any particular dam is a function of a suite of
factors that might include the dam’s type (e.g. concrete or earth), its purpose
(e.g. hydrologic regulation, hydroelectric generation, tailings impoundment),
its size, location and operating protocol (e.g. timing of drawdowns, epilim
netic versus hypolimnetic draw, peaking versus base load generation, etc.).
Finally, The observations are derived from an ecocentric rather than any
anthropocentric ethical premise, that is a rather uncompromising ecological
view of the relationship between nature and humans.
The ecosystem approach can work for dam owners/operators to a greater
extent than they may perceive it to be working against them. Here, the
Physicochemical eﬀects (physical or chemical factors) have separated from
biological eﬀects. Generally changes in the biological community are driven
by changes in the physical or chemical (physicochemical) dimensions of the
habitat (niche).
Impacts are also organized spatially as those that occur: (a) downstream
of the dam,(b)within the water column of the reservoir/impoundment, (c)
at the sediment/water interface (which may or may not feed into the wa
ter column) and (d) a category for impacts that act outside the immediate
dam/reservoir system. These latter impacts might be viewed as externalities
with less relevance for operators. In any event, the evidence for external
impacts is substantially weaker than that for ac.
Downstreamimpacts could include changes in Physicochemical factors
and Community structure. The physicochemical factors are: temper
ature, oxygen content, water chemistry, particulate loading, transparency,
discharge patterns, ﬂow rates/shear and channel simpliﬁcation. The com
munity structures are: environmental conditions exceed tolerance limits of
original community members, species richness, genetic diversity, productivity
and feedback.
Water column impacts could include changes in Physicochemical parame
ters and Community structure. The Physicochemical parameters consists of
temperature, oxygen content, water chemistry, transparency, morphometry,
internal seiches and altered currents. The community structures are longer
migration times, environmental conditions exceed tolerance limits of original
34 CHAPTER 1. INTRODUCTION
community members, productivity, diversity and feedback.
Sediment impacts could include changes in Physicochemical parameters
and Community structure. The ﬁrst one consists of temperature, oxygen
content, light, remineralization, compression and sedimentation. The second
one consists of environmental conditions exceed tolerance limits of original
community members, species richness, productivity and feedback
Externalities could occur at a variety of scales and might include Local,
Regional or Global externalities. The local externalities can cause riparian
losses, hydrologic budget, habitat fragmentation. The regional externalities
can cause climate change while the global externalities can cause climate
change and orbit dynamics. These externalities could lead to loss of biodi
versity and decreasing ecological integrity.
Dams can be operated in ways that decrease their environmental impacts.
Recent examples include the installation of a selective water withdrawal sys
tem on the Hungry Horse Dam and the (initially seredipitous) drawdown at
Glen Canyon. Doing so requires increased communication and trust among
engineers, ecologists and other professionals in order to properly bound the
problem. Engineers need to evaluate the extent to which regard for the en
vironment is codiﬁed as good practice. There will have to be a willingness
to experiment (e.g. demonstration projects) with a commitment to redeﬁn
ing operational technical directives as new knowledge emerges ( e.g. optimal
performance standards might equate ecological integrity with revenue gener
ation).
Ecological integrity (the genetic diversity, species diversity and ecological
diversity that maintains the ﬂows of energy and materials through ecosys
tems) is the key mechanism for adapting to changes in environmental con
ditions through evolutionary change. Protecting diversity means protecting
habitat, not only against loss but against fragmentation.
1.9 THE HISTORY OF DAMS DESIGN
Throughout the history, dams have been designed to fulﬁll a speciﬁc purpose
to civilizations. The following historical outline illustrate the main dam
purposes and the development of their design.
1.9. THE HISTORY OF DAMS DESIGN 35
1.9.1 Irrigation Dams
A steady supply of water for irrigation is extremely important for any civi
lization to thrive. Heavy ﬂooding in the wet season and long droughts in the
dry season makes farming very diﬃcult. However, if the water is contained
during the wet season and released periodically through the dry season, a
constant water supply will be present.
Ancient Dams
The ﬁrst dam designed for irrigation purposes was the Marib dam in Yemen
Capital City of Saba. The Marib was an embankment dam constructed
about 510 BC. It reached a height of 20m and was about 700m long. The
embankment slopes on each side were 1:1.8 with no road crest along the top.
The most interesting aspect of this dam is that the ﬁll was placed in layers
parallel to the slopes, instead of the typical horizontal layer conﬁguration.
This damalso did not have any type of impermeable element in its design. On
each end of the dam an outlet structure was built of excellent aslar masonry.
The sills of both outlets were almost located at the same height from the
riverbed. A 50m long spillway, with a sill height was also included in the
dam design located in the northern section of the dam. the capacity of the
Marib dam was 30 milliom m
3
, about 15% of the average annual rainfall for
the area. the ﬁnal failure came about 1,300 years after the completion of the
dam. It was not repaired and lead to immigration of about 50,000 people
that depended on the water supply. Recently, a modern embankment dam
with a capacity of 400 million m
3
was built in 1986, 3 km upstream of the
ancient dam site.
The Kirsi dam of Iraq was built under the Assyrian king of Sennacherib
about 355 km north of Baghdad. This dam was also a diversion dam, used
to direct the river ﬂow into a 15 km long canal for irrigation purposes. It was
designed as a gravity dam constructed rubble masonry weirs. The upstream
face was vertical with a board overﬂow crest and a stepped downstream face.
Romans
The Romans were highly advanced in hydraulic engineering from their many
engineering feast such as aqueducts. However, they did not attempt dam con
struction until around 150 BC when they annexed Greece. One advantage
36 CHAPTER 1. INTRODUCTION
the Romans had, was that they already possessed fully developed construc
tion technique based on traditional tools such as levers, picks and shovels.
The Romans also used pulleys in multiple conﬁgurations to lift heavy ob
jects vertically and horizontally. Some of the Romans advance tools included
rules, squares, plumblines, and spirit levels. Traditional building materials
were used in their design however, the Romans also added concrete to their
dams. This allowed them to develop new shapes using formwork and con
crete. They added volcanic ash or ground brick into the concrete mix to
ensure that it would harden, even when submersed under water.
Medieval and Postmedieval Europe
Damming for irrigation purposes was practiced throughout this era. Irriga
tion was used throughout Europe, however, it was essential in the southern
countries of Europe like Italy and Spain and required storage reservoirs. The
dams in these southern countries were mostly of masonry, which contrast to
dams built in northern Europe that were built mainly of soil embankments.
The reason for dams built in Italy and Spain being masonry is most likely an
inheritance from Roman civilizations who used embankments primarily for
supporting elements.
The ﬁrst true arch dam built in Europe since the Roman time, was built
in 16321640 near Elche. The arch stretched across the main section of the
gorge and butted into wing walls directed upstream on either side of the
gorge. The arch thrust was directed downwards, since it did not have much
support at the crest level. The main arch was 75m long and 9m wide at the
crest and had a curve radius of 62m and a curve angle of 70
◦
. The design of
this dam was later tested using the crowncantilever computer analysis that
determined it to be satisfactory.
1.9.2 Dams Designed for Water Supply
It is impossible for a civilization to survive without a constant fresh water
supply, since it is a basic need of all humans. In many parts of the world, the
freshwater supply ﬂuctuates from enormous amounts during the rainy season
to long droughts during the dry season. for this reason dams were invented
to store large amounts of water from the end of the rainy season until the
end of the dry season, to ensure a constant supply of water.
1.9. THE HISTORY OF DAMS DESIGN 37
Ancient Dams
The oldest dam in the world is the Jawa dam found 100 km northeast of Jor
dan capital Amman. This dam was built around 3,000 B.C. and was designed
as a gravity dam. The design of this dam was quite complex considering that
it was ﬁrst dam to be built. It consisted of two dry masonry walls, with an
earth core that acted as the waterretaining element. An impervious blanket
was also provided in front of the upstream heel. A downstream embankment
provided stability of the structure. The dam was later raised by 1m, using
similar design methods as the original dam.
Romans
The Romans were the ﬁrst civilization to use an arch dam. It is remarkable
that the Roman did not use the arch dam earlier, since they had already be
come masters at employing it into their bridges and buildings. The ﬁrst arch
dam was constructed at Barcinas on Cubillas river 2km north of Granada,
in southeast Spain. the arch design for this dam was not actually used to
increase stability, but to lengthen the dam crest for easier passage of ﬂood
water.
the ﬁrst true arch dam was built by the Romans in the Vallon de Baume
4km south of SaintRemy de Province in southeastern France. This dam was
12 high and 18m long, with a radius about 14m and a 73
◦
central angle.
It consisted of 1.3m upstream wall and a 1m downstream wall made out of
masonry. The earth core between the two walls was about 1.6m wide, which
means that the downstream wall was subjected to the entire load.
Medieval and Postmedieval Europe
The Roman standard of public fresh water supply continued throughout this
period. the water supply for Istanbul in Turkey, during the beginning of
the ear, was dealt with by many underground cisterns within the city. It
was not until after the conquest of Istanbul in 1453 that the Ottoman Turks
reconstructed some of the old Roman aqueducts. The ﬁrst storage dam was
built in 1560 on the remains of the Roman Belgrade dam. It was massive
gravity dam, with a height of 15m and a base width of 56% of its height.
The length of the dam was about 85m and had a 1.3 million m
3
reservoir
capacity. This dam also included a couple of unnecessary buttresses to add
stability to the structure.
38 CHAPTER 1. INTRODUCTION
1.9.3 Flood Control Dams
Since large number of natural fatalities, injuries and damage are caused by
rapid ﬂooding it is important to have ﬂood control methods. Dams have
been used for ﬂood control since the ﬁrst civilizations were formed.
Ancient dams
The SaddelKafara dam of Egypt was built around 2600 BC, about the same
time as the beginning of the Pyramids. It stands 14m high and has a 113m
crest. the lack of knowledge in hydrology and soil mechanics lead to the dam
being over, designed. It had three distinct sections; the downstream shell
made of rock, the upstream shell also made of rock and a centre core made
out of ﬁne silty sand and gravel. 17,000 blocks weighting 300kg each were
carefully placement over the upstream and downstream walls to protect the
dam against erosion. Construction problems were most likely encountered
since it took longer than one dry season to built, and no evidence of any type
of diversion could be found.
Romans
the Romans built the Cavdarhisar dam near Kutahya in western Turkey to
protect against raging ﬂoodwater. It was designed as a gravity structure
with height of 7m, wide of 8m and a length of 80m. It also contained a large
bottom outlet with a clear section about 11m
2
that shows this dam was not
designed to be a reservoir, not a ﬂood control measure.
Medieval and Postmedieval Europe
Throughout the period few dams were actually built to for ﬂood protection,
however, many dams were built to cause ﬂooding. Controlled ﬂooding was
used for transporting wood by ﬂuming. Originally, the dams built for ﬂum
ing were made from timber, however, few of these dams have any remains.
from around the end of the 17
th
century the timber dam were replaced with
masonry ﬂuming dams. An example of a ﬂuming dam is the Belcna dam in
western Slovenia built in 1769. It was 18m high with a length of 35m, a crest
width of 6.8m and a 12.4m base width. It was built of solid masonry and
contained two outlets 5.1m in height and a width of 3.7m which allowed a
large amount of water to rush out in a short time.
1.9. THE HISTORY OF DAMS DESIGN 39
1.9.4 Power Dams
Waterpower was a technology that did not really take oﬀ until after the 11
th
century. The invention of the dam was an essential development in the rise
of waterpower because it diversiﬁed the applications of both the vertical and
horizontal water wheels.
Romans
The Romans were the ﬁrst civilization to harness the power of water however;
they were limited to mostly grinding cereals. One of the most sophisticated
dams built by the Romans was the Monte Novo dam located 15km east of
Evora in southeast Portugal. It was a 5.7m high, 52m long arch dam with a
radius of 19m and a central angle of 90
◦
. The dam was built using blocks of
shist laid horizontally in a lime mortar. the dam’s reinforcement consisted of
the fact that the curved central part at each end was embedded into the wing
walls without any abutment blocks to absorb the horizontal arch thrust. The
designers of this dam must have been unsure of their sophisticated design,
because they also included two buttresses to increase the stability. It is
thought that the water fromtwo outlets was used to drive waterwheels further
down the river, The outlets were 1.2m and 1.4m wide, with a height of 1m
through the dam’s base.
Medieval and Postmedieval Europe
It took several centuries for Europe to recover from impacts of Christian
ization and Germanization. By the end of the 8
th
century, about 300 years
after the disintegration of the western half of the roman empire technology
had gain sprouted new wings. Dams were built in order to remove the water
wheels from rivers where debris, ice jams and ﬂoods played havoc with the
waterwheels and their structures.
One of the interesting dam built in this period for waterpower was the
Castellar dam and millhouse in southwestern Spain. The dam’s reservoir
capacity is very small(0.3 milliom m
3
), however, the height of the dam was
19m high. This large height gave a very high potential head of the water
in the reservoir, which could be transferred into waterwheel power. The
millhouse was located at one end of the weir and supplied directly with water
from the reservoir, a method of design that is used in modern lowhead dam
power plants. Another equally modern design characteristic was the location
40 CHAPTER 1. INTRODUCTION
of the millhouse, which was near the toe of the 19m high reservoir. An equal
number of vertical shafts lead the water to the horizontal water wheels. The
sturdy end and intermediate walls of the mill house act as buttress because
the width of the dam was only 36% of its height. Many of the concepts built
into this dam have direct links to modern methods, although it was built
around 1500 AD.
1.9.5 The Moslem World
Arabia
Arabs quickly developed their country within one century after the impact
of new Islamic religious founded by prophet Mohammed. Several irrigation
dams around the new power centres of Mecca and Madina, were built. The
following is a summary of 7
th
and 8
th
century dams in central Arabia.
Table 1.4 Summary of 7
th
and 8
th
century dams in central Arabia
Nearest City Name Height (m) Length (m)
Mecca Agrab 4.0 113
 Ardab 5.5 315
 Darwaish 10.0 150
 Sammallaagi 11.0 225
Medina Hashquq 2.0 130
 Qusaybah 30.0 205
All of these dams were of the gravity type. They all have two outer walls
of dry masonry and an earth or rubble core in between. The walls of some
of these dams were vertical. Whereas, the other had inclined walls as much
as 60%. Strong inclinations were applied to dams with relatively wide base,
like Darwaish.
Although, most of these dams are already destroyed, one third of them
like Saammallagi, are operative and more or less in their original shape.
South Western Asia
The Moslems conquered southwestern Asia in Seventh and Eighth centuries
as far as the Indus river and Uzbakistan. During this time, Adhaim dam was
built on the Adhaim river. This dam was located 150 km north of Baghdad.
The dam was 130 meters long and 12 meters high. Adhaim Dam had an
1.9. THE HISTORY OF DAMS DESIGN 41
equal width all along its length. Upstream face was vertical and downstream
one was stepped to 43% average inclination.
In the 10th century, a whole series of irrigation and power dams were built
on Kur river, east of Shiraz in southern of Iran. Table 1.5 shows , dimensions
and the name of the dams on the Kur river.
The Amir dam is a large power dam including 30 water wheel being
driven by a retained water. This dam has a trapezoidal cross section with
a relatively steep down stream face and a base width of more than twice its
height.
Table 1.5 Dimensions and name of the dams on the Kur river
Name Height (m) Length (m)
Amir 9 103
Feizband 7 222
Tilkan 6 162
Mawan 6 66
Djahanabad 5 50
Towards the end of the 10th century, several large gravity dams were
built under Ghaznavid rule. These dams were built in countries such as Iran,
Afghanistan and Uzbakestan. Table 1.6 contains some information about
Ghaznavid gravity dams. Khan dam was built of granite ashlar masonry
and was barely stable.
Table 1.6 Ghaznavid Gravity Dams
Name Nearest city Height (m) Length (m)
Gishtbank Samarkand 8 25
Khan Saamarkand 15.2 52
Sheshtaraz Mashhad 25 35
Soltan Mahmud Kabul 32 220
Genghis khan from Mongolia started to conquer the world in 1205. Mon
golians initiated the construction of several dams in Iran. Table 1.7 shows
the summary of the data.
Table 1.7 Mongolians Dams
42 CHAPTER 1. INTRODUCTION
Year of completion Name Nearest city Type Height (m) Length (m)
1285 Saveh Tehran Gravity 25 65
~1300 Kebar  Arch 26 55
~1350 Kalat Mashhad Gravity 26 74
1400 Abbas Tabas Arch 20 ?
~1450 Golestan Mashhad Gravity 16 
The Saveh dam was the ﬁrst structure built during this time. It was a
gravity dam. The design of this dam was not successful. The river crushed
the dam after the ﬁrst impoundment.
In 16
th
and 17
th
century, the construction of arch dams discontinued and
only gravity dams were built. Most of the well known dams during this time,
were located in northeastern Iran and Uzbakistan.
Abdullahkan dam was built in the beginning of the 16th century. This
dam is 150 km northwest of samarkand. The height of this dam is 15 meters
and it has a length of 85 meters. Its downstream face was stepped to an
average inclination of 69%. Khajoo dam is one of the most beautiful dams
not only in the Moslem, but in the entire world which is located in Isfahan
of Iran.
1.9.6 Development of the Modern Dams
Unlike past periods in history, the modern dam is designed using speciﬁc
principles. The largest change in the design of a modern dam compared
to the earlier dams, has been the formation of engineering schools. This
changed the individualistic craftsmanship of engineering into a profession
based on scientiﬁc principles. Engineers were subdivided into speciﬁc disci
plines, increasing the advancements of both dam principles and construction
methods.
The modern embankment dam has improved greatly with the increase in
research on hydrology and soil mechanics. Henri Gautier was one of the ﬁrst
to study slope stability of diﬀerent soils. His experiments determined the
natural slope response of soils. As time went on, engineers were then able
to choose diﬀerent soils based on their internal angle of friction, to increase
the height of embankment dam designs. Studies were conducted on how
soil permeability changed for diﬀerent soils. In 1186 Philipp Forchheimer,
professor of hydraulics in Austria, developed a graphical method of Pierre
S. deLaplace’s formula of how water ﬂows in a soil. This graphical method
is commonly known as a ﬂownet. The single most important theoretical
1.9. THE HISTORY OF DAMS DESIGN 43
development of soil mechanics was published in 1925 by Karl von Terzaghi.
He found the explanation for why clays consolidate, by determining that the
consolidation was caused by the dissipation of the water pressure in the soil
pores. This ﬁnding led Terzaghi to the concept of eﬀective stress, that is
equal to the total stress minus the pore water pressure. This concept is still
considered the single most important concept of modern soil mechanics. In
1980, the Nurek earth ﬁlled embankment dam in Tajikistan was completed.
It stretched to a record height of 300m and had an embankment volume of
59 million m
3
.
The analysis of stress in structures in their elastic condition, the concepts
of the modulus of elasticity, and safety stresses was introduced in 1819 by
Louis M. H. Navier. These concepts were the beginning of understanding
how the force in gravity dam designs acted and helped to increase the height,
without increasing the amount of materials used.
Breakthrough advancements in arch dam design came in 1880’s when
Hubert Visher and Luther Wagnore developed a new method of arch dam
design. This new method did not merely design the arch dams as astach
of individual arches but as an interdependent system of arches. It also con
sidered the median vertical section of the dam to be a cantilever ﬁxed at
the base. The distribution of the loads resulted in a considerable relief for
the lower arches, and the upper ones had to partially support the cantilever.
Throughout this period, the arch dam took on many new shapes.
With the modern ear, improved mechanical construction method have
caused the greatest impact on all types of modern dam designs. Scraper,
bulldozers and vibrators all grew in size, resulting in the modern dams being
completed in record time. New techniques for placing concrete, have played
a roll in the construction and design methods for gravity dams. Roller com
pacted concrete is among the newest techniques implements in gravity dam
construction. In the early 1950’s Switzerland introduced bulldozers to spread
concrete, and vibrators were attached to caterpillar tractors. In the late
1970’s early 1980’s vibrations rollers replaced the tractor mounted immersed
vibrating system. These new construction techniques have spread through
out the world at a rapid pace, increasing both the dimensions of dams and
the speed of completion.
The number of new dams being built has decreased considerably. This
trend will not be able to continue as the world population continues to in
crease and fresh water, irrigation and ﬂood control needs increase as well.
Engineers need to constantly search for a better solution, taking all aspects
44 CHAPTER 1. INTRODUCTION
of design and the repercussions of that design into account. The past is a col
lection of design ideas, both good and bad, that should never be overlooked
when faced with a new design challenge.
As time progress onwards, dams continue to grow in height and capacity.
As they become larger, their impacts on society and the environments also
increases. It is important that proper monitoring programs are setup to
ensure that failure do not occur, because their eﬀects will be devastating,
One method of monitoring is a system of alarms downstream of the dam to
warn of potential failures. the environment is an aspect that needs to be
carefully controlled. With the growing size and capacity of modern dams,
the environmental impact and awareness therefore increases as well.
1.10 BEAVERS
A mammal belonging to the order of rodents, or gnawing animals, the beaver
has been recognized as a master engineer. By using teeth and paws, beavers
construct dams, lodges, storehouses, and canals. The animal is also known
for its aquatic lifestyle as well as for its beautiful fur.
The beaver has a thick body covered with a coat of long, reddishbrown
outer hairs and soft, dense, brown underfur. This warm, waterproof coat
allows the beaver to swim in icy water in the wintertime without discomfort.
Most of the beaver’s physical characteristics of the fur, toes, tail, ears, nose,
and lips are so constructed that the animal is well equipped for life in the
water as well as on land. The toes on the beaver’s large hind feet are webbed
for swimming. The second toe on each hind foot ends in a double claw with
which the animal combs its fur. The front feet are small and handlike and
are used for picking up and carrying various objects. The tail is shaped like
a paddle, broad and ﬂat, and is covered with scaly skin. The tail serves as a
prop when the beaver sits upright and as a rudder and scull when it swims.
As a danger signal to other beavers, the normally placid animal makes a
loud noise by slapping its tail on the water’s surface. The beaver’s facial
features also allow for its aquatic lifestyle. The short, thick head has small
rounded ears and a nose which are equipped with valves that close when
the animal swims underwater. A beaver may remain submerged for up to
15 minutes. The animal carries objects underwater in its mouth by closing
loose lips behind prominent front teeth, thus keeping water out of its mouth.
Like other rodents, the beaver has welldeveloped front teeth. These teeth
1.10. BEAVERS 45
have a very hard layer on the front surface and a softer backing. Since this
softer part wears away quickly, it leaves the thin chisel edge of the front layer
exposed. The animal’s teeth are always growing to make up for wear. The
beaver has a total of 20 teeth.
Beavers are social animals; they live in colonies and work together. The
life span of the animal may be as long as 19 years. A beaver begins its life in
a litter of from two to eight young, or kits. Four is the usual number. They
are born in the spring, about four months after conception. A mother will
sometimes raise not only her own oﬀspring, but also the young of another
female that has died. Newborn kits weigh from 8 to 24 ounces (225 to
680 grams) and are about 15 inches (38 centimeters) long, with tails that
measure 3 1/2 inches (9 centimeters). Their eyes are open at birth. They are
out learning to swim when they are only a month old, and they are weaned
by six months. A family usually consists of a mature pair of beavers and two
sets of oﬀspring. A female ﬁrst breeds when it is by their third summer the
young beavers are mature and ready to mate. They usually mate for life.
A mated pair locates a fairly deep, slowmoving stream. They dig a burrow
into the bank, starting below the surface of the water and slanting upward to
a small room above the highwater mark. This is only a temporary residence
in which the ﬁrst litter will be born in the spring. Not until the following
autumn does the couple set about building their permanent home—the lodge.
Beavers live most of their lives in or near water. They settle along banks of
streams, rivers, and lakes bordered by timberland. Large beaver populations
have been credited with reducing ﬂooding because of the dams they build
across streams. On the other hand, they may also cause the ﬂooding of roads
and woodlands because of the reservoirs of water that build up behind the
dams. Constant ﬂooding can also damage valuable timber and block routes
of migrating salmon. The reservoirs created by the dams are places where the
beavers feel secure. It is in these artiﬁcially created ponds that beavers build
their lodges and storehouses. The adult beavers select a narrow, shallow site
in the water as a place to build a dam. They gnaw down a number of aspen,
birch, or willow saplings. These they drag to the site and bury in mud with
the butt ends pointing upstream. Into this foundation the beavers ﬁt and pile
more saplings, adding mud and stones until a strong barrier is completed.
This structure allows enough seepage or overﬂow to keep the water in the
reservoir fresh. Beaver dams come in all shapes and sizes. Normally a family
of beavers can build a dam 35 feet (10 meters) long in about a week. Some
dams more than 1,000 feet (300 meters) long have been found, but these are
46 CHAPTER 1. INTRODUCTION
the work of generations of beavers.
Beavers live in a structure called a lodge. The lodge is built in the river
bank or in the pond created by the dam. From a distance, the lodge resembles
a heap of tree branches and mud. It consists of a platform of carefully in
terlaced branches held together by clay and dead leaves. When the platform
has been built up a few inches above the water, the beavers fashion a dome
shaped roof over it. Before the coming of winter, the entire structure, which
may enclose a room more than 5 feet (1.5 meters) high, must be plastered
with mud. Entrances to the beaver lodge often open underwater, so that
the animals may pass in and out below the winter ice. There are at least
two, and up to ﬁve, such entrances. A steep and narrow entrance is used
by the beavers for entry and exit. Another entrance is used for the trans
portation of wood for winter food. These underwater entrances help protect
the beavers from attacks by predators which include, in North America, the
wolverine, the lynx, and the wolf. Beavers move awkwardly on land; they
prefer to swim. When they live in ﬂat areas they sometimes build canals.
These canals allow the beavers to more easily transport the logs that are too
heavy to drag overland. A canal may measure more than 1,000 feet (300
meters) in length and from 2 to 3 feet (60 to 90 centimeters) in width.
When construction on dams, lodges, and canals is ﬁnished, the beavers
gnaw whole groves of trees and sink the wood in the pond near their lodge.
This collection forms the underwater winter ”storehouse.” Their diet consists
primarily of fresh green bark and wood such as poplar, willow, and birch.
In the summer they also eat water plants, berries, swampwood, and fruit.
Beavers do most of their building and food gathering at night. After an
autumn of toil, beavers spend the winter resting. They swim out of their
warm, dry lodges only to pluck a twig or branch from the storehouse. Because
the water in the pond is a meter deep or more, it does not freeze to the bottom
and the beaver can swim under the ice in winter time to get sticks from the
food pile accumulated the previous fall, return to the feeding chamber of the
lodge and leisurely eat the food, all in a perfect safety from predators.
Beavers’ dam is usually a high ﬂat dam and having unwittingly created
the lake which aﬀords them safety and food. The dam is quite steep on
the downstream side but slopes gently into the pond on the upstrean side.
This result is obtained because the beaver adds material to the dam from
the upstream side and, by doing so, unintentionally achieves a dam with the
strongest possible conﬁguration and therefore provides maximum security for
the colony.
1.10. BEAVERS 47
Although once plentiful throughout the wooded parts of the Northern
Hemisphere, beavers had become an endangered species by the mid19th
century. They have been hunted for their fur, their tails, and their musk
glands. Both sexes possess scent glands at the rear of the body that produce
a liquid called castoreum, which is used in perfumes. It was also a popular
medicine in the Middle Ages, apparently used to heal ailments ranging from
headaches to dropsy. The healing ability of castoreum comes from its salicylic
acid—a basic ingredient of aspirin—which the beaver acquires by eating willow
bark. The beaver has in the past been hunted for its scaly tail, which was
considered a culinary delicacy. Beaver ﬂesh in general was highly esteemed
during the Middle Ages. It is the quest for the beautiful pelt that has most
drastically reduced the beaver population. The soft, thick underfur of the
beaver, which is at its best in late winter and spring, has long been highly
valued. During the 17th, 18th, and early 19th centuries, beaver skins, made
into caps and capes, were a staple of the world’s fur trade. In fact, much of the
exploration of North America, beginning early in the 1600s and continuing
through the early 19th century, was prompted by the search for beaver fur. At
one time beaver pelts were a medium of exchange. Beaver populations have
been diminishing for centuries, possibly because of various natural causes as
well as trapping. Almost too late, conservation laws were passed throughout
the world, and the beaver was saved and was resettled in some areas where
it had once been common. The beaver population in Russia and in North
America has recovered to the extent that bans on hunting imposed earlier in
the 20th century have been slightly modiﬁed.
48 CHAPTER 1. INTRODUCTION
Chapter 2
RESERVOIR
2.1 INTRODUCTION
The ﬁrst dams built by early man were low earth or rock structures designed
to impound and divert water for agricultural use. Today’s dams are a critical
part of the nations infrastructure. They are vital for hydropower generation,
drinking water, irrigation, navigation, recreation and ﬂood control. Dams
diﬀer from other structure because of their size and their containment of
water. Special attention must be taken to understand their behaviors.. The
reservoir provides an extra eﬀect of dynamic action into the dam response.
The ultimate fate of all dams and reservoirs, is deterioration and failure or
ﬁlling by sedimentation. Every reservoir that impounds water behind a dam
is a real or potential threat to those who live and work at the downstream
side of the dam. In some locations, the eﬀects of a sever earthquake may be
dangerous to the integrity of the damreservoir structure and may tend to
the destruction of the system. Modern technology combining the knowledge
of construction and accurate design to reduce the risk that is inherent in dam
and reservoir system. The capacity of the reservoir may vary depends on the
shape of the valley and topography of the dam site. The more capacity of
the reservoir, the more threat to the area in case of a failure.
In this chapter, we will establish reservoir’s governing equation of the
motion and associated boundary conditions of the damreservoir system.
49
50 CHAPTER 2. RESERVOIR
2.2 GENERAL FORMOF RESERVOIR’S EQUA
TION OF MOTION
Fluids are composed of molecules in constant motion and collisions. A ﬂuid
has no structure. They are not distinguished by their microscopic (molecular)
structure. To take account of each molecule in a ﬂow, may be diﬃcult for
engineering purposes. Instead we are interested in average measuring of the
molecule manifestations such as density, pressure and temperature which are
their macroscopic structure. The continuum concept oﬀers a great deal of
simpliﬁcation in analysis.
Aﬂuid point (particle) represent a spatial average over some small volume
V . Consider some variable as a =
1
V
R
a
0
dV , in which a
0
has a random
characteristic(ﬁgure 2.1). The variable a can be density, ρ, velocity v, etc..
There must be a volume V such that a range of V value exist over which a
is constant. Therefore for a big V we get diﬀerent values of a for a range
of V . However if we make V successfully smaller we get same value of a.
We want δ ¿ V
1
3
¿ L, in which δ is the largest length scale of the ﬂuid
structure (say mean free path) and L is the length scale of the domain of the
problem. The restriction V
1
3
¿ L is for a to be measurable at a point while
the criteria δ ¿ V
1
3
is for a to be statistically signiﬁcant and deterministic
(always get same answer if we start from same initial condition and boundary
conditions).
2.2.1 Velocity Field
In a deformable system there are an inﬁnite number of particles. In order to
deﬁne the velocity of a particle we must deﬁne the position of the particles,
spatial coordinates, at the speciﬁc time. Using this, the velocity of all parti
cles can be as v(x, y, z, t) which is consisted of three components of velocity
v
x
, v
y
, v
z
in x, y and z direction of coordinates, respectively. v is called
ﬁeld velocity vector. For steady ﬂow, the value of the velocity at a position
remains invariant with time, v = v(x, y, z).
LAGRANGIAN AND EULERIAN VIEWPOINTS
In investigation of the ﬂuid motion, two procedures may be used to study
ﬂuid particles. We can stipulate ﬁx coordinates x
1
, y
1
and z
1
in the velocity
ﬁeld functions and letting time pass. we can express velocity of function
2.2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 51
Figure 2.1: Fluid point
at any time passing this position as v(x
1
, y
1
, z
1
, t). Using this approach we
study the motion of a continuous string of particles which pass the ﬁxed
point. This viewpoint is called the Eulerian viewpoint.
On the other hand, we can study the motion of a single particle in the
ﬂow by following the particle in its path. This approach means continuous
variation of the x, y and z to locate the particle. This approach is called the
Lagrangian viewpoint where v = v(x, y, z). The x, y and z can be expressed
as a function of time x = x(t), y = y(t) and z = z(t) (ﬁgure 2.2).
The Lagrangian viewpoint is used mainly in particle mechanics. In contin
uum mechanics this method requires the description of motion of an inﬁnite
number of particles and thus becomes extremely cumbersome. The Eulerian
viewpoint is easier to use in continuum mechanics because it is concerned
with the description of motion at a ﬁxed position.
In the partial time derivative, from the Eulerian point of view, suppose we
are standing on a bridge and note how the concentration of ﬁsh (c) just below
us changes with time. In this case, the position is ﬁxed in space, therefore
by
∂c
∂t
we mean partial of c with respect to t, holding x, y and z constant. We
can also assume ourselves on a boat and moving with around the river. The
change of ﬁsh concentration with respect to time reﬂects the motion of the
boat as well as time variation. This is called total time variation and can be
52 CHAPTER 2. RESERVOIR
written as:
Dc
Dt
=
∂c
∂t
+
∂c
∂x
dx
dt
+
∂c
∂y
dy
dt
+
∂c
∂z
dz
dt
in which
dx
dt
,
dy
dt
and
dz
dt
are the components of the velocity of the boat. Now if
we turn oﬀ the engine of the boat and ﬂoat along counting ﬁsh (the lagrangian
viewpoint), the velocity of the boat would be simply the velocity of the
streams. The change of ﬁsh concentration in this case is called substantial
time derivative and is written as follows:
Dc
Dt
=
∂c
∂t
+v
x
∂c
∂x
+v
y
∂c
∂y
+v
z
∂c
∂z
in which v
x
, v
y
and v
z
are the components of the local ﬂuid velocity v.
Figure 2.2: a)Lagrangian viewpoint b)Eulerian viewpoint
2.2.2 System and Control Volume
An identiﬁed quantity of matter is called system. It may undergo shape
changes, position changes or temperature changes, but entailed the same
2.2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 53
matter. On the other hand, control volume is a deﬁnite volume in space and
its boundary called control surface. The amount of matter inside control
volume may change but its shape is ﬁxed. the concept of system and control
volume are related to the Lagrangian and Eulerian viewpoints.
2.2.3 Reynold’s Transport Equation
Consider N as properties of a substance whose measure depends on the
amount of the substance presented in a system. An arbitrary ﬂow ﬁeld of
v = v(x, y, z, t)is observed from some reference xyz. We consider a system
of ﬂuid with ﬁnite mass at time t and t + ∆t (ﬁgure 2.3). The streamlines
correspond to those at time t. The control volume is considered with system
at time t and is ﬁxed.
The distribution of N per unit mass will be given as η, such that N =
RRR
ηρdV with dV representing an element of volume with mass density of
ρ. We are trying to establish a relation between the rate of change of N in
system with variation of this property inside the control volume.
The system at time t and t+∆t can be divided into three regions. Region
II is common to the system at both time t and t +∆t. Rate of change of N
with respect to time for the system can be written as:
µ
dN
dt
¶
system
=
DN
Dt
=
lim
∆t→0
¡RRR
III
ηρdV +
RRR
II
ηρdV
¢
t+∆t
−
¡RRR
I
ηρdV +
RRR
II
ηρdV
¢
t
∆t
The above equation can be arranged as following:
DN
Dt
= lim
∆t→0
Ã¡RRR
II
ηρdV
¢
t+∆t
−
¡RRR
II
ηρdV
¢
t
∆t
!
+ (2.1)
lim
∆t→0
¡RRR
III
ηρdV
¢
t+∆t
∆t
− lim
∆t→0
¡RRR
I
ηρdV
¢
t
∆t
In the ﬁrst term on the right hand side of equation. 2.1, we can say that,
as ∆t goes to zero the region II becomes that of control volume ,cv,. Then:
54 CHAPTER 2. RESERVOIR
Figure 2.3: Moving system
2.2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 55
lim
∆t→0
Ã¡RRR
II
ηρdV
¢
t+∆t
−
¡RRR
II
ηρdV
¢
t
∆t
!
=
∂
∂t
ZZZ
cv
ηρdV (2.2)
The second term in equation 2.1 shows the amount of the property N
that has crossed part of the control surface, cs, shown as ARB. Therefore,
the second term is the average of eﬄux of N across ARB during time interval
of ∆t. As ∆t goes to zero this becomes exact rate of eﬄux of N through the
cs. Similarly, the third term approximates the amount of N that has passed
into the cv during ∆t through remaining of the cs (inﬂux). Thus, the lase
two integrals give the net rate of eﬄux of N from cv at time t as:
lim
∆t→0
¡RRR
III
ηρdV
¢
t+∆t
∆t
− lim
∆t→0
¡RRR
I
ηρdV
¢
t
∆t
= Net efflux rate of N fromcs
(2.3)
In conclusion, the rate of change of N for system at time t are sum of
ﬁrst, rate of change of N inside cv having the shape of the system at time t
(equation 2.2)and second, the net rate of eﬄux of N through the cs at time
t (equation 2.3).
In equation 2.3, the term dV can be written as:
dV = v.dAdt (2.4)
which is the volume of ﬂuid that has crossed the cs in time dt. dA is the
normal outward vector on control surface. Multiplying (equation 2.4) by ρ
and dividing by dt then gives the instantaneous rate of mass ﬂow of ﬂuid
ρv.dA leaving the control volume through the indicated area dA. For the
ﬂuid entering the control volume the expression dV = v.dAdt is negative
while for the ﬂuid going out of the control volume the expression is positive.
Therefore equation 2.3 can be written as follows:
Efflux rate through cs ≈
ZZ
ARB
ηρv.dA
Influx rate through cs ≈ −
ZZ
ALB
ηρv.dA
Net efflux rate through cs ≈
ZZ
ARB
ηρv.dA+
ZZ
ALB
ηρv.dA
56 CHAPTER 2. RESERVOIR
In the limit as ∆t →0, the approximation becomes exact, so we can express
the right side of the above equation as
RR
cs
η(ρv.dA), where the integral is
the closed integral over the entire control surface.
Now the equation 2.1 can be written as:
DN
Dt
=
ZZ
cs
η(ρv.dA)+
∂
∂t
ZZZ
cv
ηρdV (2.5)
This is called the Reynold Transport equation and it is a change from system
approach to control volume approach. In the Reynole transport equation v is
measured relative to some reference xyz and the control volume was ﬁxed in
this reference. Thus, v in the above equation is measured relative to control
volume. As a consequences, the time rate of change of N is in eﬀect observed
from the control volume and velocities and time rates of changes are those
seen from control volume. Since we could use a reference xyz having an
arbitrary motion, it means that the control volume can have any motion.
Thus the Reynold transport equation will then instantaneously be correct if
we measure the time derivatives and velocities relative to the control volume,
no matter what the motion of the control volume may be.
2.2.4 Continuity Equation
A system always entails the same quantity of matter. Therefore, the mass
M would be constant. To go from system approach to control approach, we
use Reynold transport equation in which N is for our case M, the mass of a
ﬂuid system and it is M =
RRR
ρdV . Therefore, η = 1. Thus for a constant
mass in a system at any time t, we can write the Reynold transport equation
as following:
DM
Dt
= 0 =
ZZ
cs
(ρv.dA)+
∂
∂t
ZZZ
cv
ρdV
The velocity v and the time derivative are measured relative to the control
volume. Above equation can be written in the following form:
ZZ
cs
(ρv.dA) = −
∂
∂t
ZZZ
cv
ρdV (2.6)
Equation 2.6 which is the ﬁnal form of the continuity equation expresses
that the net eﬄux rate of mass through the control surface equals the rate of
2.2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 57
decrease of mass inside the control volume. Two special cases of continuity
can be considered. The case of steady ﬂow in which all ﬂuid properties at any
ﬁxed position in the reference must remain invariant, the continuity equation
can be written as:
ZZ
cs
(ρv.dA) = 0
The other case is the case of incompressible ﬂuid having constant ρ.The
continuity equation can be written as:
ZZ
cs
v.dA = 0
A more detailed deﬁnition of compressibility is brought in the proceeding
sections.
2.2.5 Linear Momentum Equation
If we take linear momentum as parameters N which is used as a general term
(vector and scalar) in the Reynold transport equation, we have N = P = mv.
The term η in this case would become momentum per unit mass, or simply v,
then, P =
RRR
v(ρdV ). Then the Reynold transport equation can be written
as followings:
DP
Dt
=
ZZ
cs
v(ρv.dA)+
∂
∂t
ZZZ
cv
vρdV (2.7)
The equation of linear momentum is a vector equation and can be divided
into its three components.
Newton’s second law states that:
X
F =
d
dtsystem
·ZZZ
vdm
¸
=
dP
dt system
In above equation,
P
F if the resultant of the all external forces acting on
the system and v the time derivative are taken from the inertial references.
Since, the second law of Newton is based on the Eulerian viewpoint, we are
following a system. Thus, it can be written as:
58 CHAPTER 2. RESERVOIR
X
F =
D
Dt
·ZZZ
vdm
¸
=
DP
Dt
There are two types of the external forces. Surface forces, T(x, y, z, t) given
as forces per unit area on the boundary surface, bs. Force acting on the
material inside the boundary, ib are called body force, B(x, y, z, t) given as
force per unit mass. Thus, The above equation can be written as followings:
ZZ
bs
TdA+
ZZZ
ib
BρdV =
DP
Dt
(2.8)
Equation 2.8 is the Newton’s second law for a ﬁnite system. Knowing
that the control volume and system are coincident at time t, if we ﬁx the
control volume in inertial space, then the derivatives in the right hand side
is taken from a inertial references and we may use the Newton’s second law
(equation 2.8) to replace it in the Reynold equation and then we get:
ZZ
cs
TdA+
ZZZ
cv
BρdV =
ZZ
cs
v(ρv.dA)+
∂
∂t
ZZZ
cv
vρdV (2.9)
Equation 2.9 means that: Sum of surface forces acting on the control
surface and body forces acting on the control volume are equal with the sum
of the rate of the eﬄux of the linear momentum across the control surface
and the rate of increase of linear momentum inside the control volume.
Equation 9 is a vector equation and can be written for each components.
For example in x direction we can write it as follows:
ZZ
cs
T
x
dA+
ZZZ
cv
B
x
ρdV =
ZZ
cs
v
x
(ρv.dA)+
∂
∂t
ZZZ
cv
v
x
ρdV (2.10)
In this book for simplicity, the positive direction of the reference axis xyz
are the same as the positive direction of the velocity components as well as
the surface and body force components. Sign of v.dA is independent of the
sign consideration.
2.2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 59
2.2.6 The Equation of the Motion
The continuity equation and the linear momentum equation was developed
for ﬁnite systems and their were related to control volume. The equations
give an average values of quantities or only components of resultant forces.
They don’t give any detail information of the ﬂow everywhere inside the
control volume. Here, we are trying to ﬁnd out diﬀerential equations valid
at any point of ﬂuid.
Writing continuity equation in the form of diﬀerential equation, we con
sider an inﬁnitesimal control volume in the shape of a rectangular paral
lelepiped ﬁxed in xyz for a general ﬂow v(x, y, z) measured relative to xyz.
Figure 2.4: Volume element
The net eﬄux rate along the plane of control volume that are perpendicu
lar to the x axis (x planes of the element) is, ρv
x

x+∆x
∆y∆z −ρv
x

x
∆y∆z
(ﬁgure 2.4) If we write the similar equation for the axis in y and z direction
and add them up we get the net eﬄux rate as following:
60 CHAPTER 2. RESERVOIR
Net eﬄux rate = [ρv
x

x+∆x
∆y∆z −ρv
x

x
∆y∆z]
+[ρv
y

y+∆y
∆x∆z −ρv
y

y
∆x∆z]
+[ρv
z

z+∆z
∆x∆y −ρv
z

z
∆x∆y]
The rate of mass accumulation inside the control volume would be
∂ρ
∂t
∆x∆y∆z.
Equating the net eﬄux rate and the rate of decrease of mass inside the con
trol volume and dividing all resulting equation by ∆x∆y∆z and taking the
limit as ∆x, ∆y and ∆z approach zero, we get:
∂(ρv
x
)
∂x
+
∂(ρv
y
)
∂y
+
∂(ρv
z
)
∂z
= −
∂ρ
∂t
(2.11)
This is called diﬀerentia continuity equation and one special cases of it would
be case of steady ﬂow which results in:
∂(ρv
x
)
∂x
+
∂(ρv
y
)
∂y
+
∂(ρv
z
)
∂z
= 0
The other case of incompressible ﬂow would be as following:
∂v
x
∂x
+
∂v
y
∂y
+
∂v
z
∂z
= 0
Introducing the divergence operator which for a vector ﬁeld v can be
deﬁned as:
∇.v =
∂v
x
∂x
+
∂v
y
∂y
+
∂v
z
∂z
Then we can write the general form of the continuity equation 2.11 in the
following form:
∇.(ρv) = −
∂ρ
∂t
Writing the linear momentum equation in the form of diﬀerential equa
tion, we consider an inﬁnitesimal system in the shape of a rectangular paral
lelepiped ﬁxed in xyz (ﬁgure 2.5 ). The linear momentum equation is a form
of Newton’s second law which has an Eulerian viewpoint. Consider a volume
element ∆x∆y∆z as shown in ﬁgure2.5. The surface stress on each side of
2.2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 61
Figure 2.5: Rectangular Parallelpiped element
a cubic element cab be written as T
ij
, i, j = x, y, z where i is an indication
of the side, plane, in which the stress is located on and j is the direction of
the stress. For example T
xy
is the components of the stress on a plane normal
to x axis and directed toward y axis. It can be shown that for equilibrium
we must have T
ij
= T
ji
. As it was already mentioned equation of linear
momentum is a vector equation (equation 2.9). Thus we can write for each
components of it. Here, we start by considering the x direction(equation
2.10). The ﬁrst term on the left hand side of the equation 2.10 is all of
the surface forces acting on the element at the x direction. The resultant
forces acting on the x planes along the x directions is T
xx

x+∆x
∆y∆z −T
xx

x
∆y∆z. The resultant of forces acting on the y plane along the x direction
is T
yx

y+∆y
∆x∆z − T
yx

y
∆x∆z. Similarly, for the forces on the z plane
along the x direction is T
zx

z+∆z
∆x∆z −T
zx

z
∆x∆y. If we add all of the
surface forces we will get:
[T
xx

x+∆x
∆y∆z −T
xx

x
∆y∆z] +
[T
yx

y+∆y
∆x∆z −T
yx

y
∆x∆z] +
62 CHAPTER 2. RESERVOIR
[T
zx

z+∆z
∆x∆z −T
zx

z
∆x∆y]
The second term on the left hand side of the equation 2.10 is the body
force and is:
ρg
x
∆x∆y∆z
where g
x
is the component of the gravitational force along the x axis.
The ﬁrst term on the right hand side of equation 2.10 is the net rate of
eﬄux. the rate at which x component of momentum enters at x is v
x
ρv
x

x
∆y∆z and the rate at which it leaves at x +∆x is v
x
ρv
x

x+∆x
∆y∆z. The
same can be written for the y and z planes. Thus if we add them up we get:
[v
x
ρv
x

x+∆x
−v
x
ρv
x

x
] ∆y∆z +
[v
x
ρv
y

y+∆y
−v
x
ρv
y

y
] ∆x∆z +
[v
x
ρv
z

z+∆z
−v
x
ρv
z

z
] ∆x∆y
The second term on the right hand side of the equation 2.10 is the accumu
lation of the momentum inside the element and is:
∂(ρv
x
)
∂t
∆x∆y∆z
If we substitute all the terms into the equation 2.10 from the above equations
and dividing the entire resulting equation by ∆x∆y∆z and taking limit as
∆x, ∆y and ∆z approaches zero, we obtain the xcomponent of the equation
of the motion as:
∂T
xx
∂x
+
∂T
xy
∂y
+
∂T
xz
∂z
+ρg
x
=
∂(v
x
ρv
x
)
∂x
+
∂(v
x
ρv
y
)
∂y
+
∂(v
x
ρv
z
)
∂z
+
∂(ρv
x
)
∂t
(2.12)
Similarly, for the components in the xx and y direction it can be written as:
∂T
xy
∂x
+
∂T
yy
∂y
+
∂T
zy
∂z
+ρg
y
=
∂(v
y
ρv
x
)
∂x
+
∂(v
y
ρv
y
)
∂y
+
∂(v
y
ρv
z
)
∂z
+
∂(ρv
y
)
∂t
(2.13)
∂T
xz
∂x
+
∂T
yz
∂y
+
∂T
zz
∂z
+ρg
z
=
∂(v
z
ρv
x
)
∂x
+
∂(v
z
ρv
y
)
∂y
+
∂(v
z
ρv
z
)
∂z
+
∂(ρv
z
)
∂t
(2.14)
2.2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 63
The quantity ρv
x
, ρv
y
and ρv
z
are the components of the ρv ,mass veloc
ity vector. The terms ρv
x
v
x
, ρv
x
v
y
, ρv
x
v
z
, ρv
y
v
z
, etc. are the nine com
ponents(six components because of symmetry) of the momentum ﬂux ρvv,
which is the dyadic product of ρv and v. Similarly, T
xx
, T
xy
, T
xz
, T
yx
, etc.
are the nine components( six components because of symmetry) of the T,
known as the stress tensor. We can combine, the above three equations and
write them in a vector format as followings:
∇.T+ ρg = ∇.ρvv +
∂ρv
∂t
(2.15)
There should be noticed that ∇.ρvv and ∇.T are not simple divergence
because of tensorial nature of the ∇.ρvv and ∇.T. The above equation is
the general form of the linear momentum equation.
If we use the equation of continuity to substitute it in the equation 2.12,
we will get:
∂T
xx
∂x
+
∂T
yx
∂y
+
∂T
zx
∂z
+ ρg
x
= ρ
Dv
x
Dt
The same can be written for y and z components. When three components
added together vectorially, we get:
ρ
Dv
Dt
 {z }
mass per unit volume
time acceleration
= ∇.T
{z}
surface forces on element
per unit volume
+ ρg
{z}
body forces on element
per unit volume
(2.16)
This is an statement of the Newton’s second law which was developed for
a volume of ﬂuid element moving with acceleration because of the forces
acting upon it. It can be seen that momentum balance is equivalent to New
ton’s second law of motion. The above equation is valid for any continuous
medium. In equation 2.16 we can insert expressions for various stresses in
forms of velocity gradient and pressure. Therefore, we need to ﬁnd a relation
between various stress and velocity gradient..
64 CHAPTER 2. RESERVOIR
2.3 VISCOSITY
A ﬂuid with zero viscosity is called a nonviscous, or inviscid ﬂuid. Lami
nar ﬂow can be described as a wellordered pattern whereby ﬂuid layers are
assumed to slide over one another. In this case while ﬂuid has irregular
molecular motion, is macroscopically, well ordered ﬂow. For a laminar ﬂow,
whereby ﬂuid particles move in straight, parallel lines, called Newtonian ﬂu
ids, the shear stress on an interface tangent to direction of ﬂow is proportional
to the distance rate of change of velocity, wherein diﬀerential is taken in a
direction normal to the interface, T
xy
∝
∂v
x
∂y
or T
xy
= µ
∂v
x
∂y
, Where µ is called
the coeﬃcient of viscosity with the dimension of (F/L
2
)t (ﬁgure 2.6).
Figure 2.6: Wellordered parallel ﬂow
This unit in a system of centimeter (cm)gram (g)second(sec)is g −
cm
−1
− sec
−1
and is called the poise. At room temperature, µ is about 1
centipoise for water and about 0.002 centipois for air. This is wellknown
Newton’s viscosity law. All gases and simple liquids are described by the
later formula. For gases at low density viscosity decrease as temperature
increases. Whereas for liquids the viscosity usually decreases with increasing
temperature. The reason may be that, in gases, molecules travel long dis
tance between collisions and the momentum is transported by the molecules
in free ﬂight, which as in liquids the molecules travel only very short dis
tances between collisions, the principal mechanism for momentum transfer is
the actual colliding of the molecules.
In Newtonian law of viscosity for a given temperature and pressure µ is
constant. Experiments show that in some ﬂuids, T
xy
is not proportional to
the change of rate of velocity in normal direction of it. These few industrial
2.4. NAVIERSTOKES AND EULER EQUATIONS 65
important materials, that are not described by the Newtonian law of viscosity
are referred to nonnewtonian ﬂuids. In this case µ may be described as a
function of
∂vx
∂y
or T
xy
.
In more general ﬂows, there are more general relations between the stress
ﬁeld and the velocity ﬁeld( namely constitutive law). Here, we consider
Stoke’s viscosity law. This is based on assumption that each stress is lin
early related through a set of constants to each of the six strains rates
(
.
²
ij
, i, j = x, y, z 
.
²
ij
=
.
²
ji
). In addition each normal stress is directly related to
pressure p. The constants are called viscosity coeﬃcients and ﬂuids behaving
accordingly to this relation are called Newtonian ﬂuid. The Stoke’s viscos
ity law degenerates to Newton’s viscosity law for the special case of parallel
ﬂow. Knowing that most Newtonian ﬂuids have ﬂow properties which are
independent of direction of coordinates and also considering that for most
ﬂuid p = − σ= −
1
3
(T
xx
+T
yy
+T
zz
), we reach the following relationships:
T
xx
= µ(2
∂v
x
∂x
−
2
3
∇.v) −p
T
yy
= µ(2
∂v
y
∂y
−
2
3
∇.v) −p
T
zz
= µ(2
∂v
z
∂z
−
2
3
∇.v) −p
T
xy
= µ(
∂v
x
∂y
+
∂v
y
∂x
)
T
xz
= µ(
∂v
x
∂z
+
∂v
z
∂x
)
T
yz
= µ(
∂v
y
∂z
+
∂v
z
∂y
)
This is a common form of Stoke’s viscosity law.
2.4 NAVIERSTOKES ANDEULEREQUA
TIONS
Substituting the Stoke’s viscosity law into the equations of the linear momen
tum (equations 2.12, 2.13, 2.14) for a newtonian ﬂuid with varying density
and viscosity, we get:
66 CHAPTER 2. RESERVOIR
ρ
Dv
x
Dt
= (2.17)
−
∂p
∂x
+
∂
∂x
·
2µ
∂v
x
∂x
−
2
3
µ(∇.v)
¸
+
∂
∂y
·
µ(
∂v
x
∂y
+
∂v
y
∂x
)
¸
+
∂
∂z
·
µ(
∂v
z
∂x
+
∂v
x
∂z
)
¸
+B
x
ρ
Dv
y
Dt
= (2.18)
−
∂p
∂y
+
∂
∂x
·
µ(
∂v
y
∂x
+
∂v
x
∂y
)
¸
+
∂
∂y
·
2µ
∂v
y
∂y
−
2
3
µ(∇.v)
¸
+
∂
∂z
·
µ(
∂v
z
∂y
+
∂v
y
∂z
)
¸
+B
y
ρ
Dv
z
Dt
= (2.19)
−
∂p
∂z
+
∂
∂x
·
µ(
∂v
z
∂x
+
∂v
x
∂z
)
¸
+
∂
∂y
·
µ(
∂v
z
∂y
+
∂v
x
∂z
)
¸
+
∂
∂z
·
2µ
∂v
z
∂z
−
2
3
µ(∇.v)
¸
+B
z
These equations, along with the equation of continuity, the equation of
p = p(ρ), µ = µ(ρ), and the boundary and initial conditions determine com
pletely the pressure, density and velocity component in a ﬂowing isothermal
ﬂuid.
For constant ρ and constant µ the equation may be simpliﬁed by using
the equation of continuity (∇.v = 0) to results in:
ρ
Dv
Dt
= −∇p +µ∇
2
v +B
The operator ∇
2
=
∂
2
∂x
2
+
∂
2
∂y
2
+
∂
2
∂z
2
is called the Laplacian operator. The
above equation is called ”NavierStokes equation which ﬁrst developed by
Navier in France.
For inviscid ﬂuids, the equations 2.17, 2.18 and 2.19 will result in:
ρ
Dv
Dt
= −∇p +B
This is Euler equation which is widely used for describing ﬂow systems in
which viscous eﬀects are relatively unimportant.
2.5. COMPRESSIBLE FLUID 67
2.5 COMPRESSIBLE FLUID
Despite the existence of large pressure, ﬂuids undergo very little change in
density. Water needs a pressure change of 20, 000kPa to have 1 percent
change in its density. Fluids of constant density are termed incompressible
ﬂuids and assumed during computation the density is constant. A ﬂowing
ﬂuid is said to be compressible when appreciable density changes are brought
about the motion. The variation of density is usually accompanied by tem
perature changes as well as heat transfer. The variation of density means
that a group of ﬂuid elements can spread out into a larger region of space
without requiring a simultaneous shift to be made of all ﬂuid elements in the
ﬂuid. In compressible media a small shift of ﬂuid element will induce similar
small movements in adjacent elements and by this a disturbance called an
acoustic wave propagates at a relatively high speed through the medium. In
the incompressible ﬂows, these propagation have inﬁnite speed which adjust
ment (disturbance) took place simultaneously through the entire ﬂow and
there are now wave to be considered. Thus, compressibility means admission
of elastic waves having ﬁnite velocity.
Up to now, we needed four scalar equations (the equation of continuity
and three components of the equation of linear momentum) to describe fully
a ﬂow ﬁeld. For compressible ﬂow, the density and pressure changes are also
accompanied by temperature changes. Thus we need equation of energy and
equation of state(ρ = ρ(p, T)). The general theory of compressible ﬂow is
very complicated, not only because of large number of the equations involved,
but also because of the wave propagation phenomena that are predominant
at ﬂow speed higher than the speed of sound.
The speed of sound is deﬁned as the rate of propagation of an inﬁnitesi
mal pressure disturbance(wave) through a continuous medium. Sound is the
propagation of compressible an expansion wave of ﬁnite but small amplitude
such that the ear can detect them. Frequencies range from 20 to 20, 000
Hertz while the magnitude is typically less than 10 Pa.
Consider a long tube ﬁlled with motionless ﬂuid and having a piston at
one end, as shown in ﬁgure 2.7. By tapping the piston we may cause a
pressure increase dp on the right of A−A. Now two things will happen. Due
to molecular action, the pressure will increase to the left of A −A and this
increase in pressure will move in the tube at high speed C. We thus have a
pressure wave of speed V moving to the left due to microscopic action. The
second eﬀect is on the macroscopic level. According to Newton’s second law,
68 CHAPTER 2. RESERVOIR
the ﬂuid just to the left of wavefront described above must accelerate as a
result of the pressure diﬀerence dp to a velocity dv. Once the pressure rise
dp has been established in the ﬂuid, there is no further change in velocity,
so it remains at dv. Behind the wavefront, the ﬂuid is thus moving to the
left at speed dv. During an interval of dt the wave has progressed a distance
Cdt and is shown at position B in ﬁgure 2.7. Meanwhile, ﬂuid particle at A
move a distance dvdt to position A
1
. At an intermediate position, such as
halfway between A and B , shown in diagram as D, the ﬂuid velocity dv has
persisted for a time interval
dt
2
. Consequently, ﬂuid initially at D has moved
a distance
dvdt
2
to position D
1
.
Figure 2.7: Wave front movement and ﬂuid movement
By tapping, as inﬁnitesimal pressure disturbance which will move down
the tube at a constant speed, the ﬂuid behind the wave is slightly compressed,
while the ﬂuid ahead of the wave remains undisturbed. This is an unsteady
state problem. However, if we assume an inﬁnitesimal control volume around
the wave, travelling with the wave velocity C, we can apply a steady state
analysis. The wave is thus stationary while the ﬂuid ﬂows with an approach
velocity C, as shown in ﬁgure 2.8. Neglecting friction eﬀects, the velocity
proﬁle can be assumed ﬂat. The continuity equation may be written as:
ρAC = (ρ +dρ)(C −dv)A
Neglecting inﬁnitesimal quantities of higher order (dρdv) gives:
Cdρ −ρdv = 0 (2.20)
2.5. COMPRESSIBLE FLUID 69
Figure 2.8: Moving pressure disturbance in a motionless ﬂuid and ﬁxed wave
in a moving ﬂuid
70 CHAPTER 2. RESERVOIR
Applying linear momentum balance to the control volume, we get:
0 = C(ρAC) −(C −dv)ρAC +pA−(p +dp)A
Which further reduces to:
dp = ρCdv (2.21)
By combining equations 2.20 and 2.21 we get:
C
2
=
dp
dρ
(2.22)
This expression gives the velocity of propagation of a compressible wave.
Since the propagation of inﬁnitesimal expansion and compression waves is
called sound, C is then the speed of sound. the above equation is valid for
any continuum, be it a solid, a liquid, or a gas. Because of the high speed of
travel of the wave there is very little time for any signiﬁcant heat transfer to
take place, so the process is very nearly adiabatic. It should be noted that C
is measured relative to the ﬂuid in which the front is propagating. We also
assumed a constant value of C, that is we used an inertial control volume with
steady ﬂow relative to this control volume. The wave involve inﬁnitesimal
pressure variation. Wave with comparatively large pressure variation over a
very narrow front are called shock waves. these waves move relative to ﬂuid
at speed in excess of the acoustic speed. Then we may consider acoustic
wave as limiting cases of shock waves where the change in pressure across
wave become inﬁnitesimal. Therefore what we have developed is valid for
any weak spherical and cylindrical waves which moves caused by any small
disturbance of pressure.
If a continuum were incompressible, equation 2.22 would give an inﬁnite
speed of sound. However, no actual liquid or solid can be perfectly incom
pressible. All materials have a ﬁnite speed of sound. For example the speed
of sound in water, air, ice and steel at 15
◦c
and a pressure of 101.325 kPa
are 1490,340, 3200 and 5059 m/s, respectively.
For liquids and solids it is customary to deﬁne the bulk modulus as a
parameter relating the volume (or density) change to the applied pressure
change:
K = −
∆p
∆V
V
=
∆p
∆ρ
ρ
= ρ
dp
dρ
(2.23)
2.6. BOUNDARYLAYER THEORY 71
Water( at 20
◦c
and atmospheric pressure) has a bulk modulus of about 2.2×
10
6
Pa while steel about 200×10
6
Pa. The bulk modulus related to young’s
modulus of elasticity E by the expression:
K
E
= 3(1 −2ν)
Where ν is the Poisson’s ratio. For many common metals such as steel and
aluminum ν ≈
1
3
and E = K.
By combining equations 2.22 and 2.23 we get:
C =
s
K
ρ
2.6 BOUNDARYLAYER THEORY
For ﬂuids with small viscosity such as air and water with a high degree of
accuracy, we can consider frictionless ﬂow over entire ﬂuid except for their
regions around the contact areas. Here, because of high velocity gradients, we
could not properly neglect frictions (Newton’s viscosity law) so we consider
these regions apart from the main ﬂow, terming the boundary layers.
In 1904 Ludwing Prandti introduced the concept of boundary layer and
showed how NavierStokes equations could be simpliﬁed. This concept lit
erally revolutionized the science of ﬂuid mechanics. According to Prandti’s
boundary layer concept viscous eﬀects at high Reynolds number (Re =
ρvL
µ
=
Inertia Forces
V iscous Forces
=
v
2
L
µv/ρL
2
) are conﬁned in thin layers adjacent to solid bound
aries. Outside these solid boundary layers the ﬂow may be considered inviscid
(µ = 0) and can, thus be described by Euler equation. Within the boundary
layer the velocity component in the main ﬂow direction (x) changes from
v
x
= 0 (at the solid boundary) to v
x
= v
∞
(the free stream velocity at the
edge of the boundary layer).
We consider qualitatively the boundary layer ﬂow over a ﬂat plate. As it
is shown in ﬁgure 2.9, a laminar region begins at the leading edge and grows
in thickness. A transition region is reached where the ﬂow changes from
laminar to turbulent, with a consequent thickening of the boundary layer.
The transition depends partly on the Reynold number(
ρv∞x
µ
) where x is the
distance downstream from the leading edge. Transition occurs in the range
Re
x
= 3 ×10
5
to Re
x
= 10
6
. In the turbulent region we ﬁnd, that as we get
72 CHAPTER 2. RESERVOIR
Figure 2.9: Details of boundary layers
near the boundary the turbulence becomes suppressed to such a degree that
viscous eﬀects predominate, leading us to formulate the concept of a viscous
sublayer shown darkened in the diagram.
There is actually a smooth variation from boundary layer region to the
region of constant velocity. The velocity proﬁle merges smoothly into the
mainstream proﬁle. There are several deﬁnitions of boundarylayer thick
ness that are quite useful. One can consider that the thickness is the distance
δ from the wall out to where the ﬂuid velocity is 99 percent of the mainstream
velocity. The experimental determination of the boundary layer thickness as
deﬁned above is very diﬃcult because the velocity approaches asymptoti
cally the free stream value v
∞
. The edge of the boundary layer is poorly
deﬁned. For this reason alternative thicknesses which can be measured more
accurately are often used. The displacement thickness δ
∗
deﬁned as distance
by which the boundary would have to be displaced if the entire ﬂow were
imagined to be frictionless and the same mass ﬂow maintains at any section
(ﬁgure 2.10). Thus, considering a unit width along the z across an inﬁnite
ﬂat plate at zero angle of attack for incompressible ﬂuid:
∞
Z
0
v
x
dy = q =
∞
Z
δ
∗
v
m
dy
Hence, changing the lower limit on the second integral and solving for δ
∗
we
get:
2.6. BOUNDARYLAYER THEORY 73
δ
∗
=
∞
Z
0
(1 −
v
x
v
m
)dy
Figure 2.10: Displacement thickness in boundary layer
The motivation for using displacement thickness is to permit the use of
a displaced body in place of the actual body, such that the frictionless mass
ﬂow around the displaced body is the same as the actual mass ﬂow around
the real body. Use is made of the displacement thickness in the design of
wind tunnels, air intakes for airplane jet engine.
As the boundary layer thickness growth, the ﬂuid is pushed away from the
plate, which means that the velocity has also a component in the y direction
, v
y
. we would expect that v
y
¿ v
x
because the boundary layer thickness
is small compared to any signiﬁcant body dimension. The boundary layer
thickness, that is the value of y for which v
x
= 0.99v
∞
, can be obtained as
following:
δ ≈ 5.0
r
µx
ρv
m
=
5x
√
Re
x
A rough comparison between the various thicknesses gives:
δ
∗
=
δ
3
The eﬀects of viscosity is inﬂuential as δ
∗
increases. In the case of concrete
dams since δ
∗
/h ratio is very small, the viscous eﬀect can be neglected,
specially for the sudden ground acceleration on the base of the reservoir.
74 CHAPTER 2. RESERVOIR
2.7 IRROTATIONAL FLOW
A ﬂuid element can be subjected to three types of ﬂow, namely: translation,
rotation and deformation. These types of elemental motions are depicted
in ﬁgure 2.11. The concept of translational motion is selfevident.. The
deformational motion will occur when relative orientation of the axis changes
as shown in ﬁgure 2.11. Deformational ﬂow also exists if one or more axis
are stretched or compressed. Here we focus our attention on rotational ﬂow,
which is depicted by the ﬂuid element of ﬁgure 2.11 turning around.
Figure 2.11: Three types of ﬂuid motion
Let us examine the ﬂow in a ﬂuid with circular streamlines as shown in
ﬁgure 2.12. The ﬂuid rotates like a rigid body. each element turns around at
a certain angular velocity. The arrows shown rotate at the same rate. This
is unquestionably a case of rotational ﬂow.
Now let us consider the ﬂow between the two horizontal ﬂat plates, where
the bottom plate is stationary while the top one moves at velocity v
0
as shown
in ﬁgure 2.13. we note that the horizontal arrow is simply translated, while
2.7. IRROTATIONAL FLOW 75
Figure 2.12: Fluid rotating like a rigid body
Figure 2.13: Shearing ﬂow between two ﬂat plates
76 CHAPTER 2. RESERVOIR
the vertical one turns. It is not clear wether this is a case of rotational ﬂow
or not. To determine that we propose to use the average rate at rotation of
the two arrows as a measure. If the average rate of rotation is zero the ﬂow
is said to be irrotational, if not the ﬂow is rotational. For generality we refer
to ﬁgure 2.14.
Figure 2.14: Change of relative positions in an arbitrary ﬂow ﬁeld
We deﬁne the angular velocity ω
z
about the axis z as the average rate of
counterclockwise rotation of the two lines:
ω
z
=
1
2
(
dα
dt
−
dβ
dt
)
In the limit of small angles we would have:
dα
dt
=
d(
∆y
α
∆x
)
dt
=
d(
∆y
α
∆t
)
dx
=
∂v
y
∂x
dβ
dt
=
d(
∆x
β
∆y
)
dt
=
d(
∆x
β
∆t
)
dy
=
∂v
x
∂y
2.7. IRROTATIONAL FLOW 77
Thus,
ω
z
=
1
2
(
∂v
y
∂x
−
∂v
x
∂y
)
Similarly:
ω
x
=
1
2
(
∂v
z
∂y
−
∂v
y
∂z
)
ω
y
=
1
2
(
∂v
x
∂z
−
∂v
z
∂x
)
The vector ω is thus onehalf of the vorticity vector:The vector ω is thus
onehalf of the vorticity vector:
ω =
1
2
∇×v =
1
2
i j k
∂
∂x
∂
∂y
∂
∂z
v
x
v
y
v
z
¸
¸
To avoid the factor
1
2
we deﬁne the vorticity ζ as equal to twice the
rotational vector ω, then:
ζ =2ω = ∇×v
Referring back to simple ﬂow conﬁguration of ﬁgure 2.13 we note that
v
y
= v
z
= 0 and v
x
= v
0
y , then we have:
ω
z
=
1
2
(0 −v
0
) = −
1
2
v
0
Thus, the ﬂow is rotational. At this time we deﬁne irrotational ﬂows
as those where ω = 0 at each point in the ﬂow. Rotational ﬂows are those
where ω 6= 0 at points in the ﬂow.
The normal strain rates give the rate of stretching or shrinking of the two
lines shown in ﬁgure 2.14, while the shear strain rates give rate of change of
angularity of the two lines. What’s left of the relative motion must be rigid
body motion. Thus the expression
1
2
(
∂vy
∂x
−
∂vx
∂y
) is actually more than just
average rotation of the line segments about the z axis. It represent the rigid
body angular velocity ω
z
of the line segments about the z axis.
78 CHAPTER 2. RESERVOIR
A development of rotation in a ﬂuid particle in an initially irrotational
ﬂow would require shear stress to be present on the particle surface. Thus
the shear stress(Newton’sss viscosity law) in such ﬂows and in more general
ﬂows will depend on the viscosity of the ﬂuid and the rate of variation of
the velocity in region. For ﬂuids with small viscosity such as water and air,
assumption of irrotational ﬂow is valid for great part of the ﬂow except region
with large velocity gradient. Thus the boundary layer regions are the places
that element of ﬂuid rotates. Irrotational analysis may be carried out if the
boundary layer thickness is small in comparison with the scale of the ﬂuid.
2.8 RESERVOIR’S EQUATIONOF MOTION
Based on the information presented in the previous sections, the Euler equa
tion is valid for the case of dam’s reservoir. In the Euler equation, if we
divide the pressure into static and dynamic pressure we can write for static
pressure −∇p
s
+ B =0, therefore for only hydrodynamic pressure we can
write the following:
ρ
Dv
Dt
= −∇p
The p is the dynamic pressure in excess of static pressure. The above equation
can be written as:
ρ(∇.v) v + ρ
∂v
∂t
= −∇p
For small amplitude motion the eﬀect of convective term of the left hand side
of the above equation is negligible, thus we can write:
ρ
∂v
∂t
= −∇p (2.24)
For a linearly compressible ﬂuid, we deﬁne the following:
p = −K(²
xx
+²
yy
+²
zz
) (2.25)
Derivating of whole equation with respect to time:
.
p
= −K(
.
²
xx
+
.
²
yy
+
.
²
zz
)
The
.
²
xx
can be written as following:
2.9. RESERVOIR BOUNDARY CONDITIONS 79
.
²
xx
=
∂
∂t
(²
xx
) =
∂
∂t
µ
∂u
x
∂x
¶
=
∂
∂x
µ
∂u
x
∂t
¶
=
∂v
x
∂x
In which u
x
is the displacement in the xdirection. Thus we can write:
.
p
= −K(∇.v) (2.26)
If we use equation 2.24 to take the divergence of the whole equation we get:
∇.(ρ
∂v
∂t
) = ∇.( −∇p)
Or we can write the following:
ρ
∂
∂t
(∇.v) = −∇
2
p (2.27)
Using equation 2.26 and substitute it in equation 2.27 we get:
ρ
K
∂
2
p
∂t
2
= ∇
2
p (2.28)
We deﬁne c =
p
ρ
K
, in which is the velocity of pressure wave in ﬂuid. For
incompressible ﬂuid (K = ∞), therefore, C = ∞and then we have:
∇
2
p = 0
The above equation is the Laplace equation of motion for incompressible
ﬂuid.
The equation of motion for reservoir (equation 2.28) obtained based on
the following assumption:
a) inviscid ﬂuid, µ = 0
b)small amplitude motion, convective terms are neglected
c)linear compressible ﬂuid, equation 2.25
2.9 RESERVOIRBOUNDARYCONDITIONS
The formulation of the boundary conditions associated with the reservoir
boundaries is simply the expression in mathematical terms of the physi
cal situations. There are an inﬁnite number of solutions to the diﬀerential
80 CHAPTER 2. RESERVOIR
equation. The task is to ﬁnd the solution that is relevant to the boundary
condition. It should be noted that in addition to the spatial (or geometric)
boundary conditions, there are temporal boundary conditions which specify
the state of the variable of the interest at some points at time. This temporal
condition is called an ”initial condition”.
In case of damreservoir we have four spatial (geometric) boundary condi
tions namely, damreservoir, reservoirfoundation, free surface, and reservoir
farend boundary conditions (ﬁgure 2.15).
Figure 2.15: Boundaries of the damreservoir system
2.9.1 DamReservoir Boundary Condition
At the surface of ﬂuidstructure, it is clear that there must be no ﬂow across
the interface. This is based on the fact that face of the concrete dams are im
permeable. This results into the condition that at the normal to the boundary
there is no relative velocity or another word we can write it mathematically:
v
s
n
= v
n
in which n is the unit normal vector to the boundary at the damreservoir
interface and v
s
n
and v
n
are the velocity of the structure (dam) and ﬂuid
along the n respectively (ﬁgure 2.16). The above equation can be rewritten
as following:
v
s
n
= v.n
2.9. RESERVOIR BOUNDARY CONDITIONS 81
Diﬀerentiating the above equation with respect to time will give:
∂v
s
n
∂t
=
∂v
∂t
.n
It must be noted that n is normal to the surface and it direction is constant
as well as its value at a point. From the Euler equation and knowing that
∂v
s
n
∂t
is normal acceleration of the dam at the interface we can write:
a
s
n
= −
1
ρ
∇p.n
or:
∂p
∂n
= −ρa
s
n
Figure 2.16: Damreservoir interface
2.9.2 ReservoirFoundation Boundary Condition
If there is no absorption or penetration of water into the reservoir bottom, the
same damreservoir boundary condition that was obtained previously, can be
82 CHAPTER 2. RESERVOIR
used for reservoirfoundation boundary. In case of reservoir with sediment at
the bottom, we can account it in a very simpliﬁed manner.
The boundary condition at the reservoir bottom relates the hydrodynamic
pressure to the sum of the normal acceleration and acceleration due to inter
action between impounded water and the reservoir bottom material. Here,
we consider only interaction in the normal direction due to the assumption
that the hydrodynamic pressure waves incident on the reservoir bottom only
excite vertically propagating dilatational waves in the reservoir bottom ma
terials. The hydrodynamic pressure p(n, t) in the water is governed by the
onedimensional wave equation:
∂
2
p
∂n
2
=
1
C
2
∂
2
p
∂t
2
n ≥ 0 (2.29)
Similarly, the interaction displacement κ(n, t) in the layer of reservoir
bottom materials is governed by:
∂
2
κ
∂n
2
=
1
C
2
r
∂
2
κ
∂t
2
n ≤ 0 (2.30)
Where C
r
=
q
Er
ρ
r
, E
r
is the Young’s modulus of elasticity and ρ
r
is the density
of the reservoir bottom materials. At the reservoir bottom, the accelerative
boundary condition states that the normal pressure gradient is proportional
to the total acceleration:
∂p(0, t)
∂n
= −ρ
£
a
n
(t)+
..
κ (0, t)
¤
(2.31)
Where
..
κ (0, t) is the acceleration of the reservoir bottom due to interaction
between the impounded water and the reservoir bottom materials. Equilib
rium at the surface of the reservoir bottom materials requires that:
p(0, t) = −E
r
∂κ
∂y
(0, t) (2.32)
the D’alembert solution to equation 2.30 is κ = g
r
(n + C
r
t) where g
r
is
the wavefront of the refracted wave propagating vertically downward in the
reservoir bottom materials. An upward propagating wave does not exist
because of the radiation condition for the assumed inﬁnitely thick layer of
reservoir bottom materials. Note that
∂κ
∂t
(0, t) = C
r
g
0
r
(C
r
t) and
..
κ (0, t) =
2.9. RESERVOIR BOUNDARY CONDITIONS 83
C
2
r
g
00
r
(C
r
t), where the prime indicates the derivative of g
r
with respect to
argument (n +C
r
t). Diﬀerentiating equation 2.32 with respect to t gives:
∂p(0, t)
∂t
= E
r
C
r
g
00
r
(C
r
t)
or:
∂p(0, t)
∂t
= −
E
r
C
r
..
κ (0, t) (2.33)
The solution for
..
κ (0, t) from equation 2.33, when substituted into equation
2.31 with the identity
C
r
E
r
=
1
ρ
r
C
r
, gives:
∂p(0, t)
∂n
−q
∂p(0, t)
∂t
= −ρa
n
(t) (2.34)
Where q =
ρ
ρ
r
C
r
. For rigid reservoir bottom materials, C
r
= ∞and q = 0, so
the second term on the right hand side of the above equation is zero giving the
boundary condition for a rigid reservoir bottom. The fundamental parameter
that characterizes the eﬀects of absorption of hydrodynamic pressure waves
at the reservoir bottom is the admittance or damping coeﬃcient q. The wave
reﬂection coeﬃcient κ, which is the ratio of the amplitude of the reﬂected
hydrodynamic pressure wave to the amplitude of a vertically propagating
pressure wave incident on the reservoir bottom can be related to the damping
coeﬃcient.
The wave reﬂection coeﬃcient κ can be obtained by considering the re
ﬂection of a harmonic pressure wave in the impounded water impinging verti
cally on the reservoir bottom. The downward vertically propagating incident
wave of unit amplitude is p
i
= exp
£
i
ω
C
(n +Ct)
¤
, while the resulting upward
propagating wave is p
r
= κexp
£
i
ω
C
(−n +Ct)
¤
. Both p
i
and p
r
satisfy the
wave equation (equation 2.29) The substitution of the sum p
i
+p
r
of the two
hydrodynamic pressure waves into the boundary condition at the reservoir
bottom, equation 2.34, with no normal acceleration results in an equation
which can be solved to results in:
κ =
1 −qC
1 +qC
The wave reﬂection coeﬃcient is independent of excitation frequency ω. In
general, it depends on the angle of incident of the pressure wave at the reser
voir bottom. The wave reﬂection coeﬃcient may range within the limiting
84 CHAPTER 2. RESERVOIR
values of 1 and −1. For rigid reservoir bottom C
r
= ∞ and q = 0, resulting
in κ = 1. For very soft material, C
r
approaches zero and q = ∞, resulting in
κ = −1. The material properties of the reservoir bottom medium are highly
variable and depend upon many factors. It is believed the κ values from 1
to 0 would cover the wide range of materials encountered at the bottom of
actual reservoirs.
2.9.3 Free Surface Boundary Condition
The geometry of free surface boundary is not known a priori. This shape is
part of the solution, which means we have a very diﬃcult boundary condition
to cope with. In the case of surface wave of negligible surface tension, we
call them gravity wave. The free surface of a wave can be described as:
F(x, y, z, t) = z −ϕ(x, y, z, t) = 0 (2.35)
where ϕ is the displacement of the free surface above the horizontal plane,
Figure 2.17: Free surface wave
say z = 0,. If the surface varies with time, as would the water surface, then
the total derivative of the surface with respect to time would be zero on the
surface. In other word, if we move with the surface, it does not changes.
DF(x, y, z, t)
Dt
= 0 =
∂ϕ
∂t
+v
x
∂F
∂x
+v
y
∂F
∂y
+v
z
∂F
∂z
on F(x, y, z, t) = 0
Or:
2.9. RESERVOIR BOUNDARY CONDITIONS 85
−
∂F
∂t
= v.∇F = v.n∇F
Where the unit vector normal to the surface has been introduced as n =
∇F
∇F
.
Rearranging the above equation, we get:
v.n =
−
∂F
∂t
∇F
on F(x, y, z, t) = 0
The boundary condition at the free surface using equation 2.35 is:
v.n =
∂F
∂t
q
¡
∂ϕ
∂x
¢
2
+
¡
∂ϕ
∂z
¢
2
+ 1
on z = ϕ(x, t)
Where:
n =
−
∂ϕ
∂x
i−
∂ϕ
∂y
j +k
q
¡
∂ϕ
∂x
¢
2
+
¡
∂ϕ
∂z
¢
2
+ 1
Carrying out the dot product:
v
z
=
∂ϕ
∂t
+v
x
∂ϕ
∂x
+v
z
∂ϕ
∂z
on z = ϕ(x, t)
Neglecting the convection terms and approximating pressure we get:
p = ρgϕ at z = 0 (2.36)
v
z
=
∂ϕ
∂t
at z = 0 (2.37)
We also know:
∂v
z
∂t
= −
1
ρ
∂p
∂z
+g (2.38)
Diﬀerentiating equation 2.36 with respect to t and cancelling ϕ with equation
2.37 gives:
1
ρg
∂p
∂t
= v
z
86 CHAPTER 2. RESERVOIR
A second diﬀerentiation with respect to t and elimination of v
z
with equation
2.38 then substitution of p
total
= p + ρgϕ, in which p represent the excess
pressure due to motion and ρgϕstands for hydrostatic pressure, we get:
1
g
∂
2
p
∂t
2
+
∂p
∂z
= 0 at z = 0
This is an approximate to the surface boundary condition.
The above boundary condition is usually replaced with the boundary
condition that:
p = 0 at z = 0
This assumption of no surface wave is common assumption in concrete dams
and is valid. It was shown that the surface waves are negligible.
The more complicated free surface boundary condition can be established
in which we can assume pressure distribution due to interaction of wind and
surface. These cases are above the scope of the book and can be found in
appropriate references of water wave mechanics.
2.10 SOLUTIONOF THE RESERVOIREQUA
TION
Westergard’s classic work (1933) on the hydrodynamic water pressure on
dams during the earthquake started a new area for the researcher in this ﬁeld.
Westergaard’s solution to wave equation for rigid dams during earthquakes
was obtained based on the assumptions of dam with rectangular reservoir
subjected to horizontal earthquake. In the solution, the reservoir extended
to inﬁnity in the upstream direction and the eﬀect of surface wave was ne
glected. The system was subjected to the horizontal ground acceleration
which was perpendicular to the dam axis. The motion was assumed to be
small amplitude motion and water was taken inviscid ﬂuid.
The pressure equation for the system shown in ﬁgure 2.18 is found under
the following boundary conditions:
p = 0 at y = h
u
y
= 0 at y = 0
..
u
x
= ag cos(
2πt
T
) at x = 0
p = 0when x →∞
2.10. SOLUTION OF THE RESERVOIR EQUATION 87
Figure 2.18: Rigid daminﬁnite reservoir system
The hydrodynamic pressure for the above mentioned boundary conditions
is as follows:
p(x, y, t) =
8aρgh
π
2
cos(
2πt
T
)
∞
X
n=1,3,5,...
µ
1
n
2
c
n
e
−q
n
sin(
nπ(h −y)
2h
)
¶
(2.39)
in equation 2.39 q
n
and c
n
are deﬁned as followings:
q
n
=
nπc
n
x
2h
c
n
=
r
1 −
16ρh
2
n
2
KT
2
=
r
1 −
16h
2
n
2
C
2
T
2
The maximum pressure at x = 0 occurs when t = 0, T, 2T, ... and can be
written as:
p
max
(x, y, t) =
8aρgh
π
2
∞
X
n=1,3,5,...
µ
1
n
2
c
n
sin(
nπ(h −y)
2h
)
¶
(2.40)
The maximum pressure at a given time on the upstream face (x = 0) happens
at the bottom of the reservoir. The shape of the pressure diagram is such
that the curve has a horizontal tangent at the top and a vertical tangent at
the bottom.
88 CHAPTER 2. RESERVOIR
In an approximation, the maximum pressure (equation 2.40 ) can be
replaced by a parabolic curve. The maximum hydrodynamic pressure to be
added to the hydrostatic pressure is as follows:
p
max
(0, y, t) =
0.0255 ton −ft
3
a
p
h(h −y)
q
1 −0.72(
h sec
1000T ft
)
To further simplify the hydrodynamic pressure on the dam, one may
consider the hydrodynamic pressure as a certain body of water moves with
dam. This is called as added mass approach. This is possible because of the
fact that the hydrodynamic pressure obtained (equation 2.39) has same phase
compared to ground acceleration. Thus, we need to ﬁnd the mass of water
that attached to the dam and moves with it. For water with unit weight of
0.03125ton−ft
3
the width of water along the xdirection is as follows(ﬁgure
2.19):
b =
7
8
p
h(h −y)
The mass of water that moves with the dam body can be transferred in to
extra mass of concrete attached to the dam body. For a dam concrete with
the unit mass of 144lb.percu.ft the width of concrete would be:
b
0
= 0.38
p
h(h −y)
As it was already mentioned the Westergaard approach gives a pressure
distribution which is in phase with ground acceleration. Also we imposed the
boundary condition at farend of the reservoir that as x → ∞ then p → 0.
This boundary condition is an indication of energy dissipation at the farend.
The Westergaard solution is valid if the value of c
n
is real. Another word, we
can write:
1 −
16h
2
n
2
C
2
T
2
≥ 0
This will result in:
T >
4h
nc
n = 2m−1, m = 1, 2, 3, ...
The term,
4h
(2m−1)c
is the m
th
period of the reservoir, T
m
,. Therefore, the
westergaard’s solution is valid if the period of loading, T, is greater than the
2.10. SOLUTION OF THE RESERVOIR EQUATION 89
Figure 2.19: Added mass approach
90 CHAPTER 2. RESERVOIR
ﬁrst period of the reservoir, T
1
=
4h
c
,. If T = T
m
, then p = ∞ and we have
resonance.
The Westergaard pressure equation at the upstream face of the dam can
be written as followings:
p(0, y, t) =
4aρg
π
cos ωt
∞
X
m=1
(−1)
m−1
(2m−1)
q
λ
2
m
−
ω
2
C
2
cos λ
m
y
¸
¸
in which ω
m
and T
n
are the circular frequency and period of the m
th
mode
of the reservoir and ω is the circular frequency of the earthquake.. They can
be written as follows:
λ
m
=
ω
m
C
ω =
2π
T
ω
m
=
2π
T
m
=
(2m−1)πC
2h
m = 1, 2, 3, ...
Another solution to the wave equation presented by Chopra. Chopra
(1967) presented a closed form solution for hydrodynamic pressure in the
case of rigid dam with the vertical upstream face under the horizontal and
vertical ground motion. For the excitation frequency less than the fundamen
tal frequency of the reservoir, both Westergaard and Chopra’s solution are
the same. The greater excitation frequency causes an outofphase pressure
compared to Westergard’s solution. In this case the hydrodynamic pressures
cannot be represented by inertia eﬀect on an added mass moving with the
dam. Also it was found that the response to vertical ground motion is a real
valued function and there is no decay with time so that the system is truly
undamped in this case.
In his solution for the horizontal earthquake, the wave equation for a
rigid dam with rectangular reservoir (ﬁgure 2.18)presented subjected to the
following boundary conditions:
p = 0 at y = h
∂p
∂y
= 0 at y = 0
∂p
∂x
= −ρ
..
u
g
at x = 0
The hydrodynamic pressure for a harmonic ground excitation of
..
u
g
= e
iwt
would be:
2.10. SOLUTION OF THE RESERVOIR EQUATION 91
p(x, y, t) =
4ρ
π
∞
X
m=1
(−1)
m−1
(2m−1)
q
λ
2
m
−
ω
2
C
2
exp
Ã
−x
r
λ
2
m
−
ω
2
C
2
!
cos λ
m
y e
iwt
(2.41)
It can be seen that as x → ∞ the pressure does not approach zero if
λ
m
<
ω
C
which was the fourth boundary condition imposed in Westergaard
solution. To compare the above solution with Westergaard’s solution, we
can consider the real part of the ground motion as well as coeﬃcient ag
(
..
u
g
= ag cos(
2πt
T
)).The Solution in this case consist of the real part of the
pressure solution (equation 2.41) and would be as:
p(0, y, t) =
4aρg
π
sin ωt
m
1
−1
X
m=1
(−1)
m−1
(2m−1)
q
ω
2
C
2
−λ
2
m
cos λ
m
y +
4aρg
π
cos ωt
∞
X
m=m
1
(−1)
m−1
(2m−1)
q
λ
2
m
−
ω
2
C
2
cos λ
m
y
in which m
1
is the minimum value of m such that λ
m
>
ω
C
. If m
1
= 1 which
is an indication of λ
1
>
ω
C
or ω < ω
1
, then the term involving sin ωt vanishes
and the solution of the wave equation is same as Westergaard’s solution.
Here, again it can be noticed that the Westergaard’s solution is valid if the
frequency of the excitation is less than the fundamental frequency of the
reservoir system. For ω > ω
1
the term involving sin ωt does not vanish and
represent existence of pressure at inﬁnite distance up the reservoir. The
pressure response has two terms one is in phase with ground motion and the
other is in opposite phase of ground motion. Therefore, the hydrodynamic
pressure cannot be represented by added mass approach.
For m
1
> 1 we have pressure at inﬁnite distance. This means that for the
case of ω
m
< ω < ω
m+1
the pressure would be zero at x = ∞ for ω
m+1
and
higher modes. While for ω
m
and lower modes we expect to have pressure at
very farend of the reservoir.
92 CHAPTER 2. RESERVOIR
The solution of wave equation subjected to the vertical ground motion
for a rigid dam with rectangular reservoir is investigated. The reservoir have
the following boundary conditions:
∂p
∂y
= −ρ
..
u
y
at y = 0
1
g
∂
2
p
∂t
2
+
∂p
∂y
= 0 at y = h
∂p
∂x
= 0 at x = 0
The pressure solution when the reservoir is subjected to the vertical
ground motion of
..
u
y
= ag cos(ωt) would be as follows:
p(x, y, t) =
agρC
ω
sin
ω
C
(h −y) −
g
ωC
cos
ω
C
(h −y)
cos
ω
C
h +
g
ωC
sin
ω
C
h
cos(ωt)
The solution is independent of the x−coordinate.
If we ignore the free surface wave and change the second boundary con
dition by:
p = 0 at y = h
then the solution would be:
p(x, y, t) =
agρC
ω
sin
ω
C
(h −y)
cos
ω
C
h
cos(ωt) (2.42)
The diﬀerence between the two solutions depends on the magnitude of the
quantity
g
ωC
. Analysis of strong ground motions of the past earthquakes
seems to indicate that frequencies of most of the signiﬁcant harmonic com
ponents lie in the range 1 < ω < 120 rad/sec. Thus, the largest value of
g
ωC
may be taken as 6.82 × 10
−3
(ω = 1, C = 4720fps). However, such low
frequency harmonics contribute little to the response in the case of typical
reservoir depths encountered, because natural frequencies for the reservoir
are rather high: e.g. for a 300 ft deep reservoir ω
1
= 24.72 radians per sec.
The magnitude of
g
ωC
for harmonics of signiﬁcance is therefore considerably
less than 6.82×10
−3
, and errors introduced by dropping terms involving
g
ωC
will be small, except possibly near the free surface. Therefore, equation2.42
is suﬃciently accurate.
Most investigator have ignored the waves that may be generated at the
free surface of water. For compressible ﬂuid, and harmonic excitation it
was found that the error ,e , introduced by ignoring surface wave, varies as
follows:
e < 0.05 if
h
T
> 4.2
√
h
2.11. RESERVOIRFARENDTRUNCATEDBOUNDARYCONDITION93
0.05 < e < 0.20 if 2.6
√
h <
h
T
< 4.2
√
h
e < 0.20 if
h
T
< 2.6
√
h
Most of signiﬁcant harmonic in typical strong ground motion have a pe
riod below 3 sec. Thus for T = 3 sec, the error e < 0.05 if h > 158.5m.
For a reservoir depth 100 ft, e < 0.05 if T < 1.32 sec . Because the natural
periods of the system are very small (ﬁrst natural period for 100 ft depth
of water is 0.085 sec), the contributions to the total response from harmonic
components of ground motion with the periods longer than 1.32 sec will be
small. This is particularly so because we are dealing with an ﬂuid. Hence,
it may be concluded that errors introduced by neglecting surface wave is on
the order of 0.05. Similar conclusions can be derived for other depth between
100 ft and 520 ft. On this basis, it seems that the eﬀect of surface waves can
be ignored with little loss of accuracy.
2.11 RESERVOIRFARENDTRUNCATED
BOUNDARY CONDITION
The truncated boundary for the ﬁnite element and boundary element model
ing of the inﬁnite reservoir was worked by so many researchers. Sommerfeld
boundary condition is the most common one that is based on the assumption
that at long distance from the dam face, water wave can be considered as
plane wave. A plane wave can be represented by an equation of the form:
p = F(x −Ct)
This equation leads to the following condition:
∂p
∂x
=
∂p
∂n
= −
1
C
∂p
∂t
(2.43)
in which n is the normal at the truncation boundary. This represents well
known Sommerfeld radiation condition. It introduces a damping in the sys
temand models the loss of energy in the outgoing waves. Under this condition
94 CHAPTER 2. RESERVOIR
for the far end reservoir, the classical solution for the hydrodynamic pressure
was obtained in which there are inphase and outofphase pressures. As a
rule for the truncated boundary there is no reﬂection for the outgoing wave
and all of our eﬀort is in modeling the energy loss in outgoing wave such
that all energy can be absorbed on the truncated boundaries. The solution
to the wave equation can be investigated for a rigid dam with rectangular
reservoir subjected to horizontal ground motion and the following boundary
conditions:
p = 0 at y = h
∂p
∂y
= 0 at y = 0
∂p
∂x
= −ρ
..
u at x = 0
∂p
∂x
=
∂p
∂n
= −
1
C
∂p
∂t
at x = L
The hydrodynamic pressure for a harmonic ground excitation of
..
u
g
= e
iwt
would be:
p(x, y, t) =
4ρ
π
∞
X
m=1
(−1)
m−1
(2m−1)q
m
e
q
m
(L−x)
+Z
m
e
−q
m
(L−x)
e
q
m
L
−Z
m
e
−q
m
L
cos λ
m
y e
iwt
(2.44)
Where:
q
m
=
r
λ
2
m
−
ω
2
C
2
, Z
m
=
q
m
−i
ω
C
q
m
+i
ω
C
, i =
√
−1 (2.45)
It was found that the Sommerfeld radiation damping condition does not
yield good approximation for the inﬁnite reservoir for excitation frequencies
between the ﬁrst and second natural frequencies of the reservoir. Within this
range of frequencies an increase in the length of reservoir does not lead to
any increase in the accuracy(Humar and Roufaiel, 1983).
Humar and Roufaiel (1983) used a radiation condition which adequately
models the loss over a wide range of excitation frequencies, also it was shown
that the conditions give much better results as compared to the plane wave
Sommerfeld condition. Their radiation condition derived only for the hori
zontal ground motion in the case of dam with vertical upstream face. The
Solution of the equation 2.41 when
..
u
g
= e
iωt
can be represented as following:
p(x, y, t) =
m
1
−1
X
m=m
1
A
m
q
n
e
−q
n
x
cos λ
m
y +
∞
X
m=m
1
A
m
q
λ
2
m
−
ω
2
C
2
cos λ
m
y
¸
¸
e
iωt
2.11. RESERVOIRFARENDTRUNCATEDBOUNDARYCONDITION95
It was already shown that at large distance from the dam, the second term
in the equation vanishes, and the pressure is given as:
p(x, y, t) =
m
1
−1
X
m=1
A
m
e
−
µ
q
λ
2
m
−
ω
2
C
2
¶
x
cos λ
m
y e
iωt
=
m
1
−1
X
m=1
p
m
(2.46)
From the above equation we can write:
p
∂p
∂x
= −
1
ω
m
1
−1
X
m=1
Ã
r
ω
2
C
2
−λ
2
m
!
p
m
∂p
∂t
(2.47)
For the case of m = 2 (when ω lies between ω
1
and ω
2
) equation 2.46 gives
p = p
1
and equation 2.47 becomes:
∂p
∂x
= −
1
C
Ã
r
1 −
³
ω
1
ω
´
2
!
∂p
∂t
Therefore, it can be said that the following condition should be applied at
the truncated boundary:
∂p
∂n
= 0 ω < ω
1
(2.48)
∂p
∂n
= −
1
C
Ã
r
1 −
³
ω
1
ω
´
2
!
∂p
∂t
ω > ω
1
(2.49)
Apparently, the preceding condition is not exact when ω is greater than ω
2
.
For large value of ω equation 2.49 reduces to equation 2.43.
The hydrodynamic pressure obtained by applying the modiﬁed boundary
condition at x = L represented by equations 2.48 and 2.49 will be given by
an expression similar to equation 2.44, except that Z
m
will now be as follows:
Z
m
= 1 ω < ω
1
Z
m
=
q
m
−i
ω
C
q
1 −
¡
ω
1
ω
¢
2
q
m
+i
ω
C
q
1 −
¡
ω
1
ω
¢
2
ω > ω
1
96 CHAPTER 2. RESERVOIR
The modiﬁed boundary condition at the far end was obtained with ig
noring the second term in the pressure equation at the large distance from
the dam. It was found that the modiﬁed radiation boundary condition gives
much better results as compared to the Sommerfeld one for excitation fre
quencies between 0.0 and 2.
Sharan (1984) showed that the condition of
∂p
∂n
= 0 for incompressible
ﬂuid at the far end is the form of that for a rigid stationary boundary and
the behavior of the ﬂuid motion at the truncated boundary is not truly
presented. A large extent of the ﬂuid domain is required to be included in
the analysis. The above condition is same as Sommerfeld condition for the
incompressible ﬂuid (C = ∞). Under the assumption of incompressible ﬂuid
in the rectangular reservoir and rigid dam, the governing equation of the
pressure would be the Laplace equation. Knowing the pressure equation, he
found that at large distance from upstream face condition would be:
∂p
∂n
= −
π
2h
p
It can be observed that at very large distance away from the dam face p = 0.0
therefore
∂p
∂n
= 0 which is same condition at the inﬁnity for the inﬁnite
reservoir. He found that under the assumption of the rigid dam and rigid
rectangular reservoir bottom, for the horizontal vibration. the truncated
boundary can be located very close to the structure. It was found that
although the proposed boundary condition was derived for the dam with the
vertical upstream face, the results for an inclined face are relatively more
accurate than those for a vertical face.
Sharan (1985) used the radiation condition for the submerged structure
surrounded by unbounded extent of compressible ﬂuid as following:
∂p
∂n
= −Ψp
Deferent geometry of the solidﬂuid interface was used to ﬁnd the accuracy
of the proposed radiation boundary condition.
Sharan (1987) proposed a damper radiation boundary condition for the
time domain analysis of the compressible ﬂuid with small amplitude. The
structure submerged in unbounded ﬂuid in the upstream direction. The
proposed radiation condition was:
∂p
∂n
= −
π
2h
p −
1
C
∂p
∂t
2.11. RESERVOIRFARENDTRUNCATEDBOUNDARYCONDITION97
The above boundary condition was found to be very eﬀective and eﬃcient
for a wide range of the excitation frequency. For the ﬁnite value of c and h,
the eﬀectiveness of the proposed damper depends on the period of excitation
T, (
2π
ω
). If the value of T be near the natural period of vibration of the ﬂuid
domain (
TC
h
→ 4), the magnitude of pand
∂p
∂t
become inﬁnitely large. In
this case approximated boundary condition for the truncated surface may be
written as
∂p
∂n
= 0 . Therefore for such a case, neither Sommerfeld damper nor
the proposed one would be eﬀective. In the case of
TC
h
¿ 4, the magnitude
of ,
pC
h
, becomes small compared to that of
∂p
∂t
. For the limiting case p →0
when T → 0. On the other hand, for
TC
h
À 4,
∂p
∂t
is much less than that
of
pC
h
. For the limiting case
∂p
∂t
→ 0 when T → ∞. Thus in both case, one
of the terms on the right hand side of proposed equation of damping would
be small compare to the other. As a general, it was found that proposed
boundary condition at the truncated surface is more eﬃcient and needs less
computational time. In the case of
TC
h
= 1 (high excitation frequency) both
methods shows discrepancy with the case of L = ∞. It is not an important
problem because in high frequency excitation the hydrodynamic forces are
not comparable so that its error in hydrodynamic pressure is not signiﬁcant.
Sharan (1985) expressed the condition at the truncated face as:
∂p
∂x
= −
α
C
∂p
∂t
+
βω
C
p
For (Ω =
ωh
C
) less than
3π
2
which is for excitation frequency less than the
second natural frequency of the reservoir if
x
h
is suﬃciently large, then α = 0
and β = −
p
(
π
2Ω
)
2
−1 and for
π
2
< Ω <
3π
2
, β = 0 and α =
p
1 −(
π
2Ω
)
2
.
For Ω <
π
2
,
∂p
∂n
is function of p instead of
∂p
∂t
, therefore Sommerfeld radiation
boundary condition is not justiﬁed for Ω <
π
2
. For these range of frequencies
values of p,
∂p
∂t
,and
∂p
∂x
approach zeroes as the ratio
x
h
is increases. This
explains why satisfactory results may be obtained even with use of improper
boundary condition such as
∂p
∂n
= 0 and
∂p
∂n
= −
1
C
∂p
∂t
by considering a very
large extent of the reservoir. For Ω >
3π
2
even for large value of
x
h
, the
parameters α and β are sensitive to x and y coordinate point on the truncated
surface.
98 CHAPTER 2. RESERVOIR
Chapter 3
FINITE ELEMENT
MODELLING OF THE
DAMRESERVOIR SYSTEM
The ﬁrst step in dynamic analysis of a system is to properly model the actual
system. In mathematical modelling, we need to formulate the equation of the
motion. Having formulated the equation of the motion, we then procced to
solve the equation of the motion. In previous Chapter, reservoir’s equation of
the motion was formulated. Here we will ﬁnd the structure’s equation of the
motion and its ﬁnite element equation in addition to reservoir ﬁnite element
equation.
3.1 FINITE ELEMENTMODELLINGOF THE
STRUCTURE
3.1.1 SingleDegreeOfFreedom Systems
A singledegreeoffreedom system (SDF) is deﬁned as a system whose motion
can deﬁned by a single parameter. Systems shown in ﬁgure 3.1 are systems
of SDF. As shown in the ﬁgure only one parameter (u, x or θ ) is required
to deﬁne the position of the particle at each second.
For a system shown in 3.1−a, if we try to apply a dynamic force f(t), the
equation of the motion can be written as:
99
100CHAPTER3. FINITEELEMENTMODELLINGOF THEDAMRESERVOIRSYSTEM
Figure 3.1: Systems of single degree of freedom
f(t) = F
I
+F
D
+F
S
in which F
I
, F
D
and F
S
are inertial, damping and spring (stiﬀness) forces,
respectively. The inertial force is equivalent to the total acceleration of the
mass M and can be written as M
..
u when there is no base motion. If there is
a base motion, the inertial force can be written as M(
..
u +
..
u
g
). As is shown
in ﬁgure3.2
..
u is the acceleration of the mass M relative to the base and
..
u
g
is
the ground acceleration.The spring force can be written as Ku, which is a
linear spring or linear stiﬀness. Other form of spring may be proposed which
result in nonlinear equation. The damping force is viscous damping force
and is given by F
D
= C
.
u which C is the damping coeﬃcient. This is a
valid assumption when velocity is small. For high velocity, damping force is
proportional to the square of the velocity. Thus the equation of the motion
for a SDF system can be written as:
M(
..
u +
..
u
g
) +C
.
u +Ku = f(t)
This is a diﬀerential equation of second order. The equation can be
rewritten into the following form:
M
..
u +C
.
u +Ku = f(t) −M
..
u
g
3.1. FINITE ELEMENT MODELLING OF THE STRUCTURE 101
Figure 3.2: Forces on a single degree of freedom
It can be seen that the base acceleration is equivallent of a force acting on
the mass in the opposite direction of its motion. In the other word, the base
motion can be interpreted as a force acting on the mass in opposite direction
of its motion.
3.1.2 MultiDegreeOfFreedom System
In a multidegreeoffreedom (MDF) sytem, it is required to know the values
of the displacement in more than one point in order to deﬁne its motion at
any instant of time. Consider a system with n degrees of freedom which
accelerate in the y direction as shown in ﬁgure 3.3.
Figure 3.3: An example of multidegreeoffreedom (MDF) sytem with de
grees of freedom in y direction
102CHAPTER3. FINITEELEMENTMODELLINGOF THEDAMRESERVOIRSYSTEM
The inertial forces can be written as:
F
I
=
F
I1
F
I2
.
.
F
I(n−1)
F
In
=
M
1
......................
....M
2
..................
...........................
...........................
...............M
I(n−1)..
......................M
In
¸
¸
¸
¸
¸
¸
¸
¸
..
u
1
..
u
2
.
.
..
u
n−1
..
u
n
g
or it can be write as:
F
I
= [M]{
..
U
}
t
The mass matrix [M] is a diagonal matrix and in general it can be a non
diagonal symmetric matrix.
The forces due to elasticity can be written as:
F
S
=
F
S1
F
S2
.
.
.
F
Sn
=
K
11
K
12
.................K
1n
K
21
K
21
.................K
2n
..................................
..................................
.................................
K
n1
K
n2
................K
nn
¸
¸
¸
¸
¸
¸
¸
¸
u
1
u
2
.
.
.
u
n
or it can be written as:
F
S
= [K]{U}
where [K] is the stiﬀness matrix. Adetailed description of the stiﬀness matrix
can be found in ﬁnite element books.
The damping matrix can be written as :
F
D
=
F
D1
F
D2
.
.
.
F
Dn
=
C
11
C
12
.................C
1n
C
21
C
21
.................C
2n
..................................
..................................
.................................
C
n1
C
n2
................C
nn
¸
¸
¸
¸
¸
¸
¸
¸
.
u
1
.
u
2
.
.
.
.
u
n
or it can be written as:
3.1. FINITE ELEMENT MODELLING OF THE STRUCTURE 103
F
D
= [C]{
.
U
}
where [C] is the damping matrix.
Equating of the external forces and internal forces will results in:
[M]{
..
Ut
} + [C]{
.
U
} + [K]{U} = {f(t)}
which {f(t)} is the external forces act on the structures and can be written
as:
{f(t)} =
f
1
(t)
f
2
(t)
.
.
.
f
n
(t)
If there is a base displacement like ground acceleration along y direction
as shown in ﬁgure 3.3, the equation of motion for a multidegreeoffreedom
system can be written as:
[M]{
..
Ut
} + [C]{
.
U
} + [K]{U} = {f(t)}
in which
..
{U
t
} = {
..
U
gy
} +{
..
U
}. The vector of ground acceleration can be
written as:
..
{U
g
}=
..
u
gy
..
u
gy
.
.
..
u
gy
..
u
gy
=
..
u
gy
1
1
.
.
1
1
=
..
u
gy
{I}
Where
..
u
gy
is the ground acceleration at the base along y direction. Thus, the
equation of motion can be written as:
[M]{
..
U
} + [C]{
.
U
} + [K]{U} = {f(t)} −[M}{
..
Ugy
}
104CHAPTER3. FINITEELEMENTMODELLINGOF THEDAMRESERVOIRSYSTEM
Figure 3.4: An example of MDF system with two degrees of freedom at each
mass
For a system of MDF system when subjected to ground acceleration in two
directions and there are two degree of freedoms for each mass,as shown in
ﬁgure 3.4,we can write the equation of motion based on the both degrees of
freedom at each mass.For this system, we can write the displacement vector
as:
{U} =
u
1
v
1
u
2
v
2
.
.
.
u
n
v
n
The same can be written for {
.
U
} and {
..
U
}. The equation of motion can be
written as following:
[M]{
..
U
} + [C]{
.
U
} + [K]{U} = {f(t)} −[M]{
..
Ug
}
In which {
..
Ug
} can be written as:
3.1. FINITE ELEMENT MODELLING OF THE STRUCTURE 105
{
..
Ug
} =
..
u
gx
1
0
1
0
.
.
1
0
1
0
+
..
u
gy
0
1
0
1
.
.
0
1
0
1
When
..
u
gx
and
..
u
gy
are the ground acceleration along the x and y direction,
respectively. In the above equation of motion {f(t)} can be written as:
{f(t)} =
f
1x
(t)
f
1y
(t)
f
2x
(t)
f
2y
(t)
.
.
.
f
nx
(t)
f
ny
(t)
Which f
ix
(t) and f
iy
(t) are the forces acting on the ith mass along x and y
direction, respectively. In the case of concrete dams, {f(t)} can be seperated
into hydrodynamic pressure {f} and resultant of all the other forces, {f
1
},
that act on the structure. Thus, the ﬁnal form of the equation of the motion
can be written as:
[M]{
..
U
} + [C]{
.
U
} + [K]{U} = {f} +{f
1
} −[M]{
..
Ug
}
In a MDF sytem, a system is discretized into its degrees of freedom. The
number of degrees of freedom depend on the number of elements and the
degree of freedom for each element. Usually in the case of concrete gravity
dams isoparametric plane stress or plane strain are used. These types of
elements are used for two dimensional problems. In some cases triangular
106CHAPTER3. FINITEELEMENTMODELLINGOF THEDAMRESERVOIRSYSTEM
elements may be used. In both plane stress or plane strain the displacement
ﬁeld is given by the u and v in directions of cartisian x and y axis. In the case
of concrete gravity dam, assumption of 2D analysis is a valid assumption.
In 3D analysis the simplest case is the case of tetrahedron element. Based
on the nature of the problem each node may have 3 diﬀerent components u, v
and w in the direction of X, Y and Z. For more complex type of element we
may add rotational degrees of freedom in to each node. Three dimensional
analysis of concrete gravity dams requires a large number of DOF.
3.2 COUPLING MATRIX OF THE DAM
RESERVOIR
The coupling matrix relates the pressure of the reservoir and the forces on
the damreservoir interface as following:
[Q]{p} = {f}
where {f} is the force vector acting on the structure due to the hydrodynamic
pressure.
Figure ?? shows a line element on the interaction boundary of the dam
reservoir. The work done by the hydrodynamic pressure on the interaction
surface of the structure must be equal to the work of the equivalent nodal
forces on the interface boundary of an element. Thus, for unit thickness
elements as shown in ﬁgure ??, the following expression can be written:
Z
s
e
pU
n
ds = {f}
e
T
{δ} =
©
fx
1
fy
1
fx
2
fy
2
ª
u
1
v
1
u
2
v
2
(3.1)
where p and U
n
are the values of the hydrodynamic pressure and normal
displacement along the element interface, respectively. {δ} and {f}
e
are the
displacement and force vector of an interface element. u
i
and v
i
(fx
i
and fy
i
)
are the displacements (forces) at node i of the interface element along the
global X and Y coordinates, respectively. The integration is performed along
each element on the damreservoir interface. The superscript and subscript
’e’ refer to the element on the damreservoir interface. Writing u and v,
3.2. COUPLING MATRIX OF THE DAMRESERVOIR 107
displacements along the global X and Y coordinates of the interface element,
in terms of structure shape functions, then:
u = u
1
N
1
+u
2
N
2
v = v
1
N
1
+v
2
N
2
where N
i
is the structure shape function at nodei of the interface element.
For the normal displacement along the element surface, U
n
, we have:
U
n
= u
n
+v
n
= ηu
1
N
1
+ ηu
2
N
2
+ βv
1
N
1
+ βv
2
N
2
(3.2)
In equation 3.2, ηand βare the absolute values of the normal vector on the
boundary in the global directions of X and Y , respectively. Equation 3.2 can
be written in the following form:
U
n
=
©
ηN
1
βN
1
ηN
2
βN
2
ª
{δ} = {N
s
n
}
T
{δ} (3.3)
The hydrodynamic pressure can be expressed as shape function of the ﬂuid
in the form:
p = {N
f
}
T
{p}
e
=
©
N
f
1
N
f
2
ª
{p}
e
(3.4)
where N
f
i
is the ﬂuid shape function at node iof the interface element. Com
bining equations 3.1, 3.3 and 3.4, there is obtained:
{f}
e
=
Z
s
e
{N
s
n
}{N
f
}
T
ds {p}
e
= [Q]
e
{p}
e
(3.5)
where [Q]
e
and {p}
e
are the coupling matrix and hydrodynamic pressure
vector of an element on the damreservoir interface. The total coupling
matrix[Q] is obtained by assembling all element coupling matrices. From
equation 3.5, [Q]
e
is written as:
[Q]
e
=
Z
se
{N
s
n
}{N
f
}
T
ds
For an interface element as shown in ﬁgure 3.5, then:
[Q]
e
=
Z
se
ηN
1
N
f
1
ηN
1
N
f
2
βN
1
N
f
1
βN
1
N
f
2
ηN
2
N
f
1
ηN
2
N
f
2
βN
2
N
f
1
βN
2
N
f
2
¸
¸
¸
¸
108CHAPTER3. FINITEELEMENTMODELLINGOF THEDAMRESERVOIRSYSTEM
Figure 3.5: Interface element on the damreservoir interaction boundary
3.3 FINITEELEMENTMODELLINGOF THE
RESERVOIR
As it was discussed in previous Chapter, the hydrodynamic pressure distrib
ution in the reservoir is governed by the pressure wave equation. Assuming
that water is linearly compressible and neglecting its viscosity, the small am
plitude irrotational motion of water is governed by the twodimensional wave
equation:
∇
2
p(x, y, t) =
1
C
2
..
p
(x, y, t) (3.6)
where p(x, y, t) is the hydrodynamic pressure in excess of hydrostatic pres
sure, C is the velocity of pressure wave in water andxand y are the coordinate
axes.
The hydrodynamic pressure in the impounded water governed by equation
3.6, is due to the horizontal and the vertical accelerations of the upstream
face of the dam, the reservoir bottom as well as the far end of the reservoir in
the case of ﬁnite reservoir length. The motion of these boundaries is related
3.3. FINITE ELEMENT MODELLING OF THE RESERVOIR 109
to the hydrodynamic pressure by the boundary conditions.
For earthquake excitation, the condition at the boundaries of the dam
reservoir, reservoirfoundation and the reservoirfarend are governed by the
equation:
∂p(x, y, t)
∂n
= −ρa
n
(x, y, t)
whereρ is density of water and a
n
(x, y, t) is the component of acceleration on
the boundary along the direction of the inward normal n. No wave absorption
is considered at the boundaries of the reservoir.
Neglecting the free surface wave, the boundary condition at the free sur
face is written as:
p(x, y, t) = 0
whereh is the height of the reservoir.
Using ﬁnite element discretization of the ﬂuid domain and the discretized
formulation of equation 3.6, the wave equation can be written in the following
matrix form:
[G]{
..
p
} + [H]{p} = {F} (3.7)
where G
ij
=
P
G
e
ij
, H
ij
=
P
H
e
ij
and F =
P
F
e
i
. The coeﬃcient G
e
ij
,
H
e
ij
and F
e
i
for an individual element are determined using the following
expressions:
G
e
ij
=
1
C
2
Z
Ae
N
i
N
j
dA
H
e
ij
=
Z
A
e
(
∂N
i
∂x
∂N
j
∂x
+
∂N
i
∂y
∂N
j
∂y
)dA
F
e
i
=
Z
se
N
i
∂p
∂n
ds
where N
i
is the element shape function, A
e
is the element area and s
e
is the prescribed length along the boundary of the elements. In the above
formulation, matrices [H] and [G] are constant during the analysis while
the force vector {F} and the pressure vector {p} and its derivatives are the
variable quantities in equation 3.7.
110CHAPTER3. FINITEELEMENTMODELLINGOF THEDAMRESERVOIRSYSTEM
3.3.1 Truncated Boundary of the Reservoir’s FarEnd
The Sharan boundary condition, which was already described, at the farend
truncated boundary can be written as:
∂p
∂n
= −
π
2h
p −
1
C
.
p
Implementation of the truncated boundary condition in the ﬁnite element
model, can be done by separating the force vector {F} in equation 3.7 into
two components:
{F} = {FF
1
} +{FF
2
} (3.8)
where {FF
1
} is the component of the force due to acceleration at the
boundaries of the damreservoir and reservoirfoundation while {FF
2
} is due
to truncation at the far boundary and can be written as:
{FF
2
} = −
π
2h
[D]{p} −
1
C
[D]{
.
p
} (3.9)
where D
ij
= D
e
ij
and D
e
ij
is deﬁned as:
D
e
ij
=
Z
`
e
T
N
i
N
j
d`
T
(3.10)
In equation 3.10, `
e
T
is the side of the element on the truncated boundary.
Substituting equations 3.8 and 3.9 into equation 3.7 results in:
[G]{
..
p
} +
1
C
[D]{
.
p
} + ([H] +
π
2h
[D]){p} = {FF
1
} (3.11)
In general form, we can write the ﬁnite element form of the equation of the
reservoir as:
[G]{
..
p
} +[C
0
]{
.
p
} +[K
0
}{p} = {F} −ρ[Q]
T
({
..
U
} +{
..
Ug
}) = {F
2
} −ρ[Q]
T
{
..
U
}
Putting equation 3.11 in the format of the above equation, the following
relationships are obtained:
[C
0
] =
1
C
[D]
3.4. EQUATION OF THE COUPLED DAMRESERVOIR SYSTEM 111
[K
0
] = [H] +
π
2h
[D]
{F
2
} −ρ[Q]
T
{
..
U
} = {FF
1
}
Where {F
2
} is forces due to ground acceleration on the damreservoir bound
aries and total acceleration on the rest of the boundaries.
3.4 EQUATION OF THE COUPLED DAM
RESERVOIR SYSTEM
As we discussed, it can be seen that the damreservoir interaction is a clas
sic coupled problem which contains two diﬀerential equations of the second
order. The equations of the dam structure and the reservoir can be written
in the following form:
[M]{
..
U
} + [C]{
.
U
} + [K]{U} = {f
1
} −[M]{
..
Ug
} + [Q]{p} = {F
1
} + [Q]{p}
[G]{
..
p
} +[C
0
]{
.
p
} +[K
0
}{p} = {F} −ρ[Q]
T
({
..
U
} +{
..
Ug
}) = {F
2
} −ρ[Q]
T
{
..
U
}
where[M], [C] and [K] are mass, damping and stiﬀness matrices of the
structure and [G], [C
0
] and [K
0
]are matrices representing mass, damping and
stiﬀness of the reservoir, respectively. Detailed deﬁnitions of the[G],[C
0
] and
[K
0
] matrices and vector {F}, are presented in the previous sections.[Q] is the
coupling matrix; {f
1
} is the vector of body force and hydrostatic force; and
{p} and {U} are the vectors of hydrodynamic pressures and displacements.
{
..
U
g
} is the ground acceleration and ρ is the density of the ﬂuid. The dot
represents the time derivative.
112CHAPTER3. FINITEELEMENTMODELLINGOF THEDAMRESERVOIRSYSTEM
Chapter 4
DYNAMIC ANALYSIS OF
DAMRESERVOIR SYSTEM
4.1 INTRODUCTION
The damreservoir system can be categorized as a coupled ﬁeld system in
which two physical domains of ﬂuid and structure interact only at their in
terface. In such a problem, the presence of interaction implies that the time
response of both subsystems must be evaluated simultaneously. Diﬀerent ap
proaches to the solution of the coupled ﬁeld problem exist. Field elimination,
simultaneous solution and partitioned solution are the three classes of solu
tions for the coupled ﬁeld system. The advantages and disadvantages of each
method were addressed by Felippa and Park (1980). The ﬁeld elimination
approach is not feasible in the case of nonlinear problems. The reduced sys
tem of equations has high order derivatives which cause some diﬃculties in
applying the initial conditions. The simultaneous solution is time consuming
and involves many operations, especially when a large number of elements
is used. This method contains matrices with a large bandwidth and con
sequently requires a large amount of memory especially for the cases when
the existing matrices are not symmetric. In addition, the main disadvantage
of the ﬁrst two classes of solution arises from the diﬃculties encountered in
using available software while the partitioned solution has the capability of
using existing software for each subsystem. Staggered solution was described
by Felippa and Park (1980) as a partitioned solution procedure that can be
organized in terms of sequential execution of singleﬁeld analyser.
113
114CHAPTER4. DYNAMICANALYSIS OF DAMRESERVOIRSYSTEM
Most of the physical systems are made of subsystems which interact with
each other. These physical systems which are referred to as coupled sys
tems, have been investigated by several researchers. Methods of solution
vary depending on the governing diﬀerential equations of the subystems and
may lead to diﬀerent degrees of accuracy and stability of the solution (Park
1980). Coupled problems and their numerical solutions were addressed by
Felippa and Park (1980); Park and Felippa (1984). Zienkiewicz and Chan
(1989) proposed an unconditionally stable method for staggered solution of
soilpore ﬂuid interaction problem. Huang (1995) proposed two uncondition
ally stable methods for the analysis of soilpore ﬂuid problem. The methods
were named pressure correction method and displacement correction method.
Zienkiewicz and Chan (1989) presented an unconditionally stable method
for staggered solution procedure for the ﬂuidstructure interaction problem.
Their method was proved to be unconditionally stable when no damping term
was included in the equations of the ﬂuid and the structure. However, when
the damping term is included in the equation of the subsystems, the pro
posed method may not be unconditionally stable. The problem of solutions
instability when the damping term is included in the diﬀerential equation,
was recognized by researchers. Most of the staggered solution applications in
the ﬁeld of ﬂuidstructure interaction were conducted using a method which
is not unconditionally stable.
There are two method of solution for dynamic analysis of a system. Time
domain and frequency domain solutions are two methods of solutions and
have diﬀerent application depending on the nature of the system. For a
linear system both method can be used while for a nonlinear system only
time domain solution can be used to evaluate the dynamic response of the
system.
The behaviour of a nonlinear system is diﬀerent than the linear system.
Lack of superposition property, multiplicity of equilibria and domains of at
traction, local and global stability, frequency dependence of amplitude of
free oscilation, the jump response, subharmonic and superharmonic genera
tion are some of characteristics of a nonlinear system. In particular, when a
linear system is subjected to harmonic excitation, the steady state response
will be another harmonic with the same frequency and will be independent of
the initial conditions. A nonlinear system may behave in a diﬀerent manner.
This character of nonlinear systems make them unable to be solved using
frequency domain method.
In this Chapter, two methods of staggered solution procedure are applied
4.2. THE COUPLED DAMRESERVOIR PROBLEM 115
to the damreservoir interaction problem. Both methods are shown to be
unconditionally stable when the two diﬀerential equations of the ﬂuid and
structure include damping terms. The accuracy of the solution using both
of the proposed methods, is investigated. Two diﬀerent conﬁgurations of
concrete gravity dams are analysed to illustrate the application of the pro
posed procedure and to compare the solution with available ﬁnite element
solutions.
4.2 THE COUPLEDDAMRESERVOIRPROB
LEM
As it was discussed in previous Chapter, The damreservoir interaction is
a classic coupled problem which contains two diﬀerential equations of the
second order. The equations of the dam structure and the reservoir can be
written in the following form:
[M]{
..
U
}+[C]{
.
U
}+[K]{U} = {f
1
}−[M]{
..
Ug
}+[Q]{p} = {F
1
}+[Q]{p} (4.1)
[G]{
..
p
} +[C
0
]{
.
p
} +[K
0
}{p} = {F} −ρ[Q]
T
({
..
U
} +{
..
Ug
}) = {F
2
} −ρ[Q]
T
{
..
U
}
(4.2)
where[M], [C] and [K] are mass, damping and stiﬀness matrices of the struc
ture and [G], [C
0
] and [K
0
]are matrices representing mass, damping and stiﬀ
ness of the reservoir, respectively. Detailed deﬁnitions of the[G],[C
0
] and [K
0
]
matrices and vector {F}, are presented in the following sections.[Q] is the
coupling matrix; {f
1
} is the vector of body force and hydrostatic force; and
{p} and {U} are the vectors of hydrodynamic pressures and displacements.
{
..
U
g
} is the ground acceleration and ρ is the density of the ﬂuid. The dot
represents the time derivative.
4.3 DIRECTINTEGRATIONOF THE EQUA
TION OF MOTION
In direct integration the equation of motion are integrated using a numerical
stepbystep procedure. The term direct means that no transformation of the
116CHAPTER4. DYNAMICANALYSIS OF DAMRESERVOIRSYSTEM
equation of motion into the diﬀerent formis carried out.The direct integration
of the equations of motion provides the response of the system at discrete
intervals of time which are usually spaced. In this procedure three basic
parameters of displacement, velocitiy aand acceleration are computed. The
integration algorithms are based on appropriately selected expressions that
relate the response parameters at a given interval of time to their values at
one or more previous time points. In general two independent expressions
of this nature must be speciﬁed. The equation of motion written for the
time interval under consideration provides the third expression necessary
to determine the three unknown parameters. If the equation of motion is
written at a time step which the three parameters are unknown(t +∆t), the
time integration scheme is called implicite integration method. In explicit
integration method, the equilibrium equation is written at the time step
which all the three basis parameters are known (t).
In time integration scheme, the basic parameters are known at the be
gining of the integration or any points preceding the integration time. These
speciﬁed or computed values permit the marching scheme to be begun so
that response can be computed at as many subsequent points as desired.
The accuracy and stability of the scheme depend on the magnitude of the
time interval and marching algorithm.
There are several method of direct integration. Newmark method, Wil
son, Houbolt are implicit integration methods while Central Diﬀerence Method
is the explicit integration method.
The stability of each method depends on the time steps taken and some of
the methods are unconditionally stable. In an unconditionaly stable method,
solution does not blow up choosing a big time step. However, the accuracy
of the solution might be aﬀected. To achieve an accurate solution a small
time step must be chosen based on the natural period of the system as well
as the characteristics of the exciting force.
All the above integration schemes are for the solution of a single equation
of motion. While, in the couples damreservoir equation we are dealing with
a couple of equilibrium equation. The stability of an integration scheme for a
single equation does not guarantee the stability of a coupled equation. There
fore, It is needed to develope an approach for solution of coupled equation
using the above methods.
4.4. USINGNEWMARKβ METHODFORTHECOUPLEDEQUATIONS117
4.4 USING NEWMARKβ METHOD FOR
THE COUPLED EQUATIONS
Direct integration scheme is used to ﬁnd the displacement and hydrody
namic pressure at the end of the time increment i +1 given the displacement
and hydrodynamic pressure at time i. The Newmarkβ method is used for
discretization of both equations (implicitimplicit method). In this method
{
.
U
}
i+1
, {U}
i+1
, {
.
P
}
i+1
and {P}
i+1
can be written as follows:
{
.
U
}
i+1
= {
.
U
}
p
i+1
+ γ∆t{
..
U
}
i+1
(4.3)
{
.
U
}
p
i+1
= {
.
U
}
i
+ (1 −γ)∆t{
..
U
}
i
{U}
i+1
= {U}
p
i+1
+ β∆t
2
{
..
U
}
i+1
(4.4)
{U}
p
i+1
= {U}
i
+∆t{
.
U
}
i
+ (0.5 −β)∆t
2
{
..
U
}
i
{
.
p
}
i+1
= {
.
p
}
p
i+1
+ γ∆t{
..
p
}
i+1
(4.5)
{
.
p
}
p
i+1
= {
.
p
}
i
+ (1 −γ)∆t{
..
p
}
i
{p}
i+1
= {p}
p
i+1
+ β∆t
2
{
..
p
}
i+1
(4.6)
{p}
p
i+1
= {p}
i
+∆t{
.
p
}
i
+ (0.5 −β)∆t
2
{
..
p
}
i
where γand β are the integration parameters.The governing ﬁeld equations
at time i + 1 can be written as follows:
[M]{
..
U
}
i+1
+ [C]{
.
U
}
i+1
+ [K]{U}
i+1
= {F
1
}
i+1
+ [Q]{p}
i+1
(4.7)
[G]{
..
p
}
i+1
+ [C
0
]{
.
p
}
i+1
+ [K
0
]{p}
i+1
= {F
2
}
i+1
−ρ[Q]
T
{
..
U
}
i+1
(4.8)
The coupled ﬁeld equations 4.7 and 4.8 can be solved using the stag
gered solution scheme. The procedure can be started by guessing {P}
i+1
118CHAPTER4. DYNAMICANALYSIS OF DAMRESERVOIRSYSTEM
in equation 4.7 to solve for {U}
i+1
and its derivatives. Then equation 4.8
can be solved to ﬁnd {P}
i+1
. This method can not guarantee the uncondi
tional stability of the solution. Similarly, guessing{
..
U
}
i+1
at ﬁrst to calculate
{P}
i+1
from equation 4.8 and then calculating {U}
i+1
from equation 4.7 can
not provide unconditionally stable procedure.
In the following sections, two methods of staggered solution are proposed
which are shown to be unconditionally stable.
4.5 STAGGEREDDISPLACEMENTMETHOD
In this method, equation 4.7 can be approximated as following:
[M]{
..
U
}
∗
i+1
= {F
1
}
i+1
+ [Q]{p}
p
i+1
−[C]{
.
U
}
p
i+1
−[K]{U}
p
i+1
(4.9)
Combining equations 4.9 and 4.7 gives:
[M]{
..
U
}
i+1
= [M]{
..
U
}
∗
i+1
+β∆t
2
[Q]{
..
p
}
i+1
−γ∆t[C]{
..
U
}
i+1
−β∆t
2
[K]{
..
U
}
i+1
(4.10)
Taking advantage of the lumped mass which results in a diagonal mass
matrix, equation 4.10 can be modiﬁed as:
[M]{
..
U
}
i+1
= [M]{
..
U
}
∗
i+1
+ β∆t
2
[Q]{
..
p
}
i+1
(4.11)
Substituting equation 4.11 into equation 4.8, then:
([G]+ρβ∆t
2
[Q]
T
[M]
−1
[Q]){
..
p
}
i+1
+[C
0
]{
.
p
}
i+1
+[K
0
]{p}
i+1
= {F
2
}
i+1
−ρ[Q]
T
{
..
U
}
∗
i+1
(4.12)
In equation 4.12, the right hand side terms are known, thus, {P}
i+1
can be obtained. In order to correct the approximation made in equation
4.11, {P}
i+1
can be substituted in equation 4.7 to calculate {U}
i+1
and its
derivatives.
Therefore, the procedure of the staggered displacement method can be
summarized by the following steps:
1. Solving equation 4.9 to calculate {
..
U
}
∗
i+1
.
2. Substituting {
..
U
}
∗
i+1
in equation 4.12 to calculate {P}
i+1
.
3. Substituting {P}
i+1
in equation 4.7 to calculate {U}
i+1
and its deriv
atives.
4.5. STAGGERED DISPLACEMENT METHOD 119
4.5.1 Stability of the Staggered Displacement Method
In an unconditionally stable solution method, instability can be attributed
to that of structure. While in a conditionally stable method, the instabil
ity may be due to numerical or structural instability. To show that the
described method of staggered displacement is unconditionally stable, con
sider a modally decomposed system with scalar values. In such a system,
displacement and the pressure must not grow. Thus forµ < 1 we have:
{U}
i+1
= µ{U}
i
{
.
U
}
i+1
= µ{
.
U
}
i
{
..
U
}
i+1
= µ{
..
U
}
i
(4.13)
{p}
i+1
= µ{p}
i
{
.
p
}
i+1
= µ{
.
p
}
i
{
..
p
}
i+1
= µ{
..
p
}
i
(4.14)
Using ztransformation of µ =
1+z
1−z
, the condition for stability requires
that the real part of z is negative (Re(z) ≤ 0 ) and that the RouthHurwitz
criterion is satisﬁed. For β = 0.25 and γ = 0.5, equations 4.3, 4.4, 4.5 and
4.6 become:
{
..
U
}
i+1
=
4z
2
∆t
2
{U}
i+1
{
.
U
}
i+1
=
2z
∆t
{U}
i+1
(4.15)
{
.
U
}
p
i+1
=
2z −z
2
∆t
{U}
i+1
{U}
p
i+1
= (1 −z
2
){U}
i+1
{
..
p
}
i+1
=
4z
2
∆t
2
{p}
i+1
{
.
p
}
i+1
=
2z
∆t
{p}
i+1
(4.16)
{
.
p
}
p
i+1
=
2z −z
2
∆t
{p}
i+1
{p}
p
i+1
= (1 −z
2
){p}
i+1
Rewriting equation 4.7 without the force term, then:
[M]{
..
U
}
i+1
+ [C]{
.
U
}
i+1
+ [K]{U}
i+1
−[Q]{p}
i+1
= 0 (4.17)
Combining equations 4.10 and 4.12 and substituting them into equation
4.8 without the force term, gives:
[G]{
..
p
}
i+1
+[C
0
]{
.
p
}
i+1
+[K
0
]{p}
i+1
+ρ[Q]
T
[M]
−1
([M]+γ∆t[C]+β∆t
2
[K]){
..
U
}
i+1
= 0
(4.18)
120CHAPTER4. DYNAMICANALYSIS OF DAMRESERVOIRSYSTEM
The modally decomposed system is represented by a single degree of
freedom equation. The single degree of freedom equivalent of equations 4.17
and 4.18 will be obtained by substituting the mass, damping and stiﬀness
values m, c and k instead of [M], [C] and [K] in equation 4.17 and g, c
0
andk
0
instead of [G], [C
0
] and [K
0
] in equation 4.18. The coupling matrix
[Q] would be represented by scalar quantity q. The characteristic equation
of the coupled ﬁeld can be written by substituting equations 4.15 and 4.16
into equations 4.17 and 4.18 as follows:
¯
¯
¯
¯
m
4z
2
∆t
2
+c
2z
∆t
+k −q
ρq
m
(m+
∆t
2
c +
∆t
2
4
k)
4z2
∆t
2
g
4z
2
∆t
2
+c
0 2z
∆t
+k
0
¯
¯
¯
¯
= 0 (4.19)
or
a
0
z
4
+a
1
z
3
+a
2
z
2
+a
1
z +a4 = 0 (4.20)
where:
a
0
=
16mg
∆t
2
a
1
=
8mc
0
∆t
3
+
8gc
∆t
3
a
2
=
4mk
0
∆t
2
+
4cc
0
∆t
2
+
4gk
∆t
2
+
4ρq
2
∆t
2
+
2ρq
2
c
m∆t
+
ρq
2
k
m
a
3
=
2c
0
k
∆t
+
2ck
0
∆t
a
4
= kk
0
The RouthHurwitz conditions for stability are:
a
0
> 0 a
1
, a
2
, a
3
, a
4
≥ 0
¯
¯
¯
¯
a
1
a
3
a
0
a
2
¯
¯
¯
¯
> 0
¯
¯
¯
¯
¯
¯
a
1
a
3
0
a
0
a
2
a
4
0 a
1
a
3
¯
¯
¯
¯
¯
¯
> 0
(4.21)
For the structural system of dam and reservoir;m,c, k, g, c
0
andk
0
are
positive quantities. Therefore, a
0
, a
1
, a
2
,a
3
and a
4
are always positive. The
values of the two determinants in equation 4.21 are given as:
a
1
a
2
−a
3
a
0
= (4.22)
32
∆t
5
m
2
c
0
k
0
+
32
∆t
5
mc
02
c +
32ρ
∆t
5
mc
0
q
2
+
16ρ
∆t
4
c
0
q
2
c
+
8ρ
∆t
3
c
0
q
2
k +
32
∆t
5
gc
2
c
0
+
32
∆t
5
g
2
ck
4.6. STAGGERED PRESSURE METHOD 121
+
32ρ
∆t
5
gcq
2
+
16ρ
∆t
4
gc
2
q
2
+
8ρ
∆t
3
gcq
2
k
a
1
a
2
a
3
−a
2
1
a
4
−a
2
3
a
0
= (4.23)
64
∆t
6
(mk
0
√
cc
0
−gk
√
cc
0
)
2
+
64
∆t
6
mc
03
ck +
64
∆t
6
mc
02
c
2
k
0
+
64ρ
∆t
6
mc
02
q
2
k +
64ρ
∆t
6
mc
0
q
2
ck
0
+
32ρ
∆t
5
c
02
q
2
ck
+
32ρ
∆t
5
c
0
q
2
c
2
k
0
+
16ρ
∆t
4
c
02
q
2
k
2
+
16ρ
∆t
4
c
0
q
2
kck
0
+
64
∆t
6
gc
2
c
02
k +
64
∆t
6
gc
3
c
0
k
0
+
64ρ
∆t
6
gcq
2
c
0
k
+
64ρ
∆t
6
gc
2
q
2
k
0
+
32ρ
∆t
5
gc
2
q
2
c
0
k +
32ρ
∆t
5
gc
3
q
2
k
0
+
16ρ
∆t
4
gcq
2
k
2
c
0
+
16ρ
∆t
4
gc
2
q
2
kk
0
All the terms in equation 4.22 and 4.23 are positive. Recalling the condi
tion of stability (equation 4.21), then the method of staggered displacement
is unconditionally stable.
4.6 STAGGERED PRESSURE METHOD
In this method, the pressure can be approximated using equation 4.8 as
following:
[G]{
..
p
}
∗
i+1
= {F
2
}
i+1
−[C
0
]{
.
p
}
p
i+1
−[K
0
]{p}
p
i+1
(4.24)
Substituting equation 4.24 into equation 4.8, there is obtained:
[G]{
..
p
}
i+1
= [G]{
..
p
}
∗
i+1
−ρ[Q]
T
{
..
U
}
i+1
−γ∆t[C
0
]{
..
p
}
i+1
−β∆t
2
[K
0
]{
..
p
}
i+1
(4.25)
or:
([G] + β∆t
2
[K
0
] + γ∆t[C
0
]){
..
p
}
i+1
= [G]{
..
p
}
∗
i+1
−ρ[Q]
T
{
..
U
}
i+1
(4.26)
Substituting equation 4.26 into equation 4.7 with [H] = [G] +β∆t
2
[K
0
] +
γ∆t[C
0
], gives:
122CHAPTER4. DYNAMICANALYSIS OF DAMRESERVOIRSYSTEM
([M] + ρβ∆t
2
[Q][H]
−1
[Q]
T
){
..
U
}
i+1
+ [C]{
.
U
}
i+1
+ [K]{U}
i+1
= (4.27)
{F
1
}
i+1
+ [Q]({p}
p
i+1
+ β∆t
2
[H]
−1
[G]{
..
p
}
∗
i+1
)
Using equation 4.27, the variable {
..
U
}
i+1
can be calculated. Substituting
{
..
U
}
i+1
into equation 4.7 gives {p}
i+1
and its derivatives.
Therefore, the procedure of the staggered pressure method can be sum
marized by the following steps:
1. Solving equation 4.24 to calculate {
..
p
}
∗
i+1
.
2. Substituting {
..
p
}
∗
i+1
in equation 4.27 to calculate {
..
U
}
i+1
.
3. Substituting {
..
U
}
i+1
in equation 4.26 to calculate {p}
i+1
and its deriv
atives.
4.6.1 Stability of the Staggered Pressure Method
For stability check, similar procedure as that used in the displacement method
can be applied. Rewriting equations 4.7 and 4.8 without the force terms,
then:
[M]{
..
U
}
i+1
+ [C]{
.
U
}
i+1
+ [K]{U}
i+1
−[Q]{p}
i+1
= 0 (4.28)
[G]{
..
p
}
i+1
+ [C
0
]{
.
p
}
i+1
+ [K
0
]{p}
i+1
+ ρ[Q]
T
{
..
U
}
i+1
= 0 (4.29)
The characteristic equation of the coupled ﬁeld for a modally decomposed
system with scalar values, can be written by substituting equations 4.15 and
4.16 into equations 4.28 and 4.29:
¯
¯
¯
¯
m
4z
2
∆t
2
+c
2z
∆t
+k −q
ρq
4z
2
∆t
2
g
4z
2
∆t
2
+c
0 2z
∆t
+k
0
¯
¯
¯
¯
= 0
or
a
0
z
4
+a
1
z
3
+a
2
z
2
+a
1
z +a4 = 0 (4.30)
where:
a
0
=
16mg
∆t
4
a
1
=
8mc
0
∆t
3
+
8gc
∆t
3
a
2
=
4mk
0
∆t
2
+
4cc
0
∆t
2
+
4gk
∆t
2
+
4ρq
2
∆t
2
a
3
=
2c
0
k
∆t
+
2ck
0
∆t
a
4
= kk
0
4.7. MODIFIED STAGGERED PRESSURE METHOD 123
The coeﬃcients of the polynomial are all positive. The determinants in
the RouthHurwitz conditions (equation 4.21), give:
a
1
a
2
−a
3
a
0
= (4.31)
32
∆t
5
m
2
c
0
k
0
+
32
∆t
5
mc
02
c +
32ρ
∆t
5
mc
0
q
2
+
32
∆t
5
gc
2
c
0
+
32
∆t
5
g
2
ck +
32ρ
∆t
5
gcq
2
a
1
a
2
a
3
−a
2
1
a
4
−a
2
3
a
0
= (4.32)
64
∆t
6
(mk
0
√
cc
0
−gk
√
cc
0
)
2
+
64
∆t
6
mc
03
ck +
64
∆t
6
mc
02
c
2
k
0
+
64ρ
∆t
6
mc
02
q
2
k +
64ρ
∆t
6
mc
0
q
2
ck
0
+
64
∆t
6
gc
3
c
0
k
0
+
+
64ρ
∆t
6
gc
2
c
02
k +
64ρ
∆t
6
gc
2
q
2
k
0
+
64ρ
∆t
6
gcq
2
c
0
k (4.33)
These terms are all positive. Therefore, given the stability condition of
equation 4.21, the method of staggered pressure is unconditionally stable.
4.7 MODIFIED STAGGERED PRESSURE
METHOD
Most of the available nonlinear solutions assume a diagonal mass matrix for
the purpose of analysis. The staggered displacement method is the most
suitable coupled ﬁeld problem solution procedure for the case of nonlinear
analysis. In the case of the staggered pressure method some diﬃculties may
arise due to added mass eﬀect in equation 4.27 which changes the mass matrix
from diagonal to a full matrix. For this reason the staggered pressure method
was modiﬁed to apply to nonlinear analysis.
The staggered pressure method is modiﬁed by rewriting equation 4.27 in
the following approximate form:
[M]{
..
U
}
i+1
+[C]{
.
U
}
i+1
+[K]{U}
i+1
= {F
1
}
i+1
+[Q] ({p}
p
i+1
+β∆t
2
[H]
−1
( [G]{
..
p
}
∗
i+1
−ρ[Q]
T
{
..
U
}
i
) )
(4.34)
Therefore, the procedure of the modiﬁed staggered pressure method can
be summarized by the following steps:
124CHAPTER4. DYNAMICANALYSIS OF DAMRESERVOIRSYSTEM
1. Solving equation 4.24 to calculate {
..
p
}
∗
i+1
.
2. Substituting {
..
p
}
∗
i+1
in equation 4.34 to calculate {
..
U
}
i+1
.
3. Substituting {
..
U
}
i+1
in equation 4.26 to calculate {p}
i+1
and its deriv
atives.
The modiﬁed staggered pressure method does not guarantee uncondi
tional stability of the solution. In the following analysis, the modiﬁed stag
gered pressure method is used instead of the staggered pressure method and
the results are compared with those obtained using the staggered displace
ment analysis procedure.
4.8 USINGαMETHODFORTHE COUPLED
EQUATIONS
In this method, {
.
U
}
i+1
, {U}
i+1
, {
.
p
}
i+1
and {p}
i+1
can be written same as
Newmarkβ method. The governing ﬁeld equations at time i + 1 can be
written as follows:
[M]{
..
U
}
i+1
+[C]{
.
U
}
i+1
+(1+α)[K]{U}
i+1
= {F
1
}
i+1
+[Q]{p}
i+1
+α[K]{U}
i
(4.35)
[G]{
..
p
}
i+1
+[C
0
]{
.
p
}
i+1
+(1+α)[K
0
]{p}
i+1
= {F
2
}
i+1
−ρ[Q]
T
{
..
U
}
i+1
+α[K
0
]{p}
i
(4.36)
where α is the integration parameter which is introduced in the coupled ﬁeld
equation. The coupled ﬁeld equations 4.35 and 4.36 can be solved using the
staggered displacement solution scheme.
4.8.1 Staggered Displacement Method
In this method, equation 4.35 can be approximated as following:
[M]{
..
U
}
∗
i+1
= {F
1
}
i+1
+[Q]{p}
p
i+1
−[C]{
.
U
}
p
i+1
−(1+α)[K]{U}
p
i+1
+α[K]{U}
i
(4.37)
Combining equations 4.37 and 4.35 gives:
[M]{
..
U
}
i+1
= [M]{
..
U
}
∗
i+1
+β∆t
2
[Q]{
..
p
}
i+1
−γ∆t[C]{
..
U
}
i+1
−(1+α)β∆t
2
[K]{
..
U
}
i+1
(4.38)
4.8. USING αMETHOD FOR THE COUPLED EQUATIONS 125
Taking advantage of the lumped mass which results in a diagonal mass
matrix, equation 4.38 can be modiﬁed as:
[M]{
..
U
}
i+1
= [M]{
..
U
}
∗
i+1
+ β∆t
2
[Q]{
..
p
}
i+1
(4.39)
Substituting equation 4.39 into equation 4.36, then:
([G] + ρβ∆t
2
[Q]
T
[M]
−1
[Q]){
..
p
}
i+1
+ [C
0
]{
.
p
}
i+1
+ (1 + α)[K
0
]{p}
i+1
=
{F
2
}
i+1
−ρ[Q]
T
..
U
}
∗
i+1
+ α[K
0
]{p}
i
(4.40)
In equation 4.40, the right hand side terms are known, thus, {p}
i+1
can
be obtained. In order to correct the approximation made in equation 4.39,
{p}
i+1
can be substituted in equation 4.35 to calculate {U}
i+1
and its deriv
atives.
Therefore, the procedure of the staggered displacement method can be
summarized by the following steps:
1. Solving equation 4.37 to calculate {
..
U
}
∗
i+1
.
2. Substituting {
..
U
}
∗
i+1
in equation 4.40 to calculate {p}
i+1
.
3. Substituting {p}
i+1
in equation 4.35 to calculate {U}
i+1
and its deriv
atives.
It can be shown that the method of staggered displacement is uncondi
tionally stable for the linear coupled equations of the damreservoir system
with structural damping when α = 0. For the nonlinear equations, the
numerical solution is based on piecewise linear solution. The solution sta
bility depends on the length of the time steps and the introduced numerical
damping. The damreservoir interaction representation using the staggered
solution technique using αmethod is suitable for nonlinear fracture analysis
of concrete dams.
The nonlinear seismic analysis of concrete gravity dams includes opening
and closing of the cracks due to the cyclic nature of the earthquake. When
the cracks are closed, cracked elements recover their strength and therefore
the structure gains stiﬀness. As the cracks open, the stiﬀness of the structure
reduces. The eﬀect of opening and closing of cracks introduce high frequency
shock waves into the structure. The numerical diﬃculties due to opening
and closing of cracks can be overcome by using the αmethod (Hilber et al.,
1977; Hilber and Hughes, 1978).
126CHAPTER4. DYNAMICANALYSIS OF DAMRESERVOIRSYSTEM
The αmethod of time integration algorithms introduces numerical damp
ing to the system. It is an eﬃcient method that is accurate in lower modes
and dissipate energy in the higher modes when compared with other time
integration techniques. Thus, using the αmethod ensures that the response
of higher modes is damped out. Direct integration is used to determine the
displacement and hydrodynamic pressure at the time increment i + 1. The
αmethod is used for discretization of both equations of the coupled ﬁeld
problem (implicitimplicit method). The damreservoir interaction represen
tation using the staggered solution technique is introduced to the nonlinear
fracture analysis of concrete dams.
4.9 SEISMIC ENERGY BALANCE
In the design of structure subjected to earthquake loading, the energy equa
tion can be used to study the energy absorbtion of diﬀerent components.
In a satisfactory design, the energy supply must be larger than the energy
demand. In this regard, two approaches can be considered for the energy
equation. Uang and Bertero (1990) used absolute and relative energy for
mulations for a single degree of freedom system. They found that absolute
energy formulation is simple and more straightforward. Filiatrault et al.
(1994) used energy balance to study the nonlinear behaviour of diﬀerent
structures under variable earthquake ground motion. Diﬀerent time step
ping algorithms were used to investigate the eﬀect of numerical damping.
Without the numerical damping, exact energy balance can be achieved. The
two approaches to energy formulation were found to give diﬀerent energy
responses.
The energy equation of the dam structure governed by equation (2.1),
can be written as:
1
2
{
.
Ut
}
T
[M]{
.
Ut
} +
Z
{
.
U
}
T
[C]{dU} +
Z
{r}
T
{dU} = (4.41)
Z
{f
1
}
T
{dU} +
Z
{
..
Ut
}
T
[M]{dU
g
} +
Z
([Q]{p})
T
{dU}
or:
EK +ED +ER = EP +EQ+EH (4.42)
4.10. ACCURACY OF THE SOLUTION SCHEME 127
In equations ?? and 4.42, {r} is the vector of the nonlinear restoring
force. The absolute kinetic energy is EK, the viscous damping energy is
ED, the nonlinear restoring work is ER, the work of preseismic applied
force is EP, the absolute seismic input energy is EQ and the work done by
the hydrodynamic pressure is EH. The relative displacement is {U} while
{U
t
} is the total (absolute) displacement vector {U
t
} = {U} + {U
g
}. {U
g
}
is the ground displacement vector. The restoring energy, ER contributes to
the stored elastic energy in a system EE, and the energy dissipated due to
fracture EF(EF = ER −EE), Thus:
EK +EE +ED +EF = EI
EK and EE contribute to the stored energy while ED and EF represent
the dissipated energy. The input energy EI, is the sum of the seismic input
energy due to the inertial force EQ, hydrodynamic force EH and work of
preseismic applied load EP.
The energy balance error is computed as:
Error =
(EP +EQ+EH) −(EK +ED +ER)
(EQ+EH)
×100
When the damreservoir interaction eﬀects are represented by added masses,
the hydrodynamic energy EH is excluded from the energy equation, EH = 0.
However, energy is added to the seismic input energy EQ and kinetic energy
EK through the mass added to the structural system. In the analysis, the re
sults of the fracture response are presented for the time before the ﬁve percent
energy balance error is reached. The error in the energy balance represents
an excessive amount of damage when numerical damping is introduced.
4.10 ACCURACYOF THE SOLUTIONSCHEME
The accuracy of the staggered solution scheme can be improved by increasing
the number of iterations and/or by decreasing the time step. Increasing the
number of iterations of the staggering scheme is a time consuming process.
The accuracy of the proposed methods is based on the selection of the ap
propriate time step. In all of the following analyses no iterations have been
made for the purpose of improving accuracy. The staggered displacement
method and the modiﬁed staggered pressure method are compared with the
128CHAPTER4. DYNAMICANALYSIS OF DAMRESERVOIRSYSTEM
ﬁnite element solution of example problems for the purpose of evaluating the
accuracy of the analysis.
Chapter 5
NONLINEAR FRACTURE
MODELS OF CONCRETE
GRAVITY DAMS
5.1 INTRODUCTION
The cracking behaviour of concrete dams has been the subject of extensive
research during the past decade because few dams suﬀered severe cracking
during earthquakes. Rescher (1990) indicated that most concrete gravity
dams will experience cracking even under operational loading conditions and
moderate earthquake ground motions. Therefore, the assumption of linear
behavior may not be appropriate in the analysis of the seismic response of
concrete gravity dams.
Concrete dams are distinguished from other structures because of their
size and their interactions with the reservoir and foundation. The results ob
tained from the nonlinear analysis of concrete dams are strongly dependent
on the approach to modelling of these interactions. It is a diﬃcult task to
develop a comprehensive analytical model to include both nonlinearity and
interaction eﬀects. The size eﬀect can also inﬂuence the properties of the dam
concrete. The fracture properties of normal concrete can be determined using
laboratory tests. However, the dam concrete diﬀers from the normal weight
concrete because of aggregate size and its poor strength. Little information is
available on the fracture properties of the dam concrete. The fracture surface
of the dam concrete specimens is characterized by mainly aggregate failure.
129
130CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
Saouma et al. (1991) attempted to measure fracture toughness of a concrete
specimen in the laboratory which was considered to be similar to dam con
crete. They concluded that a deﬁnitive decision cannot be made concerning
the results and their accuracy. Bruhwiler and Wittmann (1990) carried out a
dynamic test to determine the material properties of the dam concrete under
high rate of loading and an initially applied compression load. They found
that the fracture energy of dam concrete is 2 to 3 times higher than that of
ordinary concrete. The reason is related to the tensile strength characteristic
of dam concrete. The tensile behavior of concrete can be divided into two
stages. In the ﬁrst stage, the behavior is linear until the tensile strength is
reached. In the second stage, strain softening behavior is observed. Fracture
energy is sensitive to the tensile stress. In addition, increasing the preloading
decreases the fracture energy.
To understand the nonlinear behaviour of concrete dams, modelling of
the cracking and damage process is needed. Bazant and Oh (1983) proposed
a fracture mechanics approach as a blunt smeared crack band. The proposed
approach represented a signiﬁcant advance in comparison to the linear frac
ture theory. The strain softening of the material was considered based on the
fracture parameters, fracture energy, uniaxial tensile strength and crack band
width. Fracture energy can be determined from the complete stressstrain
curve. Formulas were derived to give the fracture parameters.
Two classes of solutions can be found in the nonlinear study of concrete
gravity dams. Discrete crack approach is the ﬁrst class of solutions which is
based on the variable mesh approach. Two methods of linear elastic fracture
mechanics LEFM, and nonlinear fracture mechanics NLFM, can be used
in this approach. The second class of solutions is the continuum model
in which a ﬁxed ﬁnite element mesh is used. Smeared crack model and
damage mechanics are the two methods of solution in this class. These two
families of fracture model have been investigated in parallel. They also called
global approach and local approach. In the global approach, the cracks are
simulated by discontinuity in the continuum and the stresses are obtained
using linear elastic fracture mechanics(LEFM). This method is coupled to
a numerical technique such as ﬁnite elements or boundary elements. On
the other hand, the local approach of fracture is based on changes in the
constitutive law governing the behaviour of concrete. The integrity of the
structure is not required during the solution process.
Bhattacharjee and Léger (1994) applied NLFM to predict the response of
concrete gravity dams. The experimental work done on a model of a concrete
5.1. INTRODUCTION 131
gravity dam and a small beam specimen conﬁrmed the applicability of the
proposed NLFM approach. The coaxial rotation crack model gives a better
response than the ﬁxed crack model. Léger and Leclerc (1996) studied the
nonlinear response of concrete gravity dams subjected to diﬀerent earthquake
ground motions. They found that the response is sensitive to time variation
of the input motion. Most of the time, cracking response showed that the
crack starts from the downstream side and moves toward the upstream side.
This form of cracking does not promote dam instability. The cracks are either
horizontal or they sloped downward. They found that the vertical ground
motion acceleration component is not critical in seismic cracking response of
dams.
The nonlinear response of a concrete gravity dam with an initial distrib
ution of temperature gradient when subjected to the earthquake was studied
by Léger and Bhattacharjee (1995). They used frequencyindependent added
mass matrix as a representative of damreservoirfoundation interaction. The
reservoir and foundation were modelled as a series of dampers and springs
such that the same response can be obtained for the linear response of the
crest when compared with the case of actual interaction. Under earthquake
excitation, when a rigid foundation is assumed with no reservoir bottom ab
sorption, no crack was observed at the top part of the dam. A crack was
formed at the foundation level.
Bhattacharjee and Léger (1993); Léger and Bhattacharjee (1994) studied
the energy response of concrete gravity dams. They used a stiﬀness pro
portional damping with method of integration. NewtonRaphson iteration
technique was used to remove the unbalanced load. An energy balance error
approach is used as a measure of damage. The seismic analysis of Koyna dam
under both horizontal and vertical components of the earthquake was con
ducted. Without introducing the numerical damping, the analysis stopped
after the ﬁrst few seconds because of energy balance error due to spurious de
formation of some elements. No discrepancies were found in the results of the
analysis before the occurrence of instability, when compared with the case of
= 0.2 in which the analysis was successfully completed. Dissipated fracture
energy is negligible in comparison to other sources of energy dissipation. The
reservoir eﬀect was represented by added mass technique.
The eﬀect of hydrodynamic pressure inside the crack in the seismic analy
sis of concrete gravity dams was investigated by Tinawi and Guizani (1994).
The pressure inside the crack does not change the response of the dam sig
niﬁcantly. It was found that under high frequency content earthquakes, the
132CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
hydrodynamic pressure inside the crack may increase when higher modes are
signiﬁcant. At the base of the dam, the hydrodynamic pressure may be 50%
higher than the hydrostatic pressure.
The nonlinear response of the Pine Flat dam was studied using the dis
crete approach (Wepf et al. 1993). A ﬁctitious crack approach was used to
model the crack tip. Reservoir interaction was modelled using a boundary
element. Linear response of the dam was compared with EAGD84 code
analysis and good agreement was found. The nonlinear response of the dam
including reservoir interaction was strongly aﬀected in comparison with the
added mass approach. The slope of the reservoir bottom strongly inﬂuenced
the nonlinear response. The aggregate interlock eﬀect was found to be im
portant in the ﬁnal cracking conﬁguration of the dam.
The cracking response of a concrete gravity dam when subjected to earth
quake loading can be diﬀerent if nonuniform damping or uniform damping
including the damping due to cracking is considered (Barrett et al., 1991). In
the analysis, the dam was represented by a small number of elements. When
the bottom few elements were cracked, a noticeable change in the response
was observed.
Using diﬀerent computer codes, Singhal (1991) found that the Wester
gaard’s added mass approach yields higher values for crest displacement and
stress than that obtained using other approaches. The reservoir bottom ab
sorbtion and water compressibility did not change the response signiﬁcantly.
Pekau et al. (1991, 1995) and Pekau and Batta (1991) presented a method
to study the cracking of concrete gravity dams using the principle of Linear
Elastic Fracture Mechanics (LEFM) and boundary element mode superpo
sition analysis. The model was checked by a shake table test of cantilever
beam made of gypsum. The impact of cracking surface was modeled as a
load pulse.
Ayari and Saouma (1990) proposed a model for simulation of discrete
crack closure. The model was applied in the dynamic analysis of the Koyna
dam (India) under both horizontal and vertical components of the earth
quake. The results were obtained for 5 seconds only of the earthquake in
which the numerical damping was less than 10%.
Nonlinear seismic response of concrete gravity dams was studied by Skrikerud
and Bachmann (1986). Fracture mechanics analysis using discrete crack ap
proach was applied. The model was capable of initiation, opening, closing
and reopening of discrete cracks. Special treatment was used to model ag
gregate interlock eﬀect. The model was applied to a dam of rigid foundation
5.1. INTRODUCTION 133
with empty reservoir. The crack pattern was found to be very sensitive to the
parameters chosen for the analysis. The ﬁrst four seconds of an artiﬁcially
generated time history was used for the purpose of analysis. The analysis
stopped due to excessive damage. Nonlinear response of concrete gravity
dams was also studied by Feltrin et al. (1990). A rigid foundation was as
sumed for Pine Flat dam and the reservoir interaction was included. The
nonlinearity in concrete behaviour included the strain softening and aggre
gate interlock. Response of the linear model with and without the reservoir
interaction was determined. Nonlinear response of the empty reservoir was
studied by scaling the ground motion until cracking occurred. The cracks
started at the top part from down stream face of the dam near the slope dis
continuity and moved horizontally. A diﬀerent response was observed under
the eﬀect of reservoir interaction. The ﬁrst crack started at the foundation
level and then it followed by a crack at the top part of the dam at the same lo
cation of the crack of empty reservoir case. The crest displacement was found
to be higher than that of the empty reservoir. They concluded that the eﬀect
of damreservoir interaction must be included in the nonlinear analysis.
ElAidi and Hall (1989 a,b) investigated the nonlinear response of concrete
gravity dams. The water cavitation in addition to cracking of concrete was
considered. Despite the diﬃculties involved, the nonlinear model was applied
for the case of preformed base crack, top crack and homogeneous damwithout
any cracks. In the case of homogeneous dam, the top crack initiated at
t = 1.95 sec. Soon after initiation of the top crack, it went through the dam
body and almost separated the top part from the rest of the dam. During
the rest of the analysis, no other cracks were observed and only rocking and
opening and closing of the crack were observed.
Fenves and VargasLoli (1988) proposed a method for damreservoir in
teraction which resulted in a symmetric matrix representation of the total
equation of the system. The nonlinearity of the reservoir was introduced into
the proposed method to investigate the reservoir interaction eﬀect. They
found that the eﬀect of cavitation is not signiﬁcant in the response of the
structure.
Mlakar (1987) studied the nonlinear dynamic behaviour of concrete grav
ity dams using the ADINA code. It was found that the crack ﬁrst started
at the base. Then cracking initiated at the top part near slope discontinu
ity. The cracks near the slope discontinuity propagated instantaneously and
passed through the cross section.
In this chapter the nonlinear fracture response of concrete gravity dams
134CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
due to seismic loading is investigated. The damreservoir interaction is in
cluded in the time domain analysis using the method of staggered displace
ment. Smeard crack approach based on a nonlinear fracture mechanics crack
propagation criterion is used to study the cracking and response of concrete
gravity dams.
5.2 ABRIEF STUDYOF NONLINEARPA
RAMETERS
Modelling the constitutive behaviour of concrete is the most important part
of nonlinear seismic response study of the dams. Analytical models for two
dimensional fracture propagation studies are well developed now, but the con
stitutive behaviour of concrete is the most important part of nonlinear seismic
response study of the dams. Analytical models for twodimensional fracture
propagation studies are well developed now, but the threedimensional appli
cation of the fracture models is still in its infancy. Unlike the circumstance
in arch dams, where the nonlinear joint behaviour in the arch direction could
be a decisive factor in the stability of the structure , the structural response
of gravity dams is mainly determined by gravity action. The concrete gravity
dammonoliths, usually unkeyed or lightly grouted, are expected to vibrate in
dependently under severe ground excitations. Hence, twodimensional plane
stress idealizations seem to be appropriate for nonlinear seismic response
study of concrete gravity dams.
A stateoftheart review on the constitutive models for twodimensional
ﬁnite element crack propagation analysis of concrete gravity is presented in
this part. Special emphasis is put on the applications of the models in seismic
analyses of the dams and the limitations of past investigations are examined.
5.2.1 Finite element models of crack propagation
Two approaches have generally been followed for the spatial representation of
tensile crack propagation in ﬁnite element analysis of concrete structures: the
discrete crack model and the continuum crack model. Both models have been
used over the decades because of the advantages and inconveniences that they
bring to the constitutive models for ﬁnite element crack propagation analysis
of concrete structures.
5.2. A BRIEF STUDY OF NONLINEAR PARAMETERS 135
5.2.2 Discrete crack propagation model ,DCPM,( vari
able mesh)
In the DCPM, global approach, a crack is represented as a discrete gap
along the interelement boundary. The growth of the crack is determined by
strength or fracture mechanics based constitutive models. The progressive
physical discontinuity in the system is reﬂected instantaneously in the ﬁnite
element model by modifying the mesh during the analysis. It is generally
argued that the nonlinear response of concrete dams is dominated by a few
discrete long cracks. From this consideration, the discrete crack model may
be a sensible choice for dam fracture analysis.. The speciﬁc advantages of
the DCPM are the abilities to consider explicitly the water penetration and
uplift pressure on the crackopen surfaces, the aggregate interlock in the
rough cracks, and the direct estimation of crackopening displacement (COD)
proﬁle. The principal disadvantages in applying this model are the diﬃculty
and high computational cost due to continuous change of the ﬁnite element
topology during the analysis, and the unobjective eﬀects of ﬁnite element
mesh and crack length increment. A special case of discrete crack modelling
is the application of special elements to represent the a priori weak joints in
the system, such as the damfoundation interface and construction joints.
5.2.3 Continuum crack propagation models (CCPM)
The continuum crack can be divided into Smeared crack propagation model
(SCPM) and Damage mechanics crack propagation model. In these models
the fracture is idealized to propagate as a blunt front smeared over an en
tire element or a certain band width of the element. After initiation of the
fracture process, determined by a suitable constitutive model, the precrack
material stress strain relationship is replaced by an orthotropic relationship
with material reference axis system aligned with the fracture direction. The
tension stiﬀness across the crack plane is either eliminated suddenly or a
gradual stressrelease criterion is applied. Thus, only the constitutive rela
tionship is updated with propagation of cracks and the ﬁnite element mesh is
kept unchanged. The advantage of the models lies in its simplicity and cost
eﬀectiveness, although the physical nature of crack representation is ques
tionable. The tendency of the model to cause diﬀused crack pattern and
the directional bias caused by slanted ﬁnite element mesh are still signiﬁcant
computational diﬃculties. However, the model has been extensively used in
136CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
the seismic response study of concrete gravity dams.
5.3 Constitutive models for crack propaga
tion
To conduct a comprehensive fracture propagation analysis, the selected con
stitutive model should describe the prefracture material stressstrain behav
ior, the fracture initiation and propagation criteria, and the postcrack be
haviour. It is a usual practice in concrete structure analysis to presume linear
elastic behaviour before the onset of tensile fracture process. The behaviour
of concrete under high compressive loading is predominantly nonlinear. Sev
eral models based on the concepts of elastoplasticity and elastoviscoplasticity
have been proposed, to study the nonlinear behaviour of concrete under
compression. However, the compressive stresses in concrete gravity dams
are expected to be low even under severe ground excitations. A reasonable
assumption of linear elastic behaviour under compressive loading has been
applied in almost all previous investigations. It is almost universal in the con
stitutive models of fracture analysis to assume the initiation of new cracks
in a homogenous structure when the principal tensile stress reaches the ten
sile strength of concrete. Diversity in various models lies in the deﬁnition of
the fracture propagation criteria after the crack has been introduced in the
structure. Major developments in the realms of crack propagation analysis
and their relative merits are presented in the following sections.
5.3.1 Strengthbased criteria
The early investigations on cracking of concrete have mostly applied sim
ple criteria based on the concepts of strength of material (SOM). A crack
is assumed to propagate when the predicted stress or strain at the cracktip
exceeds the critical value representative of the strength of the material. The
crack propagation criterion, in this approach, is identical to the new crack
initiation criterion. A sudden release of stress on the fracture plane is com
monly assumed upon reaching the material tensile strength. The gradual
release of stress with increasing strain has also been used, mainly for numer
ical stability reasons, in ﬁnite element analyses of cracking problems. The
SOM criterion of crack propagation has been used in discrete and smeared
crack propagation ﬁnite element models. Comparison of computed tensile
5.3. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 137
stress with the strength of the material is not rational for a cracked structure
because spurious results may be obtained depending on the size of the ﬁnite
element ahead of the propagating crack. The lack of ﬁnite element mesh
objectivity of the SOM criterion was reported.
5.3.2 Fracture mechanics criteria
Fracture mechanics is the theory dealing with propagation of cracks, based
on the concept of energy dissipation by the structure undergoing fracturing
process. It has been recognized only recently that the failure mechanism in
concrete structures is diﬀerent from the usual strength based concept, due
to the progressive growth of a fracture process. The fracture mechanics of
concrete has drawn signiﬁcant attention of the research community over the
last decade; as a consequence, the literature on the subject is voluminous.
Here, only the principal developments in fracture mechanics, relevant to the
seismic analysis of concrete dams, are reviewed.
Figure 5.1: Modes of failure: (a) mode I  Tensile fracture; (b) mode II 
planar shear fracture; (c) mode III  tearing fracture
Three elementary modes of failure are recognized in the fracture theory
(Figure 5.1). For two dimensional idealizations, modes I and II (the opening
mode and the planar shear mode) are usually considered. The third one,
the tearing mode (mode III), is considered for three dimensional fracture
propagation studies. Fracture mechanics crack propagation models can be
138CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
broadly classiﬁed into two categories the linear elastic fracture mechanics
(LEFM) models and the nonlinear fracture mechanics (NLFM) models.
According to LEFM, the fracture process occurs right at the cracktip and
the entire volume of the material remains elastic (Figure5.2a). The stress
ﬁeld around the tip of a sharp crack is characterized by the stress intensity
factors, Ki, determined from linear elastic solutions:
K
I
K
II
K
III
= lim
r→0,θ→0
√
2πt
σ
22
σ
12
σ
23
where σ
ij
are the near cracktip stresses, r and θ are polar coordinates, and
K
i
are associated with three fundamental fracture modes. Once the stress
intensity factors, K
i
, have been numerically (or analytically) computed and
the material fracture toughness (sometimes named critical stress intensity
factor), K
1c
, experimentally determined, a suitable functional relationship is
applied for propagation of an existing crack:
f(K
I
, K
II
, K
III
, K
1c
) = 0 (5.1)
Several functional forms of equation 5.1 have been proposed in the liter
ature. In LEFM models, a sudden release of stress on the surface is assumed
with the extension of the crack. Most investigators adopt the discrete crack
propagation ﬁnite element model (DCPM) with the LEFM constitutive mod
els. Techniques to apply the LEFM criteria in smeared crack propagation
ﬁnite element model have also been reported in the literature. Pekau et al.
(1991) have proposed a numerical procedure to apply the linear elastic frac
ture mechanics criterion in discrete crack propagation analysis with boundary
element model for the dams.
The question of whether the fracture process in concrete can take place at
a localized point has been a subject of intense debate for quite long time. In
reality, the fracture process zone (FPZ) must have some ﬁnite size (Figure5.2
b). It is argued that the LEFM could be applied if the FPZ is much smaller
than the dimension of the structure under consideration. Very large concrete
structures like dams are usually cited as the possible candidates for appli
cation of LEFM models. However, no rational experimental evidence has
ever been put forward appraising the extent of FPZ in dam concrete. The
disregard of nonlinear behaviour in the FPZ is an assumption of unknown
consequences in inﬂuencing the global response of the structure. It seems
5.3. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 139
Figure 5.2: Fracture process zone (FPZ); (a) LEFM; (b) NLFM
140CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
appropriate to consider the nonlinear behaviour in the FPZ if the localiza
tion of the crack proﬁle is a primary objective of the ﬁnite element analysis.
In a gravity dam, a relatively stiﬀ structure, crack opening displacements
may be very small, which means that a long fracture process zone may exist
(Dungar et al. 1991). Hence the argument of small fracture process zone in
comparison to the thickness of the structure, usually cited to apply LEFM
models, may not be true even for concrete gravity dams. The choice between
LEFM and NLFM models could also be inﬂuenced by the strain rate under
consideration.
The primary characteristic of nonlinear fracture mechanics (NLFM) is
the recognition of the strain softening behaviour of concrete in the FPZ.
Two apparently diﬀerent models have been proposed in the literature con
sidering only the model I nonlinear fracture propagation in concrete. The
most referenced work is due to Hillerborg et al. (1976). They characterized
the existence of FPZ as a ﬁctitious crack lying ahead of the real crack tip
(ﬁgure5.3a). The behaviour of concrete in the FPZ was represented by a
diminishing stress, σ, versus crack opening displacement (COD), δ, relation
ship; the tensile resistance being ceased at a critical COD value, δ, (Figure
5.3b). The area under the σ −δ curve represents the energy, G
f
, dissipated
during fracture Process on unit area:
G
f
=
Z
δ
f
0
σ(δ)dδ
G
f
is a material property and often referred to as fracture energy or speciﬁc
fracture energy. The special feature of the Hillerborg’s ﬁctitious crack model
is the dissipation of energy over a discrete line crack. The basic nature
of the model has made its extensive applications possible in discrete crack
propagation analysis.
The LEFM models, in the context of discrete crack propagation, have also
been modiﬁed to take account of the FPZ through an eﬀective crack length
calculated as the true crack length plus a portion of the ﬁctitious crack. In
some recent studies the key assumption of Hillerborg’s model, that the tensile
stress at the tip of ﬁctitious crack is equal to the tensile strength of concrete
has been modiﬁed using the concept of singular stress distribution at the
ﬁctitious crack tip.
Bazant and Oh (1983) came forward with the argument that the energy
dissipation cannot take place in a diminishing volume, it must involve a ﬁnite
5.3. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 141
Figure 5.3: Nonlinear fracture mechanics models: (a,b) ﬁctitious crack
model,(c,d) crack band model
142CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
volume of the material. The fracture process is thus assumed to propagate
as a blunt front (Figure5.3c). The width of the blunt crack (or the band
of microcracks), w
c
, is assumed to be a material property in this model.
The strain softening behaviour of concrete in the FPZ is represented by a
stressstrain relationship (Figure5.3d) and the fracture energy, G
f
, is given
by
G
f
= w
c
Z
ε
f
0
σ(ε)dε
The inherent characteristic of the proposed crack band model is the
smeared nature of crack distribution over a band width, w
c
, The crack band
width, w
c
, however, has not been determined by any direct experimental in
vestigation. Bazant and Oh (1983), in the original presentation, attempted
to establish the value of w
c
by ﬁtting their proposed model in a series of
stressstrain and fracture energy data collected from the literature. The val
ues of w
c
in the range of single aggregate dimension to six times that size
gave equally good results. The assumption of fracture energy dissipation over
a certain band area of material characteristic dimension, thus, seems to be a
numerical speculation.
An important aspect of the nonlinear fracture mechanics models is the
shape of the softening branch. Various proposals have been made in the
literature about the form of σ −δ and σ −ε softening relationships. A bilin
ear relationship is usually applied to interpret the experimental observations
(Briihwiler 1990; Nomura et al. 1991). The precise shape of the softening
diagram has been reported to have considerable inﬂuence on the numerical
results (ACI 1991). There has been an unrelenting debate as to which soft
ening model, the ﬁctitious crack model or the crack band model, should be
used. The smeared nature of the crack band model is a tempting feature for
application in ﬁnite element analysis when the direction and location of crack
propagation are not known a priori. Application of the crack band model, in
its original form, to smeared fracture propagation analysis requires the size
of ﬁnite elements to be limited to w
c
. The very small value of w
c
, according
to the deﬁnition presented by Bazant and Oh (1983), renders any practical
ﬁnite element analysis of large concrete gravity dams too expensive.
Special ﬁnite element techniques have been proposed to ease this limit
on element size, where the fracture process eﬀects are smeared over a zone
of the ﬁnite element and the average stressstrain relationship is adjusted to
conserve the fracture energy. A linear strain softening relationship has been
5.3. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 143
Figure 5.4: (a) average stressstrain curve for smeared crack element; (b)
characteristic dimension, l
c
= l
1
, l
2
; (c) characteristic dimension. lc=
√
l
0
l
00
144CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
assumed in most of these models. The area under the average stressstrain
curve for a ﬁnite element undergoing fracture process is adjusted such that
the dissipated fracture energy,G
f
, for unit area of crack extension remains
independent of the element characteristic dimension, l
c
, (Figure5.4a) . A
favored approach in the nonlinear smeared fracture models (NSFM) is to
adjust the slope of the softening branch, assuming that the softening process
initiates when the tensile stress reaches the material tensile strength. The
strainsoftening modulus, E
t
, in ﬁgure5.4a can be derived as:
E
t
=
σ
2
t
E
σ
2
t
−
2EG
f
l
c
(5.2)
if the tensile strength, σ
t
, the elastic modulus, E, and the strain fracture
energy, G
f
, are known for the material, the strain softening modulus for
the particular element size, l
c
, can be determined from equation 5.2. For
a special case of , l
c
= w
c
(w
c
is the width of crack band), the nonlinear
smeared fracture model reduces to the crack band model of Bazant and Oh
(1983). Unlike the crack band width, w
c
, the characteristic dimension, l
c
, is
a geometric property of the element. For cracks parallel to a side of square
ﬁnite elements, as in Figure5.4b, the characteristic dimension, l
c
, equals the
element dimension across the crack plane (l
c
=l
1
, l
2
,l
3
, etc.) For oblique crack
propagation, l
c
can be taken as the square root of the element area under
consideration (Figure5.4c).
The softening modulus, E
t
, given by equation 5.2, becomes stiﬀer for
increasing value of l
c
up to a certain limit, beyond which an unrealistic snap
back appears in the tensile stress strain relationship of concrete. In the limit
case, the softening constitutive model degenerates to the traditional elas
tobrittle failure criterion, dissipating the stored elastic strain energy instantly
upon reaching the tensile strength of material. The maximum ﬁnite element
size, denoted by l
0
, that can be modelled with strainsoftening behaviour is
determined from equation 5.2:
l
0
≤
2EG
f
σ
2
t
For typical dam concrete properties of E = 30 000 MPa, G
f
= 0.2 N/mm,
and σ
t
= 2.0MPa, the limiting value is l
0
≤ 3.0m. This limit on maximum
dimension, which is much higher than the material characteristic dimension,
w
c
, of Bazant and Oh (1983), also has often been considered stringent for
large scale ﬁnite element analysis at reasonable cost.
5.3. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 145
Figure 5.5: Nonlinear fracture mechanics in smeared crack propagation model
146CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
To circumvent this limit on the size of ﬁnite elements, and at the same
time respect the principle of conservation of energy, one proposition is to
reduce the fracture initiation stress, σ
0
, with increasing ﬁnite element size
and assume an elastobrittle failure criterion for element sizes greater than
l
0
(Figure5.5a). This is the socalled size reduced strength (SRS) criterion
proposed by Bazant (1984) and Bazant and Cedolin (1979, 1983). This size
reduced strength criterion (or in other words the elastic fracture criterion)
can be criticized for two reasons: (i) the size independent critical COD value,
δ
f
, of no tensile resistance, implied by the conservation of fracture energy in
the nonlinear fracture model, is violated (Figure5.5b), and (ii) when applied
with a strength based crack initiation criterion, the principle of fracture en
ergy conservation is likely to be violated in the interior element as well as in
the exterior element (ﬁgure5.5c). An important numerical side eﬀect of the
elastobrittle SRS failure criterion is the generation of spurious shock waves
in the ﬁnite element model. ElAidi and Hall (1989a) encountered such nu
merical diﬃculties in the nonlinear seismic analysis of concrete gravity dams.
The alternate proposition of Bazant (1985) is to take the softening modu
lus, E
t
, as a material property and reduce the fracture initiation stress with
increasing size of element (Figure5.5d) for conserving the fracture energy.
This approach may allow numerically stable algorithm as opposed to the
elastobrittle model, but it seems to be based on a weaker theoretical con
sideration, and the questions raised for the elastobrittle criterion are still
present. Moreover, taking E
t
, as a material property has not been justiﬁed
from experimental investigations. The limit on the maximum size of ﬁnite
elements thus appears to be a requirement to ensure a reliable application
of the nonlinear fracture mechanics criteria in smeared crack propagation
analysis.
Several other models with some variations of the crack band model have
also been proposed in the literature. Gajer and Dux (1990) decomposed the
ﬁnite element strain increment in the following form:
∆ε = ∆ε
e
+ α∆ε
cr
(5.3)
where the incremental strains ∆ε
e
and ∆ε
cr
correspond to the uncracked
concrete and the crack band respectively, and α is the averaging factor deﬁned
as the ratio between the crack band area and the gross element area. For
decomposition of strain in the fracturing direction, equation 5.3 essentially
represents the nonlinear smeared fracture model described earlier. Gajer
5.3. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 147
and Dux (1990) apparently ignored the directional property of cracking by
applying the scalar factor, α, in the global ﬁnite element direction. An
enhanced form of the crack band model is the socalled nonlocal cracking
model (Bazant and Lin 1988). In that approach, the crack strain at an
integration point is smeared over the neighboring points of the ﬁnite element
mesh. A crude form of the nonlocal model can be considered to be the
nonlinear smeared fracture model, described earlier, where the crack strain
is averaged over a characteristic dimension of single element.
Figure 5.6: Closing and reopening of partially formed cracks
Application of nonlinear fracture mechanics models in dynamic analysis
requires the deﬁnition of unloading and reloading behaviours during the frac
ture process. Very few studies have been reported in the literature in this
respect. Bazant and Gambarova (1984) proposed a hypothetical nonlinear
148CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
stressstrain relationship for closing and reopening behaviours of partially
open cracks as depicted in Figure 5.6a. Numerical simulation of such non
linear model is very complex. de Borst and Nauta (1985) applied the linear
tangent softening modulus (line 12 in Figure5.6b) to characterize the FPZ
behaviour of increasing strain with decreasing stress and a secant formula
tion (line 20 in Figure5.6b) to represent the closing of partially open cracks.
Gambarova and Valente (1990) have applied an assumption of sudden stress
release when the closing of partially open cracks is detected at any instant of
the fracture process (Figure5.6c). Dahlblom and Ottosen (1990) proposed
the following relationship for closing and reopening behaviours of partially
fractured concrete:
ε = [λ + (1 −λ)
σ
σ
max
]ε
max
(5.4)
where λ is the ratio between the residual strain upon closing of cracks
and the strain of open cracks (Figure5.6d). It appears that the techniques
applied by de Borst and Nauta (1985) and Gambarova and Valente (1990)
are subsets of this generalized model with λ = 0.0 and 1.0 respectively. The
physical phenomenon of crack closing and reopening taking place before the
complete fracture of the material is, however, yet to be investigated rationally.
5.3.3 Shear resistance of fractured concrete
After the initiation of fracture process on a plane perpendicular to the direc
tion of major principal tensile stress, it is not unlikely to expect the rotation
of principal stress directions under varying deformation modes. Hence, shear
deformation may take place on the partially formed rough fracture plane.
The deﬁnition of shear stressstrain behaviour of concrete during the fracture
process seems to be diﬃcult. Bazant and Oh (1983) ignored the shear defor
mation on the fracture plane, and the material stiﬀness matrix was derived
for normal strains only. This formulation is not compatible with the linear
isotropic stiﬀness matrix of initial state. In another proposition, Bazant and
Gambarova (1984) proposed a nonlinear mathematical model, named the
crack band microplane model, to account for shear in the partially formed
crack, propagating as a blunt front of microcracks. Gambarova and Va
lente (1990) retained the initial shear modulus unchanged until the complete
fracture had taken place and then applied an aggregate interlock model. de
Borst and Nauta (1985) and Gajer and Dux (1990) applied the concept of
5.4. POSTFRACTURE BEHAVIOUR OF CONCRETE 149
simple shear retention factor, β, proposed by Suidan and Schnobrich (1973),
to derive the tangent shear modulus of the fracture plane. The simpliﬁed
approach of applying a shear retention factor ignores the shear dilation and
the dependence of crack shear stiﬀness on the crack opening displacement
(COD). Dahlblom and Ottosen (1990) assumed a linear relationship between
shear resistance and COD. The models proposed by Bazant and Gambarova
(1980), Chen and Schnobrich (1981) Reinhardt and Wairaven (1982), Riggs
and Powell (1986), and Wairaven (1981) represent the shear resistance and
dilation on crack open surfaces of concrete. The models are very often com
putationally inconvenient because of the nonsymmetric stiﬀness matrix. A
comparative study on diﬀerent rough crack models is available in Feenstra et
al. (1991a, 1991b). A special numerical problem associated with the shear
retention in smeared crack ﬁnite element model is the spurious tension stiﬀ
ness in the direction across the fracture plane (ElAidi and Hall 1989a). This
happens due to the application of continuous shape functions in deriving the
ﬁnite element stiﬀness matrix. Discontinuous shape functions proposed by
Ortiz et al. (1986) for localized failure analysis may be a solution to this
problem.
Investigate on the Shear resistance of fractured concrete.
5.4 Postfracture behaviour of concrete
The fracture direction is generally ﬁxed based on the principal stress direction
that initiates the ﬁrst crack. An additional crack plane is allowed to form
only when the stress reaches the tensile strength on the plane orthogonal to
the ﬁrst fracture plane. This type of model is stated as ﬁxed or stationary
crack model . In the ”rotating crack” model, proposed by Cope et al. (1980),
the orthotropic material reference axis system is rotated when the principal
stress direction deviates by a certain amount from the direction that initiated
the fracture process. The rotation of physical crack direction does not seem
to be acceptable from common perception. After early opposition to the
rotating crack model, Bazant et al. (Bazant 1983; Bazant and Lin 1988)
adopted the concept based on the argument that the cracks of one direction
may close and lock in shear while cracks of another direction may form.
A special numerical technique to represent nonorthogonal multiple crack
formation has been perfected by de Borst et al. (de Borst 1987; de Borst and
Nauta 1985).
150CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
A very special postfracture problem, associated with dynamic analysis,
is the modelling of contactimpact phenomenon occurring upon closing and
reopening of the cracks. Special numerical techniques to simulate the im
pact behaviour in discrete crack models have been proposed in the literature
(Ayari and Saouma 199 1; Pekau et al. 199 1). ElAidi and Hall (1989a) pre
sented a discussion on numerical diﬃculties arising from high velocity closing
and reopening of cracks in smeared crack analysis No rigorous procedure has
been reported in the literature to deal with shock wave generated by the
sudden change in stiﬀness resulting from the closing and reopening of the
smeared cracks. And the consequences of the shockwave phenomenon in
smeared crack analysis are not conclusively known yet.
5.5 Material parameters for fracture propa
gation analysis
The present development of mathematical models is far ahead of the current
knowledge of material behaviour, specially under transient conditions. Mate
rial parameter data determined from reliable experimental studies is limited
in the literature for dam concrete. Recent experimental investigations, such
as the one by Briihwiler (1990) and Briihwiler and Wittmann (1990) are
revealing signiﬁcant diﬀerences in the mechanical properties of structural
concrete and mass concrete. Ideally, the selection of material properties for
safety analysis of concrete dams should be dealt with on a casetocase basis,
because the material properties are widely varying from dam to dam. How
ever, a review of literature is presented here to establish a reasonable limit
of parametric values.
Poisson’s ratio, ν, and elastic modulus, E, are applied to represent the
elastic behaviour of concrete in all analyses irrespective of constitutive mod
els selected for propagation of cracks. Jansen (1988) suggested the Poisson’s
ratio for 1year old dam concrete between 0.17 and 0.28. A value of 0.20
has been applied almost universally in the past studies. Briihwiler (1990)
observed the reduction of Poisson’s ratio with increasing compressive strain
rate applied to the concrete cylinder. However, the inﬂuences of rate sensi
tive ν may be insigniﬁcant in comparison to the inﬂuences of other material
parameters. The static modulus for 1 year old dam concrete was suggested
by Jansen (1988) in the range of 28000−48000 MPa. A signiﬁcant feature of
5.5. MATERIAL PARAMETERS FORFRACTUREPROPAGATIONANALYSIS151
concrete constitutive behaviour is the dynamic magniﬁcation of the elastic
modulus under rapidly varying loading condition. A review of this phenom
enon along with the tensile strength property of concrete is presented in the
following section. Separate reviews are presented for commonly used material
parameters in three major crack propagation criteria strength of material, lin
ear elastic fracture mechanics, and nonlinear fracture mechanics. The shear
resistance parameter of fractured concrete is also discussed.
5.5.1 Strengthofmaterial parameters
The governing material parameter in SOMbased fracture propagation mod
els is either critical stress or strain. From a rigorous study with some 12 000
published test results, Raphael (1984) proposed the following relationship
between tensile and compressive strengths of concrete under static loading:
σ
0
t
= 0.324f
0
2/3
c
(5.5)
where f
0
c
and σ
t
are respectively static compression and tensile strengths of
concrete in MPa. In absence of experimentally determined values, the above
equation can be applied as an approximation to the expected static tensile
strength of concrete.
Tensile strength of concrete increases signiﬁcantly with increasing rate of
applied loading, but the failure strain remains more or less unchanged under
varying load rate. In the limited dynamic tests performed on mass concrete,
the dynamic load rate eﬀect has been observed to be higher than that in usual
structural concrete. The selection of dynamic magniﬁcation of the concrete
tensile strength is not very evident from the literature. Raphael (1984), from
his study with the published data, proposed a dynamic magniﬁcation factor
of 1.50 resulting in the dynamic tensile strength of concrete:
σ
t
= 0.48f
0
2/3
c
(5.6)
Briihwiler and Wittmann (1990) observed a dynamic magniﬁcation of up to
80% for the investigated strain rates between 10
−5
and 10
−2
per second, and
this magniﬁcation decreased signiﬁcantly due to compression preloading on
the tested specimens. There is a controversy as to whether the static tensile
strength or the dynamic strength should be used in seismic response analy
sis of concrete dams. Under alternating tensile and compressive loadings,
substantial micro level damage may take place in the material, resulting in
152CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
reduced dynamic tensile strength. Nevertheless, 50% dynamic magniﬁcation
proposed by Raphael (1984) seems to have received wide recognition (Chopra
1988; Kollgaard 1987; NRC 1990).
Figure 5.7: Strengthofmaterialbased failure criterion
A confusion, however, exists about the interpretation of tensile stresses
computed from ﬁnite element analysis. Since the prepeak stressstrain rela
tionship is assumed to be linear elastic in most analyses, some investigators
have suggested to compare the predicted tensile stresses with the appar
ent strength of the material (Figure??a). Experimental evidences seem to
support an apparent static tensile strength, σ
a
, about 30% higher than the
value given by equation 5.5. Under dynamic loading, the nearpeak stress
strain nonlinearity decreases substantially (Briihwiler 1990; Hatano 1960).
Recognizing the fact that the initial tangent modulus does not increase at
5.5. MATERIAL PARAMETERS FORFRACTUREPROPAGATIONANALYSIS153
the same rate as the tensile strength under dynamic loading, the Canadian
Electrical Association (1990) report on safety assessment of concrete dams
recommended the following relationship between E
0
i
and E
i
, the dynamic and
static initial tangent moduli respectively:
E
0
i
= 1.25E
i
(5.7)
Assuming that the failure strain remains the same under static and dynamic
loadings, above relation gives the apparent dynamic tensile strength, σ
0
a
, to
be 25% higher than the apparent static strength, σ
a
. Thus,
σ
0
a
= 1.25(1.3 ×0.324)f
0
2/3
c
= 0.526f
0
2/3
c
(5.8)
Comparing equation 5.6 and equation 5.7, the apparent dynamic tensile
strength is only about 80% higher than the actual dynamic strength, instead
of 30% as proposed by Raphael (1984). Equation 5.8 respects the experimen
tal evidences of rate insensitive failure strain, ε
t
, and reduced nearpeak σ−ε
nonlinearity under dynamic loading. The apparent tensile strength, however,
may have to be lowered by about 10%o −20% due to the existence of rela
tively weak construction lift joints in dams, if such joints are not modelled
explicitly (NRC 1990).
Apparently, no empirical formula is readily available in the literature for
dam concrete to select the initial static tangent modulus,E
i
, from the static
compressive strength, f
0
c
of concrete. Many more experimental investigations
are required to establish an acceptable mathematical model for the eﬀects of
precompression load and the fatigue behaviour under cyclic loading. In the
experimental study by Mlakar et al. (1985), the tensile strength of concrete
under tension compression loading was observed to increase with increasing
rate of loading; the failure strain showed the general tendency of rate in
sensitivity. The behaviour under biaxial loading can be expressed using the
standard failure envelopes in principal stress space (Figure??b).
5.5.2 Linear elastic fracture mechanics parameters
The principal parameter applied in the linear elastic fracture mechanics crack
propagation models is the fracture toughness, K
1c
, of concrete. No deﬁnite
relationship is readily available in the literature of dam concrete to determine
the fracture toughness from standard material parameters such as strength,
elastic modulus, and aggregate size. And only a handful of experimentally
154CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
determined results has been reported so far. Saouma et al. (1989) found a K
1c
value of 1.1 MPam
1/2
. Linsbauer (1990) reported K
1c
, values in the range of
2.0 −3.5 MPam
1/2
. The following guideline has been proposed by Saouma
et al. (1990) to select K
1c
: a zero value as a ﬁrst approximation; should the
response be unacceptable, a value of K
1c
= 1.0 MPa  m
1/2
is used; and if
this value still results in unacceptable crack lengths, laboratory experiments
could be performed on recovered core specimens. Due to multiaxial conﬁning
stresses in the ﬁeld conditions, the in situ values for fracture toughness,
K
1c
, were determined as three times the unconﬁned laboratory test values
(Saouma et al. 1989, 1991b). The fracture toughness can also be estimated
from the following wellknown relationship (Irwin 1957):
K
1c
=
p
G
f
E
where E is the elastic modulus and G
f
the fracture energy. The strain rate
sensitivity of K
1c
for mass concrete is not well addressed in the literature.
In the seismic analysis of Koyna Dam (India), Ayari and Saouma (1990)
assumed an arbitrary dynamic magniﬁcation factor of 60 for K
1c
, which seems
too high for concrete. Briihwiler (1990) predicted the rate sensitivity of K
1c
to be lower than that of concrete tensile strength.
5.5.3 Nonlinear fracture mechanics parameters
Fracture energy, G
f
, is the key parameter that is combined with elastic mod
ulus, E, and tensile strength, σ
t
, to deﬁne the entire constitutive behaviour
of concrete in the nonlinear fracture mechanics models. Usually, the ten
sile strength, σ
t
, beyond which a strain softening process is assumed to take
place, is determined from uniaxial or split cylinder tests and the fracture en
ergy, G
f
, from wedge splitting tests (Briihwiler and Wittmann 1990). Values
for σ
t
and E can be selected according to the guidelines used for strength
ofmaterial failure criterion. Empirical relationships have been proposed to
determine the fracture energy from standard material parameters (Bazant
and Oh 1983; Oh and Kim 1989). Those relationships were derived from
the results of laboratory experiments performed with small size aggregates
speciﬁc of structural concrete behaviour. Extrapolation of the test results
from structural concrete does not seem to be realistic to establish the behav
iour of mass concrete that has distinct features of much larger and weaker
aggregates.
5.5. MATERIAL PARAMETERS FORFRACTUREPROPAGATIONANALYSIS155
Limited results have been reported from experimental investigations on
concrete collected from dam construction sites (Briihwiler 1990; Briihwiler
and Wittmann 1990). The G
f
value under static loading condition was de
termined to be in the range of 0.175and 0.310 N/mm; meaning that the G
f
of dam concrete is two to three times larger than that of structural con
crete. Fracture energy values for the specimens subjected to compressive
preloading were found considerably low. The G
f
parameter determined un
der simulated seismic loading rates showed substantial strain rate sensitivity,
and a maximum of 80% dynamic magniﬁcation over the pseudostatic value
was observed. Briihwiler and Wittmann (1990) attributed the rate sensi
tivity of G
f
mainly to the rate sensitivity of σ
t
. That means, the dynamic
magniﬁcation criterion selected for tensile strength can be applied to the
fracture energy of the nonlinear fracture mechanics model. More complex
relationships can also be established, including the rate sensitivity of critical
crack opening displacement, δ
f
(Briihwiler 1990).
Experiments performed by Saouma et al. (1990, 1991a) have shown that
the fracture properties are not aﬀected by the specimen or aggregate sizes
used. In contrary to this ﬁnding, a study performed by H.N. Linsbauer
(Dungar et al. 1991) reported larger G
f
values for larger specimen sizes,
and another study by H. Mihashi (Dungar et al. 1991) reported the increase
of fracture energy with increasing aggregate size. The test results reported
in the literature are thus sketchy at the present time, and they are often in
contradiction to one another.
Laboratory tests performed at the University of Colorado (Briihwiler and
Saouma 1991) showed signiﬁcant reduction of the fracture properties of con
crete with increased water pressure along the crack. The eﬀect of multiple
cracks on fracture energy dissipation phenomenon is not known precisely.
Biaxial and triaxial stressstrain eﬀects on fracture energy dissipation char
acteristic of concrete are also required to be investigated (Kreuzer et al. I 99
1).
5.5.4 Shear resistance of fractured concrete
The fractured concrete derives shear resistance from interlocking of aggre
gates protruding out of the fracture plane. Numerous investigations have
been conducted to determine the phenomenon of aggregate interlock in struc
tural concrete where the aggregates are usually very strong. The physical
characteristic of fractured dam concrete, where the aggregate strength is
156CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
substantially low, is not known clearly. The recent fracture investigation of
dam concrete, performed by Briihwiler and Wittmann (1990), have shown
that the crack does not travel around the aggregates; it goes straight through
them. The roughness on resulting fracture plane, thus, would be mild with
low shear resistance. No speciﬁc information is currently available to es
tablish a shear resistance model. Rigorous parametric analyses are required
to study the sensitivity of nonlinear seismic response of concrete dams to
the,selected shear resistance model.
5.6 NONLINEAR MODELLING OF CON
CRETE DAMS USING DAMAGE ME
CHANICS
Micro cracking in concrete is believed to occur at relatively low levels of
loading. Therefore, cracking progresses in a heterogeneous medium because
of an increase in micro cracking, and because of the linking of various micro
cracked zones. Experiments performed on cement paste as well as concrete
show that micro cracking has an arbitrary orientation. When the load is
increased, macroscopic cracks develop and the crack orientations follow the
principal stress directions in the material.
Proper understanding and mechanical modeling of the damage process of
concrete, brought about by the internal defects, is of vital importance in dis
cussing the mechanical eﬀects of the material deterioration on macroscopic
behavior. Modeling of this phenomenon has triggered intensive research ac
tivities over the past 20 years or so.
The main concept of this theory is to represent the damage state of ma
terial by an internal variable, which directly characterizes the distribution
of micro cracks formed during the loading process. Each damage model
established mechanical equations to describe the evolution of the internal
variables and the mechanical behavior of damaged material. The damage
mechanic model can be divided into isotropic and anisotropic damage mod
els. The isotropic damage mechanics model uses a single scalar parameter
and is based on Lenlaitre’s hypothesis of strain equivalence. It was used with
some success to describe the damage of concrete. Yet, few practical applica
tions to real structures were conducted using this approach. However, it was
experimentally observed that crack growth in concrete structures signiﬁcantly
5.6. NONLINEARMODELLINGOF CONCRETEDAMS USINGDAMAGEMECHANICS157
depends on the direction of the applied stress and strain. Hence, the damage
process in concrete is essentially orthotropic, the isotropic description is a
mere simpliﬁcation.
Compared with the fracture mechanics theory used in the context of dis
crete cracks the continuum models with ﬁxed mesh have the advantage of
avoiding remeshing when ﬁnite elements are adopted. However, when com
pared with the smeared crack approach there seems to be a small advantage.
Researchers have proved that diﬀerent mechanical phenomena can be for
mulated within the same framework of damage mechanics. The swelling
problem is investigated in concrete dams using this concept. Finally, dam
age mechanics permits the easy implementation of any initial damage due to
thermal stresses or any other phenomena, such as alkaliaggregate reactions,
acting on an existing dam.
5.6.1 NUMERICAL PROBLEMS RELATEDTOSTRAIN
SOFTENING
Damage process is associated with strain softening, which is usually accom
panied by sudden transition from a smoothly varying deformation ﬁeld into
a localized band. The treatment of strain softening and localization by ﬁnite
elements could lead to serious diﬃculties, as reported by Simo (1989), Bazant
and Belytschko (1985), and Wu and Freund (1984).
Some of these diﬃculties are:
(1) cracks tend to localize in a band that is generally the size of the
element used in the discretization;
(2) the fracture energy dissipated decreases as the mesh is reﬁned, in the
limit it tends towards zero;
(3) the solution obtained is extremely dependent on mesh size and orien
tation.
To overcome these diﬃculties, the following solutions were proposed:
1 . The strain softening is related to the element size (Pietruszczak and
Mroz 1981). which leads to the meshdependent hardening modulus tech
nique (Bazant and Oh 1983. Pramono and Wiliam 1989; Simo 1989).
2. Nonlocal damage is formulated as proposed by Bazant and Pijaudier
Cabot (1988), Mazars et at. (1991), Saouridis and Mazars (1992).
3. Introduction of viscoplasticity as suggested by Needleman (1988),
Loret and Prevost (1991), Sluys and de Borst (1992).
158CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
4. The use of the second gradient theory in the deﬁnition of the strain
tensor (sluys et al. 1993).
5. Cosserat continuum approach (de Borst 1991; Vardoulakis 1989).
From a purely mathematical viewpoint Simo (1989) showed that the ﬁrst
three solutions can be interpreted as regularization procedures of the dissipa
tion function. PijaudierCabot et al. (1988) performed a comparative study
of these techniques for the propagation of waves in a longitudinal bar.
5.6.2 FUNDAMENTAL EQUATIONS OF DAMAGE
MECHANICS
The concept of the damage mechanics model is based on the dissipation of
energy by means of cracking and loss of rigidity of the material. Any induced
plastic behavior is ignored.
Equilibrium of the system is deﬁned by the thermodynamic potential,
which is considered here as the strain energy W(ε, d) and is a function of the
strain vector {ε} and the damage variable d. Thus, W(ε, d) can form the
basis to formulate the following expressions:
σ
i
=
∂W
∂ε
i
and Y = −
∂W
∂d
where σ
i
is the component of the stress vector in engineering notation; and
Y is the thermodynamic force associated with d, interpreted as a damage
energy release rate. The time derivative of the speciﬁc energy yields
.
W
= {σ}
T
{
.
ε} −Y
.
d
The Clausius Duhem inequality gives a condition on the energy dissipated
W
d
by the damage process, which can be written as:
.
W
d
= Y
.
d
= {σ}
T
{
.
ε}−
.
W
≥ 0 (5.9)
The formulation of a damage model ﬁrst requires the deﬁnition of thresh
old of damage. which is the conditions that initiates the propagation of
damage. Secondly, the evolution of the damage with loading must also be
deﬁned, and it is a function of a measure of strains, stresses or energy.
5.6. NONLINEARMODELLINGOF CONCRETEDAMS USINGDAMAGEMECHANICS159
5.6.3 ISOTROPICDAMAGE MODEL FORCONCRETE
To establish the damage constitutive equation, it is necessary to relate the
damage variable d to the other internal variables by some physical hypothesis.
Here we try to brieﬂy describe an isotropic damage mechanics model. The
hypothesis of strain equivalence of Lemaitre and Chaboche (1978) is empirical
in nature. It states that any constitutive equation for a damaged material
can be derived from the same potentials as that for a virgin material by for a
damaged material by replacing the stresses by eﬀective stresses. The eﬀective
stresses are deﬁned as:
{
−
σ} = [M(d)]{σ} (5.10)
where[M(d)] is in general a symmetric matrix of rank four. For isotropic
damage model equation 5.10 is valid when the strains in the damaged mater
ial are assumed to be equivalent to strains in the virgin material, but possess
a reduced modulus by a factor (1 −d). Figure 5.8C;D shows the basis of this
assumption. In equation 5.10, [M] is a diagonal matrix equal to
1
1−d
I; and I
is identity matrix.[M] represents the damage matrix. Thus, in the context
of strain equivalence, equation 5.10 reduces to:
{
−
σ} =
1
1 −d
{σ} (5.11)
The strain energy for a damaged material is W = 1/2{ε}
T
[C(d)]{ε},
where [C(d)] is a second order matrix of material properties. In such a case,
the evolution of damage decreases the material stiﬀness and the following
expression can be written
[C(d)] = (1 −d)[C
0
] (5.12)
where [C
0
] is initial matrix of undamaged material properties. The stress
vector and the rate of energy dissipation can be easily obtained:
{σ} = (1 −d)[C
0
]{ε}
and
Y =
1
2
{ε}
T
[C
0
]{ε} = −
∂W
∂d
160CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
Since Y is a quadratic function of strain, it is positive and, thus, d is
always increasing as shown in equation 5.9. This is a characterization of the
irreversibility of damage.
Diﬀerent hypothesis have been proposed for isotropic damage model. The
basic assumption in behaviour of the damaged and equivalent undamaged
element will results in diﬀerent model. Among several models, special at
tention will be given to the two models. They are mainly based on two
concepts of the modelling. In the ﬁrst model, EnergyBased Damage Model,
a single damage variable independent of directions of stresses describes the
behavior of concrete. The damage evolution is deﬁned as a function of the
elastic strain energy of an undamaged equivalent material. To ensure mesh
objectivity of the ﬁnite element solution, the softening parameters are made
meshdependent using the energy equivalence concept. This technique leads
to reasonable mesh size and the model is suitable for the analysis of large con
crete structures. While in the second concept, StrainBased Damage Model,
the damage is described by coupling the compression and tension eﬀects to
deﬁne a single damage variable d. This variable is calculated based on a
certain measure of the material’s strain ﬁeld. The objectivity of the ﬁnite
element solution is ensured by introducing the nonlocal description of this
measure by averaging it within an inﬂuence area in the ﬁniteelement mesh.
The size of the inﬂuence area is of the order of three times the aggregate
size. In this case an extremely reﬁned mesh is required. If the aggregate size
varies between 100 and 300 mm, it becomes impractical to model real dams
with a reasonable size mesh since at least three elements are required within
the inﬂuence area of about 300900 mm.
5.6.4 ANISOTROPICDAMAGEMODEL FORCON
CRETE
A G
f
type anisotropic damage model is described in this part for diﬀerent
reasons: when roller compacted concrete is used in dams, the material itself
is initially orthotropic. Therefore, the development of an anisotropic damage
model is essential. In addition, orthotropic damage models allow the mod
eling of joints in the dam and at the damfoundation interface, even though
these problems are not addressed in this paper. It will be shown that by de
veloping an orthotropic damage model, the isotropic model becomes a special
case (Chow and Wang 1987; Ju 1989; Chow and Lu 1991).
5.6. NONLINEARMODELLINGOF CONCRETEDAMS USINGDAMAGEMECHANICS161
Figure 5.8: Material model in the damage mechanics concept; A)eﬀective ar
eas for isotropic and anisotropic damages; B)characteristic length; C)strain
equivalence hypothesis; D)stressstrain curve for equivaalence hypothesis;
E)closingopening criterion; F)initial damage formulation
162CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
When isotropic damage is considered, the eﬀective stress {
−
σ} (Lemaitre
and Chaboche 1978) and the elastic stress {σ} are related by 5.11. If damage
is no longer isotropic, because of cracking, material anisotropy introduces
diﬀerent terms on the diagonal of [M(d)]. If the concept of net area is still
considered, the deﬁnition for d
i
becomes:
d
i
=
Ω
i
−Ω
∗
i
Ω
i
where Ω
i
is tributary area of the surface in direction i; and Ω
∗
i
is lost area
resulting from damage, as shown in Figure 5.8A. The index i(1, 2, 3) corre
sponds with the Cartesian axes x,y and z. In this case the ratio of the net
area over the geometrical area may be diﬀerent for each direction.
The relation between the eﬀective stresses {
∗
σ} and the elastic equivalent
stresses {σ} becomes:
∗
σ
1
∗
σ
2
∗
σ
12
∗
σ
21
¸
¸
¸
¸
¸
=
1
1−d
1
0 0
0
1
1−d
2
0
0 0
1
1−d
2
0 0
1
1−d
1
¸
¸
¸
¸
σ
1
σ
2
σ
12
¸
¸
In this case the eﬀective stress tensor is no longer symmetric and an
anisotropic damage model, based on equivalent strains, results in a nonsym
metric eﬀective stress vector. Various attempts to restore symmetry were
proposed by, Chow and Lu (1991) and Valliappan et al. (1990). They are
based on the principle of elastic energy equivalence. This principle postu
lated that the elastic energy in the damaged material is equal to the energy
of an equivalent undamaged material except that the stresses are replaced
by eﬀective stresses.
If the symmetrized eﬀective stress vector is deﬁned as:
{
−
σ} = {
−
σ
1
−
σ
2
−
σ
12
} = {
∗
σ
1
∗
σ
2
s
∗
σ
2
12
+
∗
σ
2
21
2
}
It can be related to the real stresses by:
−
σ
1
−
σ
2
−
σ
12
¸
¸
¸
=
1
1−d
1
0 0
0
1
1−d
2
0
0 0
r
1
2
³
1
(1−d
1
)
2
+
1
(1−d
2
)
2
´
¸
¸
¸
¸
σ
1
σ
2
σ
12
¸
¸
(5.13)
5.6. NONLINEARMODELLINGOF CONCRETEDAMS USINGDAMAGEMECHANICS163
This equation can be written as {
−
σ} = [M]{σ}. where[M] is damage matrix.
The elastic strain energy stored in the damaged material is equal to:
W
e
d
=
1
2
{σ}
T
[C
d
]
−1
{σ} (5.14)
The elastic strain energy for the equivalent undamaged material is given by:
(W
e
0
)
equivalent
=
1
2
{
−
σ}
T
[C
0
]
−1
{
−
σ} (5.15)
Equating equations 5.14 and 5.15 and substituting for {
−
σ} from equation
5.13 yields:
[C
d
]
−1
= [M]
T
[C
0
]
−1
[M]
which results in the following expression for the eﬀective plane stress material
matrix:
[C
d
] =
E
1−ν
2
(1 −d
1
)
2 Eν
1−ν
2
(1 −d
1
)(1 −d
2
) 0
Eν
1−ν
2
(1 −d
1
)(1 −d
2
)
E
1−ν
2
(1 −d
2
)
2
0
0 0
2G(1−d
1
)
2
(1−d
2
)
2
(1−d
1
)
2
+(1−d
2
)
2
¸
¸
¸
(5.16)
The term C
d
(3, 3) in equation 5.16 represents the shear resistance of the
damaged material. It can be written in a manner similar to that of the
smeared crack approach, i.e., C
d
(3, 3) = βG in which β is the shear retention
factor that is given here as a function of damage scalars d
1
and d
2
.
β =
2(1 −d
1
)
2
(1 −d
2
)
2
(1 −d
1
)
2
+ (1 −d
2
)
2
If d
2
is neglected as it assumed in the smeared crack approach, β can be
expressed as a function of strain in the principal direction.
The constitutive law can be written as:
{σ} = [C
d
]{ε}
where the principal directions of damage are assumed to coincide with the
principal stresses. Transforming to the global axes, the constitutive relation
can be written as:
164CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
[C
d
]
G
= [R]
T
[C
d
][R]
Where[C
d
]
G
is material matrix in the global axes; and [R] is transformation
matrix deﬁned by:
[R] =
cos
2
β sin
2
β sin βcosβ
sin
2
β cos
2
β −sin βcosβ
−2 sinβcosβ 2 sinβcosβ cos
2
β −sin
2
β
¸
¸
and β is angle of principal strain direction. If d
1
= d
2
= d, the following
relation is obtained:
[C
d
] = (1 −d)
2
[C
0
]
which represents the damaged isotropic model. This model diﬀers from equa
tion 5.12 by a square factor because of the energy equivalence.
5.6.5 EVALUATION OF DAMAGE VARIABLE
Now we try to relate the damage variable to the state of an element using uni
axial behaviour of a concrete specimen.. Consider the elastic brittle uniaxial
behavior as shown in Figure 5.9. If a point A
t
on the stress strain curve (σ,
ε) moves to A
2
, because of damage, a certain amount of energy is dissipated
(dW
d
). Under conditions of inﬁnitesimal deformation and negligible thermal
eﬀects, the ﬁrst law of thermodynamics requires:
dW = dW
e
+dW
d
where dW
e
is elastic energy variation; and dW
d
is energy dissipated by dam
age. It can be seen that after damage the strain will reach to its original
strain at zero. The total dissipated energy is calculated as:
W
d
=
Z
dW
d
= Area(OA
1
A
2
O)
The upper bound of W
d
is total available energy of the material g
t
given
by:
5.6. NONLINEARMODELLINGOF CONCRETEDAMS USINGDAMAGEMECHANICS165
Figure 5.9: Stressstrain curve for energy dissipation due to fracture
166CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
g
t
=
Z
x
0
σdε (5.17)
The fracture energy per unit surface G
f
is deﬁned by:
G
f
= l
ch
g
t
(5.18)
and l
ch
characteristic length of the element of volume representing the mate
rial’s average behavior.
For uniaxial loading, the relation between the strain energy stored in the
damaged material W
e
d
, and the elastic energy in the virgin material W
e
0
is:
W
e
d
= (1 −d)
2
W
e
0
= (W
e
0
)
equivalent
(5.19)
Therefore, the eﬀective material modulus is:
−
E
= (1 −d)
2
E
0
5.6.6 Damage evolution for concrete subjected to ten
sile strain
An element of volume of the material which can be representative of the
global behaviour is now considered. This volume will be characterized by
its length which provides a measure of the region over which the damage
is smeared so that the global response is reproduced by this volume. This
length, l
ch
, is called characteristic length and it should be measured in the
direction normal to a potential crack plane ( Figure 5.8B).
To deﬁne the damage evolution of concrete, The principal strains are
measured for an element to determine the state of the system. The initial
threshold is the strain beyond which damage can occur and is given by:
ε
0
=
f
0
t
E
0
Using equation 5.19, a possible expression for damage is:
d = 1 −
s
W
e
d
W
e
0
(5.20)
5.6. NONLINEARMODELLINGOF CONCRETEDAMS USINGDAMAGEMECHANICS167
where W
e
d
is recoverable elastic energy of the damaged material. When the
deformations are less than the initial threshold (ε
0
) the material is elastic
and all the energy is recoverable, which implies that W
e
d
= W
e
0
, and therefore
d = 0. In the limit of damage, W
e
d
→0, which implies d = 1. A simple way to
evaluate d is to adopt a postpeak stress function. According to Rotls (1991),
the strainsoftening curve of concrete must be concave. By considering an
exponential function as proposed by Lubliner et al. (1989):
σ(ε) = f
0
t
[2 exp(−b(ε −ε
0
)) −exp(−2b(ε −ε
0
))] (5.21)
The constant b can be evaluated using equations 5.17, 5.18, and 5.21, which
leads to:
b =
3
ε
0
(
2G
f
E
l
ch
f
0
t
−1)
≥ 0 (5.22)
Using 5.20 and 5.21, the evolutionary model for damage can be expressed
as:
d = 1 −
r
ε
0
ε
[2 exp(−b(ε −ε
0
)) −exp(−2b(ε −ε
0
))] (5.23)
If a simple linear softening curve is assumed, Figure 5.8D can show that:
d = 1 −
s
ε
0
ε
−
µ
ε −ε
0
ε
f
−ε
0
¶
³
ε
0
ε
´
The fundamental issue of this approach lies in the introduction of a geomet
rical factor, l
ch
, in the constitutive model. When the ﬁnite element method
is used, a socalled meshdependent hardening modulus is obtained. This
technique was proposed by Pietruszcak and Mroz and was employed by a
number of authors. Using equation 5.18 ensures conservation of the energy
dissipated by the material. Amongst many strategies used to ensure mesh
objectivity, the meshdependent hardening technique is the most practical
for mass structures such as dams. Condition in equation 5.22 should be
interpreted as a localization limiter on the characteristic length of the vol
ume representing the global behaviour of the material. In other words, if
l
ch
≥
2EG
f
f
02
t
, it is not possible to develop strain softening in the volume. For
mass concrete, using average values for E = 30000 MPa, G
f
= 200 N/m and
f
0
t
= 2MPa yields a limit for l
ch
= 3 m. This leads therefore to a reasonable
168CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
mesh size for dam models. The introduction of such parameter is not for nu
merical convenience. The characteristic length can be related to the Fracture
Process Zone (FPZ) commonly used in ﬁctitious crack model for concrete.
5.6.7 Opening and closing of the crack and initial dam
age
When the strain is increasing, damage will also increase. During cyclic load
ing the strains are reversed and unloading will occur. Experimental cyclic
loading of concrete in tension shows evidence of permanent strain after un
loading. Under compression the material recovers its stiﬀness. The classical
split of the total strain in a recoverable elastic part ε
e
and an inelastic strain
ε
in
gives:
ε = ε
e
+ ε
in
= ε
e
+ λε
max
where ε
max
is the maximum principal strain reached by the material and λ is
a calibration factor varying from 0 to1. Figure 5.8E illustrates this criterion.
The value λ = 0.2 is selected; the unloadingreloading stiﬀness becomes :
E
unl
= E
0
(1 −d)
2
(1 −λ)
(5.24)
When the principal strain is less than ε
in
the crack is considered closed.
Alternatively, the crack will open when the principal strain is greater than
λε
max
.
The damage model presented here is based on three parameters: the
elastic modulus E, the initial strain threshold ε
0
and the fracture energyG
f
.
If the element of volume is initially damaged (d = do), the secant modulus
and the fracture energy are reduced by (1 − do) so the eﬀective values are
(Figure 5.8F):
E
0
= (1 −d
0
)E G
0
f
= (1 −d
0
)G
f
5.6.8 ANALYTICAL PROCEDURES IN A FINITE
ELEMENT MODEL
In a ﬁnite element model, the fournode isoparametric element is preferred in
the implementation of the local approach of fracturebased models and has
5.6. NONLINEARMODELLINGOF CONCRETEDAMS USINGDAMAGEMECHANICS169
been used for the implementation of the described constitutive model. The
standard local deﬁnition of damage is modiﬁed such that it refers to the status
of the complete element. The average of the strain at the four Gauss points
is obtained and the damage is evaluated from the corresponding principal
strains of the element. The constitutive matrix [C
d
] is updated depending
on the opening/closing and the damage state of the element. The stresses
at each integration point are then computed using the matrix [C
d
] and the
individual strains.
The characteristic length of the element is calculated approximately, us
ing the square root of its total area. For an eﬃcient control of the damage
propagation in the ﬁnite element mesh, some adjustments have to be consid
ered. During the softening regime the stiﬀness is reduced as a consequence
of damage ,equation 5.23. At the unloading stage the secant modulus is cal
culated using equation 5.24 until closing of the crack. When the crack closes
the material recovers its initial properties. In the reloading regime the last
damage calculated before unloading is used again.
Finite element implementation of the damage mechanics
1. Compute total displacement {u}
i+1
= {u}
i
+{∆u}
2. Initialization Iselect = 0, E
max
= 0
3. Loop over elements e = 1, nel
(a) Compute deformations for each Gauss point : {²}
gp
e
= [B]
gp
{u}
e
(b) If the element is already damaged call subroutine UPDATE
(c) If not: (selection of element with largest energy density)
Compute ε
1
, if ε
1
≤ ε
0
go to (d)
Compute E
e
=
1
2
{σ}
e
{ε}
e
if E
e
≥ E
max
then E
max
= E
e
, Iselect = e
(d) {σ}
e
= [C
d
]{ε}
e
(e) Compute the internal force vector r
e
=
R
A
e
[B]
T
{σ}
e
dA
f) Assemble the element contribution r ←r
e
4. If (Iselect 6= 0), (New element damaged)
(a) Correction of stresses due to damage
(b) Update data base of damaged elements
Subroutine UPDATE (Stress and damage update)
1. Set damage parameter d = d
old
2. Compute the average strains within the element and the corresponding
principal strains: ε
1
, ε
2
set ε = Max(ε
1
, ε
2
)
3. Check for loading/unloading and closing/opening of crack:
(a) if (ε ≥ ε
max
) the element is in a loading state
170CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
Calculate d from equation 5.23
update if ²
in
, ε
max
= ε
(b) else if ε ≥ ε
in
)
d = d
old
(c) else d = 0
4. Compute[C
d
] using equation 5.16
5.7 CONSTITUTIVEMODEL FORSMEARED
FRACTURE ANALYSIS
The constitutive model deﬁning (i) the presoftening material behaviour, (ii)
the criterion for softening initiation, (iii) the fracture energy conservation,
and (iv) the softening, closing and reopening of cracks, and the ﬁnite element
implementation of the formulations are presented in the following sections.
A linear elastic relationship is assumed between compressive stresses and
strains. The tensile stresses and strains are referred to as positive quantities
in the presentation.
5.7.1 Prefracture behaviour
Stresses {σ} and strains {ε} in a linear elastic condition are related as:
{σ} = [D]{ε}
where [D] is the constitutive relation matrix deﬁned for an isotropic plane
stress condition as:
[D] =
E
1 −υ
2
1 ν 0
ν 1 0
0 0
1−ν
2
¸
¸
Here, E is the elastic modulus; and ν is the Poisson’s ratio.
5.7.2 Strain softening of concrete and the initiation
criterion
The stressstrain relationship for concrete becomes nonlinear near the peak
strength [Figure 5.10a]. In the postpeak strain softening phase, coalescence
5.7. CONSTITUTIVEMODEL FORSMEAREDFRACTUREANALYSIS171
Figure 5.10: Constitutive modelling for smeared fracture analysis; a)softening
initiation criterion; b)fracture energy conservation; e)local axis system;
d)closing and reopening of cracks
of the microcracks causes a gradual reduction of the stress resistance. The
area under the uniaxial stressstrain curve up to the peak is taken as the
index for softening initiation:
U
0
=
Z
ε
1
0
σdε =
σ
i
ε
i
2
=
σ
2
i
2E
=
Eε
2
i
2
σ
1
> 0
where σ
i
is the apparent tensile strength, that may be approximately taken
2530% higher than the true static strength, σ
t
. It is calibrated in such a way
that a linear elastic uniaxial stressstrain relationship up to σ
i
will preserve
the value U
0
[Figure 5.10a].
In ﬁnite element analyses, a linear elastic relationship is assumed until
the tensile strain energy density,
1
2
σ
1
ε
1
(σ
1
, and ε
1
are the major principal
172CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
stress and strain, respectively), becomes equal to the material parameter, U
0
:
U
0
=
1
2
σ
1
ε
1
=
σ
2
i
2E
σ
1
> 0 (5.25)
Taking the square roots of both sides, the biaxial eﬀect in the proposed strain
softening initiation criterion is expressed as:
σ
1
σ
i
=
r
σ
1
Eε
1
The above equation is a representative of a biaxial failure envelope. The
principal stress σ
1
, and the principal strain, ε
1
, at the instance of softening
initiation, are designated by σ
0
, and ε
0
, respectively [Figure 5.10a].
Under dynamic loads, the prepeak nonlinearity decreases with increas
ing values for both σ
t
and ε
t
[Figure 5.10b]. The strainrate eﬀect on the
material parameter U
0
is considered through a dynamic magniﬁcation factor,
DMF
e
, as follows:
U
0
0
=
σ
02
i
2E
= (DMF
e
)
2
U
0
where the primed quantities correspond to the dynamic constitutive parame
ters. The increased material resistance due to inertia and viscous eﬀects un
der dynamically applied loads has been explicitly considered in the dynamic
equilibrium equations. Reviewing the literature on the dynamic fracture
behaviour of concrete, a 20 percent dynamic magniﬁcation of the apparent
tensile strength is considered adequate for seismic analyses of concrete dams.
Under dynamic loading, the material parameter U
0
in equation 5.25 is re
placed by the corresponding dynamic value, U
0
0
. At the instant of softening
initiation under a dynamically applied load, the principal stress, σ
1
, and the
principal strain, ε
1
, are designated by σ
0
0
and ε
0
0
, respectively, as shown in
Figure 5.10b.
5.7.3 Fracture energy conservation
The tensile resistance of the material is assumed to decrease linearly from the
presoftening undamaged state to the fully damaged state of zero tensile resis
tance [Figure 5.10b]. The slope of the softening curve is adjusted such that
5.7. CONSTITUTIVEMODEL FORSMEAREDFRACTUREANALYSIS173
the energy dissipation for a unit area of crack plane propagation, G
f
, is con
served. The strainrate sensitivity of fracture energy is considered through a
dynamic magniﬁcation factor, DMF
f
, applied to magnify the static fracture
energy:
G
0
f
= DMF
f
G
f
The strainrate sensitivity of fracture energy can mainly be attributed to that
of tensile strength. DMF
f
can therefore be assumed to be equal to DMF
e
.
In ﬁnite element analyses, the ﬁnal strains of no tensile resistance for static
and dynamic loading are deﬁned as:
ε
f
=
2G
f
σ
0
l
c
ε
0
f
=
2G
0
f
σ
0
0
l
c
where l
c
is the characteristic dimension deﬁned in the previous section.
5.7.4 Constitutive relationships during softening
After the softening initiation, a smeared band of microcracks is assumed to
appear in the direction perpendicular to the principal tensile strain. The
material reference axis system, referred as the local axis system, is aligned
with the principal strain directions at that instant [directions np in Figure
5.10c. The constitutive matrix, relating the local stresses to local strains is
deﬁned as:
[D]
np
=
E
1 −ηυ
2
η ην 0
ην 1 0
0 0 µ
1−ην
2(1+ν)
¸
¸
η =
E
n
E
(5.26)
where the parameter η (0 ≤ η ≤ 1) is the ratio between the softening Young’s
modulus E
n
, [Figure 5.10d], in the direction normal to a fracture plane and
E, is the initial isotropic elastic modulus; and µ is the shear resistance factor.
The following options are considered with respect to the orientation of crack
bands in ﬁnite element analyses.
5.7.5 Coaxial Rotating Crack Model (CRCM)
The local axis system n−p is always kept aligned with the directions of prin
cipal strains, ε
1
, and ε
2
. In this model, the strains ε
n
and ε
p
are, respectively,
174CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
ε
1
, and ε
2
at the newly oriented material reference state. Using an implicit
deﬁnition of the softened shear modulus in cracked elements, the parameter
µ is deﬁned for the CRCM as follows:
µ =
1 + ν
1 −ην
2
µ
ηε
n
−ε
p
ε
n
−ε
p
−ην
¶
0 ≤ µ ≤ 1 (5.27)
Here, ε
n
and ε
p
is the normal strain components in the directions normal and
parallel to the fracture plane, respectively.
5.7.6 Fixed Crack Model With Variable Shear Resis
tance Factor (FCMVSRF)
In this model, the local reference axis system is ﬁrst aligned with the princi
pal strain directions at the instance of softening initiation, and kept nonro
tational for the rest of an analysis. The shear resistance factor, µ, is derived
using the strain components ε
n
and ε
p
corresponding to the ﬁxed local axis
directions (which are not necessarily coaxial with the principal stress di
rections). The deﬁnition of a variable shear resistance factor according to
equation 5.27, that takes account of deformations in both lateral and normal
directions to a fracture plane, is diﬀerent from the usual formulations where
only the crack normal strain is often considered as the damage index.
The total stressstrain relationship matrix, deﬁned in equation 5.26, is
similar to the formulation presented by BaZant and Oh (1983), except that
they have not considered shear deformations in the constitutive relationship.
The present formulation, with the degraded shear modulus term, maintains
a backward compatibility with the presoftening isotropic elastic formulation
when η = 1 and 2. The local constitutive relationship matrix, [D]
np
, can be
transformed to the global coordinate directions as follows:
[D]
xy
= [T]
T
[D]
np
[T]
where [T] is strain transformation matrix deﬁned as follows in terms of the
inclination of the normal to it crack plane, θ [Figure 5.10c]:
[T] =
cos
2
θ sin
2
θ sin θcosθ
sin
2
θ cos
2
θ −sin θcosθ
−2 sin θcosθ 2 sin θcosθ cos
2
θ −sin
2
θ
¸
¸
5.7. CONSTITUTIVEMODEL FORSMEAREDFRACTUREANALYSIS175
With an increasing strain softening, the damaged Young’s modulus, E
n
, (Fig
ure 5.10d), and hence the parameters η and µ decrease gradually, and may
eventually reach zero values after the complete fracture (ε
n
> ε
f
or ε
0
f
). The
constitutive matrix in equation 5.26 is updated as the parameters η and µ
change their values. In the CRCM, a change in the global constitutive re
lationship may also be caused by a rotation of the local axis system, which
is always kept aligned with the directions of coaxial principal stresses and
strains. During unloading/reloading, when the strain, ε
n
, is less than the
previously attained maximum value, ε
max
[Figure 5.10d], the secant modu
lus, E
n
, remains unchanged; the parameter µ, however, may change during
that process.
The change in global constitutive relations is also caused by a rotation
of the local axis system, which is always kept aligned with the directions
of principal strains to keep the principal stresses and strains coaxial. The
CRCM is very eﬀective in alleviating the stress locking generally observed in
ﬁxed crack models. During unload ing/reloading, when the strain, ε
n
, is less
than the previously attained maximum value, ε
max
[Figure 5.10f], the secant
modulus, E
n
, remains unchanged; the parameter µ, however, changes during
that process.
5.7.7 Closing and reopening of cracks
Under reversible loading conditions, the tensile strain, ε
n
, in an element may
alternatively increase and decrease. With the reduction Of ε
n
, the shear
resistance factor, µ, gradually increases. The softened Young’s modulus in
the direction n, E
n
(which may have reached a zero value), is replaced by the
undamaged initial value, E, if the parameter µ is greater than a threshold
value µ
c
. Parametric analyses have shown that the seismic fracture response
of concrete gravity dams is not aﬀected by a value of µ
c
between 0.90 and
0.9999. A relatively ﬂexible tolerance, µ
c
= 0.95, can be used to minimize
spurious stiﬀness changes during the closing of cracks. When ε
n
> 0 in
subsequent load steps, the value µ is determined by using the damaged value
of η to determine the reopening of cracks. If µ becomes less than µ
c
, the
element behaviour is determined by either the reloading or the reopening
path depending on the ﬁnal state attained in previous tension cycles. The
appropriate value of the damage modulus, E
n
, is reused in equation 5.26 at
that state. For µ
c
≈ 1, the residual strain upon closing of a crack is given by
ε
n
= νε
2
s
. Figure 5.10d sshows the closingreopening behaviour for a special
176CHAPTER5. NONLINEARFRACTUREMODELS OF CONCRETEGRAVITYDAMS
case when ε
n
≈ 0.
2
Contents
1 INTRODUCTION 1.1 TYPES OF DAMS . . . . . . . . . . . . . . . . . . . . . . . . 1.1.1 Embankment Dams . . . . . . . . . . . . . . . . . . . . 1.1.2 Concrete Arch and Dome Dams . . . . . . . . . . . . . 1.1.3 Concrete Gravity and GravityArch Dams . . . . . . . 1.1.4 Concrete Slab and Buttress Dams . . . . . . . . . . . . 1.2 APPURTENANT FEATURES OF DAMS . . . . . . . . . . . 1.3 SAFETY OF DAMS AND RESERVOIRS . . . . . . . . . . . 1.4 HOW DAMS ARE BUILT . . . . . . . . . . . . . . . . . . . . 1.5 FAMOUS DAMS OF THE WORLD . . . . . . . . . . . . . . 1.6 POWER GENERATOR, FLOOD CONTROL AND IRRIGATION DAMS . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6.1 Power Generator Dams . . . . . . . . . . . . . . . . . . 1.6.2 Flood Control Dams . . . . . . . . . . . . . . . . . . . 1.6.3 Irrigation Dams . . . . . . . . . . . . . . . . . . . . . . 1.7 INSTRUMENTATIONS AND SURVEILLANCE OF DAMS . 1.7.1 Surveillance . . . . . . . . . . . . . . . . . . . . . . . . 1.7.2 Instrumentation . . . . . . . . . . . . . . . . . . . . . . 1.7.3 Instruments . . . . . . . . . . . . . . . . . . . . . . . . 1.8 ECOLOGICAL/ENVIRONMENTAL CONSIDERATION OF DAM OPERATION . . . . . . . . . . . . . . . . . . . . . . . 1.9 THE HISTORY OF DAMS DESIGN . . . . . . . . . . . . . . 1.9.1 Irrigation Dams . . . . . . . . . . . . . . . . . . . . . . 1.9.2 Dams Designed for Water Supply . . . . . . . . . . . . 1.9.3 Flood Control Dams . . . . . . . . . . . . . . . . . . . 1.9.4 Power Dams . . . . . . . . . . . . . . . . . . . . . . . . 1.9.5 The Moslem World . . . . . . . . . . . . . . . . . . . . 3 9 10 11 13 14 16 16 18 20 21 25 25 26 28 28 28 28 30 33 34 35 36 38 39 40
. . . . . . . .9. . . . . . . .3 Reynold’s Transport Equation . . . . .2. . 2. . . . . . . . . . . . . . 2. . . . 110 3. . . .9. . . . . .3 VISCOSITY . . . . . . . . . . . . .10 BEAVERS . . . . . . . . . . . . . . . . . . . . . . . . .4 Continuity Equation . . . .1 SingleDegreeOfFreedom Systems . . . . . . . . . . . . . 2. . . . 2. . . . . . . . . .1 Velocity Field . . . . . . . . 2. . . .2. . . 2. . .2 MultiDegreeOfFreedom System . . . 99 3. 93 3 FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM 99 3.2. . . . .9. . . . . 101 3. . . . 2. . . . . . . . . .2.2. . 2. . . .3. . . . . . . 2. . . . . .1 FINITE ELEMENT MODELLING OF THE STRUCTURE . .6 The Equation of the Motion . . . . . . . . .11 RESERVOIR FAREND TRUNCATED BOUNDARY CONDITION . 99 3. . . . . .2 ReservoirFoundation Boundary Condition . 2. . 49 . . . . . 2. . .6 Development of the Modern Dams . . . . .1 Truncated Boundary of the Reservoir’s FarEnd . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . .2. . . . . .9. .2 GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION . . . . . . . 2. . . . . . . . 50 50 52 53 56 57 59 64 65 67 71 74 78 79 80 81 84 86 .4 NAVIERSTOKES AND EULER EQUATIONS . . . . . 2. . . . . . . . . . .8 RESERVOIR’S EQUATION OF MOTION .3 FINITE ELEMENT MODELLING OF THE RESERVOIR .2 System and Control Volume . . . . . . . . . . . . . . . . . . 2. . . . .2 COUPLING MATRIX OF THE DAMRESERVOIR . . . . . . . . . . . . . . . 2. . . . . 108 3. .6 BOUNDARYLAYER THEORY . . 42 1.5 COMPRESSIBLE FLUID . 2. . . . . . . . . . . .4 CONTENTS 1. . . . . . . . . . . . . . .10 SOLUTION OF THE RESERVOIR EQUATION . . . .3 Free Surface Boundary Condition . . . . . . .1. . . . . 106 3. . . . . 44 2 RESERVOIR 2. . . . . . . . . . . .1 DamReservoir Boundary Condition .1. . . . . . . . . . . . .1 INTRODUCTION . . . . . . 49 . . . .4 EQUATION OF THE COUPLED DAMRESERVOIR SYSTEM . . . . . . 111 . . . . . . 2. . . . . . . . . . . . . . . . . .7 IRROTATIONAL FLOW . . . .5 Linear Momentum Equation . . . . .9 RESERVOIR BOUNDARY CONDITIONS . . . . . . . . . 2. . . . . . . . . . . . .
135 5. . . . 157 .( variable mesh) . . .5. . . . . . . . . 153 5. . .2 Linear elastic fracture mechanics parameters . .6 NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS . . . . . 134 5. . . . .5. . . . . 135 5. . . . . . .2 A BRIEF STUDY OF NONLINEAR PARAMETERS . .2. . . . 124 4. . 118 4. . .DCPM. .8. . .1 Finite element models of crack propagation . .1 Strengthbased criteria . . . . . . .1 NUMERICAL PROBLEMS RELATED TO STRAIN SOFTENING .5 STAGGERED DISPLACEMENT METHOD .5. . .5. . 148 5. . . 154 5. .3. . . . . 122 4. . . 113 4. . . . . . 151 5. . . . . . . . .3. . . . . . 129 5.2 Fracture mechanics criteria . . . . . . . .4 USING NEWMARKβ METHOD FOR THE COUPLED EQUATIONS . . . . . . . 136 5. . . . .8 USING αMETHOD FOR THE COUPLED EQUATIONS . . . .2. 136 5. . . . .7 MODIFIED STAGGERED PRESSURE METHOD . . . .6. . . . . . .1 Stability of the Staggered Pressure Method . . . . . . . . 124 4. . 156 5. . . . . . . . . . . . . . . . . .1 INTRODUCTION . . . . . . . . . . . . . . . . . . .4 Postfracture behaviour of concrete .3. 117 4. . . .1 Stability of the Staggered Displacement Method . .9 SEISMIC ENERGY BALANCE . . .6. . . . . . .3 Continuum crack propagation models (CCPM) . . . 127 5 NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS 129 5. 149 5. . . . . . 115 4. . . . . . . . . . . . . . .2 Discrete crack propagation model . . . . . 137 5. . . . . . . . . . . . . .1 Strengthofmaterial parameters . . .3 Nonlinear fracture mechanics parameters . . . . . . . . . . . . . . . . . . .2. . . . . . 123 4.10 ACCURACY OF THE SOLUTION SCHEME . . . . . . . . . . . . . . . . .2 THE COUPLED DAMRESERVOIR PROBLEM . . .5. . . . . . . . . . . . . . . . . . . 150 5. . . . . . . . . . . . .CONTENTS 5 4 DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM 113 4.3 Constitutive models for crack propagation . . . .5 Material parameters for fracture propagation analysis . . . . . .1 INTRODUCTION . . . 121 4.3 DIRECT INTEGRATION OF THE EQUATION OF MOTION115 4. . .1 Staggered Displacement Method . 126 4. . . . . . . . . . . . . . 119 4. . . . . . . . .3 Shear resistance of fractured concrete . . . . . 155 5. . . . . . .4 Shear resistance of fractured concrete .6 STAGGERED PRESSURE METHOD . . . . . 134 5. . . . . . . .
. . . .4 ANISOTROPIC DAMAGE MODEL FOR CONCRETE160 5. . .7. . .5 EVALUATION OF DAMAGE VARIABLE .6. . 159 5. .7 Opening and closing of the crack and initial damage . . . . . . . . . . . . . . . . 168 5. . . 174 5. . . .2 CONTENTS FUNDAMENTAL EQUATIONS OF DAMAGE MECHANICS . .8 ANALYTICAL PROCEDURES IN A FINITE ELEMENT MODEL . . . . . .7 CONSTITUTIVE MODEL FOR SMEARED FRACTURE ANALYSIS . 164 5.1 Prefracture behaviour . . . 173 5.7. 170 5.6. . . . . . . . . . .6 Damage evolution for concrete subjected to tensile strain166 5. . . . . . . . . . . . . .7. . . . . . .7. . . .3 Fracture energy conservation . .6 5. . . . . 173 5.6. . . . . . . . . . . .6 Fixed Crack Model With Variable Shear Resistance Factor (FCMVSRF) . . . . 170 5. . . . 158 5. .6. .5 Coaxial Rotating Crack Model (CRCM) . . . . . . . . . 175 . . 168 5. . . . . . . . . . . . . . .7. . .7. . . . . . .7. . .6. .7 Closing and reopening of cracks . . . . .4 Constitutive relationships during softening . . .6. . . . . . . . .6. . . . .2 Strain softening of concrete and the initiation criterion 170 5. . . . . . . . . . .3 ISOTROPIC DAMAGE MODEL FOR CONCRETE . . 172 5.
. . . . . . . . . . . . . . . . . .5 2. . . . . . . Displacement thickness in boundary layer . . . 100 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13 2. . 12 14 15 16 51 52 54 59 61 64 68 69 72 73 74 75 75 76 80 81 84 87 89 3. . . . . . . . . . a)Lagrangian viewpoint b)Eulerian viewpoint . . . . . . . . . . . . . . . 1. . . . . . . . . . . . . .10 2. . . . . . Change of relative positions in an arbitrary ﬂow ﬁeld . . . . . . . . . . . Shearing ﬂow between two ﬂat plates . .9 2. . . . . . . 2. . Boundaries of the damreservoir system . . . . . . . . . . . . . . . . Damreservoir interface . . . . . .3 2. . . . . . . . . . . . . . .4 Crosssection of a concrete buttress dam . . .11 2. . . . . Three types of ﬂuid motion . . .1 Systems of single degree of freedom . . . . .12 2. . . . . . . 1. . 1. .1 Idealized section of embankment dams a) Rockﬁll dam with symmetrical clay core b) Rock and gravel dam with reinforced concrete slab . . . . . . .4 2. . . . . . . . . Wellordered parallel ﬂow . . . . . . . . . . . . . . . . . . . . . . Rigid daminﬁnite reservoir system . . .List of Figures 1. Wave front movement and ﬂuid movement . . . . . . . . . . . . . . Added mass approach . . . . . . . . . . . . . . . . . . . . . . . . Moving pressure disturbance in a motionless ﬂuid and wave in a moving ﬂuid . . . . .2 Crosssections of several arch dams . . . . . . . . . . . . Free surface wave . . . . . . . . . . . . . . . .8 2. . . . . .2 2. . . . . . . . .19 Fluid point . . .6 2. . . . . . ﬁxed . . . . . . . . . . . . . . . . . . . . . Details of boundary layers . . Rectangular Parallelpiped element . . Fluid rotating like a rigid body . . .15 2. .17 2. . . . . . . . . . .3 Crosssection of typical concrete gravity dam . .18 2. . . . . . . . . .14 2.1 2. . . . . . . . . . . . . . . . . . Volume element . . . . . . . . . .7 2. . . . . . . . . . . . . . . .16 2. . . . . . . . . . Moving system .
. . . . . . b)fracture energy conservation. . a)softening initiation criterion. . . . . . . . . . . 137 5. . .9 Stressstrain curve for energy dissipation due to fracture . . .6 Closing and reopening of partially formed cracks . . . . (b) NLFM . .2 Fracture process zone (FPZ). . 101 . . . . . . . . . . . . . . . . . . . . . . . 152 5. . . . . . . . . . 165 5. . . . . . .d) crack band model . .8 LIST OF FIGURES 3. . . . . . . . . (b) mode II planar shear fracture. . . . . . . (b) characteristic dimension. .Tensile fracture. .tearing fracture . . . d)closing and reopening of cracks . . . . . . . . . . . . . 161 5. . . . .5 Nonlinear fracture mechanics in smeared crack propagation model . . F)initial damage formulation . . . . . . 143 5. . . (c) characteristic dimen√ sion. 3. .3 Nonlinear fracture mechanics models: (a. 145 5. . . . . . .3 An example of multidegreeoffreedom (MDF) sytem with degrees of freedom in y direction .4 (a) average stressstrain curve for smeared crack element. (a) LEFM. . 108 5. . . . 101 . . . . E)closingopening criterion. . D)stressstrain curve for equivaalence hypothesis. . . . . . . . . . . . . (c) mode III . . .1 Modes of failure: (a) mode I . . 171 . . 3. . . . . . . . B)characteristic length.10 Constitutive modelling for smeared fracture analysis. . . . . .(c. A)eﬀective areas for isotropic and anisotropic damages. . e)local axis system. . . .7 Strengthofmaterialbased failure criterion . .8 Material model in the damage mechanics concept. 139 5. . lc = l1 . . . 3.5 Interface element on the damreservoir interaction boundary . l2 . . . .4 An example of MDF system with two degrees of freedom at each mass . . . . 104 . . . . 147 5. C)strain equivalence hypothesis. .2 Forces on a single degree of freedom . . . . . . . . . . 141 5. . lc= l0 l00 . . . .b) ﬁctitious crack model. . . . . . . . . . . .
and improvement of navigation. Water supply dams collect water for domestic. The falling water spins the blades of the turbines connected to generators. recreation. The minimum ﬂow in dry months ﬁxes the amount of ﬁrm power which customers can rely upon to receive regularly. electric power. or height of stored water above the turbines. ﬁrst. Hydroelectric power dams are built to generate electric power by directing water in penstocks through turbines. upon the head of water. The higher the water the more weight and pressure bear upon the turbine blades. Many modern dams are multipurpose. ﬂood control. The earliest of these dams were used to water farms.Chapter 1 INTRODUCTION People from the beginning of recorded history have constructed barriers across rivers and other water courses to store or divert water. generated in ﬂood seasons can be sold. Sometimes extra power. waters pile up to form an artiﬁcial lake which sometime can be very long. The artiﬁcial lake backed up by a dam is called a reservoir. 9 . Behind the dam. water supply. Power dams are expected to generate power to repay the cost of construction. or runofstream power. usually at lower rates. Dams are built primarily for irrigation. Irrigation dams store water to equalize the water supply for crops throughout the year. Flood control dams impound ﬂoodwaters of rivers and release them under control to the river below the dam. For example. wheels with curved blades as spokes. The output depends. industrial. the ancient Egyptians built earth dams that raised the river level and diverted water into canals to irrigate ﬁelds above the river. and municipal uses for cities without suitable lakes or rivers nearby for a water supply. A second factor is the volume of water throughout the year.
design. . Diﬀerences depend on their geometric conﬁgurations and the material of which they are constructed. The world’s largest dams. and kinds and amount of local materials for construction. 1. and many of them are 180 m (600 ft) or more high. They slow down streams so that the water does not carry away soil. are embankment dams. they commonly require an equal amount of engineering skill in planning. and construction. Multiplearch and multipledome dams. Another problem of dams is silting.1 TYPES OF DAMS Dams range in size and complexity of construction from low earth embankment constructed to impound or divert water in small streams to massive earth or concrete dams built across major rivers to store water. containing materials of distinctly diﬀerent properties in various portions of dams. Under special circumstances. a. Their reservoirs provide recreation. Dams conserve soil by preventing erosion. INTRODUCTION Dams also provide beneﬁts other than those mentioned above. geology. Zoned dams. Dams may impair ﬁshing. such as ﬁshing and swimming. the amount of water available. feature of the basic types are combined within a particular dam to meet unusual design requirement. 2. topography. as measured by the volumes of materials used in their construction. The type of dam that is built and its size are a complex function of a demonstrated necessity for water storage or diversion. b. Concrete arch and dome dams. Dams can also create problems. Although large embankment dams do not posses the graceful and architecturally attractive conﬁgurations of many concrete dams. Their reservoirs may cover towns or historic and scenic places. b. thereby lessening its usefulness. constructed entirely from a more or less uniform natural material. They become refuges for ﬁsh and birds. Some rivers pick up clay and sand and deposit them behind the dam. a. The followings are the main types of dams: 1. There are several basic types of dams.10 CHAPTER 1. Embankment dams. Single arch and dome dams. many of the world’s highest dams are built of concrete. Homogenous dams. In contrast.
5. In the absence of such materials. and clay across a river. earth is carried to the dam by water in pipes or ﬂumes and also deposited by the water. in locations where natural impermeable materials are unavailable. In a rolledﬁll dam. 1. weathered bedrock. it may be obtained by separation from bowldery stream deposits. of concrete or other watertight materials is used. watertight materials form the core. 4. Larger embankment dams generally are zoned and constructed of a variety of materials. sand. Embankment dams are made by building an embankment of gravel. and the impermeable core is constructed of ordinary concrete or asphaltic concrete.1 Embankment Dams A broad spectrum of natural and fabricated materials have been used in the construction of embankment dams.1. glacial till. Where rock is used extensively. Concrete gravity and gravityarch dams. The placing of the earth is so controlled that the ﬁner. or steel. TYPES OF DAMS 11 3. or by quarrying.1. siderock accumulations. An important element in a zoned dam is an impermeable blanket or core which usually consists of clayey materials. In the semihydraulic ﬁll dam. earth is hauled by vehicles onto the dam and rolled tight with heavy machinery. obtained locally. A wall of rocks is then laid on the upstream side and over this is built a waterproof facing of reinforced concrete. or glacial till. or inner wall. Many small embankment dams are built entirely of a single type of material such as stream alluvium. Rock dams are made by dumping rocks across the river. Dams combining two or more basic characteristics of the above basic types.1. Concrete slab and buttress dams. or prepared by mechanical or hydraulic separation of a source material into fractions with diﬀerent properties. Construction of an embankment dam requires prior investigation of foundation geology and an inventory and soilmechanics study of materials available for emplacement in the embankment. Alternatively. often a core. Controlling factors in choosing this type of dam are the amounts and types of materials locally available for construction and the size and conﬁguration of the dam. embankment dams are built of rock or earthrock aggregates and impermeable layers . the dam is built of quarried rock or unsorted pebbly or bowldery deposits. either extracted from diﬀerent local sources. trucks bring the earth to the dam and jets of water distribute the materials. To prevent leakage. In the hydraulic ﬁll dam. timber.
Control of seepage through the dam or under it commonly requires installation of porous materials within or immediately beneath the dam. Minor settlement of an embankment dam owing to load stresses during and after construction generally is not a serious matter because of the ability of the embankment to adjust to small dislocations without failure. is that the bearingstrength requirements of the foundation are much less. A particular advantage of an embankment dam. or riveted sheet steel are placed on the upstream face of the dam.12 CHAPTER 1. Embankment dams have been built on a great variety of foundations. asphaltic concrete.1: Idealized section of embankment dams a) Rockﬁll dam with symmetrical clay core b) Rock and gravel dam with reinforced concrete slab of reinforced concrete. Crosssections of selected examples of embankment dams are shown in ﬁgure 1. . ranging from weak.1. INTRODUCTION Figure 1. as compared with a concrete dam. unconsolidated stream or glacial deposits to highstrength sedimentary rocks and crystalline igneous and metamorphic rocks.
It is clear that the foundation and abutments must be competent not only to support the dead weight of the dam on the foundation but also the forces that are directed into the abutments because of . TYPES OF DAMS 13 1. Constantradius arch dams commonly have a vertical upstream face with a constant radius of curvature. At sites where abutments are not entirely satisfactory rock may be excavated and replaced with concrete to form artiﬁcial abutments. curved structures commonly containing reinforcing. Simply stated. Variableradius dams have upstream and downstream curves (extrados and intrados curves) of systematically decreasing radii with depth below the crest When a dam is also doubly curved. that is. it is curved in both horizontal and vertical planes. Volume requirements for aggregate for manufacture of concrete are much less than in gravity and gravityarch dams. Curves that have been used in construction of arch or dome dams are arcs or sectors of circles. That portion of the bedrock abutment which receives the thrust from the load of reservoir water either by a tendency for downstream movement of the dam or ﬂattening of the arch is called the thrust block and must be suﬃciently strong to resist the forces acting on it without failure or appreciable. Some dams are constructed with two or several contiguous arches or domes. Engineering analysis of arch and dome dams assumes that two major types of deﬂections or dislocations aﬀect the dam and its abutments. tend to rotate the dam about its base by cantilever action. dislocation. an arch dam utilizes the strength of an arch to resist the loads placed upon it by the familiar ”arch action”.2 Concrete Arch and Dome Dams The ultimate complexity of design and analysis of stresses is attained in arch and dome dams. but the competency of bedrock in foundations and abutment to sustain or resist loads must be of a high order.1. or parabola. it is sometimes called a ”dome” dam. Pressure of water on the upstream face of the dam and. Their height can be as high as 272 m (905 ft).1. uplift pressures from seepage beneath the dam. so that stresses are created which act horizontally within the dam toward the abutments. Figure 1. ellipses. the pressure of reservoir water tends to ﬂatten the arch and push it downstream. Arch dams are of two kinds. in some instances. deep gorges in mountainous regions where access and availability of construction materials pose especially acute problems. Arch dams usually are built in narrow. and are then described as multiple arch or multipledome dams. These dams are thin. either steel rods or prestressed steel cables. In addition.1.2 shows Crosssection of several varieties of arch dams.
A typical gravity dam is shown in ﬁgure 1. the dam is freestanding. In regions of seismic activity consideration must also be given to the interaction of the dam and pulses of energy associated with earthquakes. Final selection of the site for a gravity or gravityarch dam is made only after comprehensive investigation of hydrologic.2: Crosssections of several arch dams arch action in response to loads created by impounded reservoir water. INTRODUCTION Figure 1. pressures exerted by ice forming on the reservoir surface. Gravity dams require maximum amounts of concrete for their construction as compared with other types of concrete dams. 1. and.14 CHAPTER 1. and. especially.1. They withstand the pressure or push of water by their weight. Properly constructed gravity dams with adequate foundation probably are among the safest of all dams and least susceptible to failure with time. that is. Concrete gravity dams have been constructed up to 285 m (950 ft)high. . In cross section. in areas of cold climates.3. they are like a triangle. with a ﬂat bottom. broad at the base and narrow at the crest. topographic.3 Concrete Gravity and GravityArch Dams A concrete gravity dam has a crosssection such that. the dam has a center of gravity low enough that the dam will not topple if unsupported at the abutments. and resist dislocation by the hydrostatic pressure of reservoir water by sheer weight. They are built in this shape because water pressure becomes greater with the depth of water.
3: Crosssection of typical concrete gravity dam subsurface geologic conditions. An important consideration in construction of a concrete gravity or gravityarch dam is the availability. . within a reasonable hauling distance.1. of adequate deposit of aggregate suitable for manufacture of concrete.1. whether the aggregate is obtained from unconsolidated deposits or is quarried. A favorable site usually is one in a constriction in a valley where sound bedrock is reasonably close to the surface both in the ﬂoor and abutments of the dam. TYPES OF DAMS 15 Figure 1. it may be possible to construct a gravityarch dam which incorporates the advantages of mass weight and low center of gravity of a gravity dam with those inherent in an arch dam. depending on the topographic conﬁguration of a valley and the foundation geology. in gravityarch dams the requirements for sound rock in abutments are somewhat more stringent than in simple gravity dams. The simplest form of a gravity dam is one in which the top or crest is straight.
They are up to 130 m (430 ft) high. buttress dams resemble gravity dams. In crosssection. canals. surge tanks and towers.4: Crosssection of a concrete buttress dam 1. gate hoist mechanisms and their supporting struc . penstocks.4 1. but with ﬂatter upstream slopes.1.4 Concrete Slab and Buttress Dams In locations where aggregate for concrete or for earthﬁll is in limited supply and the foundation rocks are moderately to highly competent. tunnels. buttress dams provide a possible alternative to other kinds of dams.2 APPURTENANT FEATURES OF DAMS Appurtenances are structures and equipment on a project site.16 CHAPTER 1. In a buttress dam a slab of reinforced concrete upstream rest on a succession of upright buttress which have thicknesses and a spacing suﬃcient to support the concrete slab and the load of the water in the reservoir that exert on the slab. powerhouse structures. other than the dam itself. but are not limited to. lowlevel outlets. They include. INTRODUCTION Figure 1. such facilities as intake towers. A crosssection of a typical buttress dam is shown in ﬁgure 1.
In valleys of steep proﬁle diversion commonly is accomplished by a tunnel or tunnels in the walls of the valley. In this section mention is made of various features that are incorporated into the designs of dams for control of ﬂow of water impounded in the reservoir through or outside of a dam. and subsurface geology of the dam site. the toe of a dam or at some distance downstream. In previous sections of this chapter the characteristics and conﬁgurations of basic types of dams have been outlined. or by conduits which subsequently are buried by the dam.1. Flow of water into the power house is controlled by . Following is a tabulation and description of several kinds of appurtenant features included in the construction of dams. Tunnels that transmit water under pressure to elevate the water to a higher level than the intake of the tunnel or to generate hydroelectric power are called pressure tunnel and usually require considerable competency in the rock through which they are constructed . topography. and all critical water control and release facilities. In many large dams the valves are installed in underground vaults or chambers to which access is gained downstream from the dam. Tunnels for control of the water level in the reservoir are commonly called gravity tunnel and serve a principal function in diverting water to some point downstream from the dam. Flow through them is controlled by valves external to the dam or in valve chambers or vaults within the dam or in bedrock outside of the dam. Coﬀer dams usually are temporary structures built upstream from a dam to divert stream ﬂow around the excavation for a dam. In valleys of low proﬁle diversion is by tunnels. but no consideration was given to the various appurtenant features that enable use of a dam and the reservoir behind it for their intended purposes. APPURTENANT FEATURES OF DAMS 17 tures.2. The powerhouse is located at. Valves control the ﬂow of water through tunnels and penstocks. Many dams are constructed to generate hydroelectric power. Commonly the diversion tunnels are put to further use to control ﬂow from reservoir either for drainage of the reservoir or for ﬂow under pressure into a hydroelectric generating plant. or in the vicinity of. Design of many of these features requires intensive prior investigations of hydrology. Also included are mechanical and electrical control and standby power supply equipment located in the powerhouse or in the remote control centres. It is not unusual in embankment dams to incorporate the coﬀer dam into the larger embankment structure comprising the designed dam. canals. Tunnels in bedrock outside of dams serve a variety of purposes.
and ﬂood control. tunnels. Pipes for conducting water to the power turbines are called penstocks. Levees are artiﬁcial riverbanks constructed high enough to prevent ﬂooding. or within the reservoir. Less . storage of industrial and municipal water supplies. Their size and location with respect to the dam is determined by the size and type of dam. Increasing use is being made of tidalstorage reservoir for power generation along coast lines and excavated reservoirs for water storage for municipal or other use. may be called a barrage. or near the dam in chutes. sluices. recreation. Overﬂow of embankment dams outside of a spillway can have especially disastrous consequences so that safety usually requires a spillway capable of containing at least a hundredyear ﬂood. Gates are devices installed in the tops of spillways to control the ﬂow of water over the spillway. and penstocks is regulated by control gates. The ﬂow of water from intake towers through spillways. within the dam downstream. Spillways are located within or on the downstream face of a dam outside of the dam on one side or the other. where water spills into a ”glory hole” and passes through a shaft and tunnel or tunnels in the abutment of the dam. The common type is a reservoir behind a dam in a valley. Reservoirs and their associated dams serve many purposes including electric power generation. A small dam that forms a millpond or ﬁshpond is called a weir. designed to accommodate ﬂows during maximum ﬂood stage so as to prevent damage to the dam and appurtenant features. INTRODUCTION valves upstream from the dam. A sluice is a passage through the dam itself for lowering the water level of the reservoir. Spillways are. 1. or shafts.3 SAFETY OF DAMS AND RESERVOIRS There are Several types of reservoirs as deﬁned by their locations. Water may pass over the crest of the dam itself. storage and diversion of irrigation water. once a certain depth of water has been reached. local topography geology and a careful review of the history of stream ﬂow at the site of the dam. A Spillway is designed to contain and control overﬂow of reservoir water when the reservoir is full. In the latter type the excavated material commonly is employed to construct an embankment on one or all sides of the reservoir.18 CHAPTER 1. A dam across a river intended to permit ﬂow. or in valve vaults excavated in rock outside of the dam. or should be.
Filling of the reservoir impose additional loads not only on the ﬂoor and wall of the reservoir but also on the upstream face of the dam. and reservoirs are constructed for this purpose. But the length of the reservoir and its capacity may change the risk associated in the case of failure. probably in the thousands. foundation subsidence. Failure of concrete dams usually is attribute to imperfect design or construction.3. dam built long ago and dam built in recent times present a growing potential for massive destruction of life and property. and the conﬁguration of the valley below it. and highways. In some stances fuel powered and thermonuclear power plants require large volumes of cooling water. and storage of sewage and waste products from mining or manufacturing operations. and property damage of a ﬂood generated by breaching of a dam or collapse of reservoir wall depends to a large extent on the magnitude of ﬂood and what lies in the pathway of the ﬂood. ruins of dams built through a long span of history in both ancient and more recent times indicates that the number must be high. Construction of a dam and reservoir imposes new loads on foundation materials. or earthquake damage. Initial adjustment in the dam and foundation occurs as the dam is being built and as the load on the foundation is increased to a ﬁnal load equal to the weight of the dam. As the reservoir level is decreased and increased these loads ﬂuctuate . However. SAFETY OF DAMS AND RESERVOIRS 19 frequent uses of reservoirs include storage and control of stream water for navigation. the cause of ﬂoods associated with failure of dams and reservoir are numerous. Pressure depends not upon how far water is backed upstream but upon its depth at the dam. goods. to use of inferior materials in the dam. Whether a dam backs up water for a long or a short distance is not important from the structural point of view. or to failure of foundation and/or abutment rocks. The magnitude of ﬂoods generated by dam failure or by collapse of the walls of a reservoir are not related to the height of a dam. and a cycle dynamic system of changing loads is superimposed on the static . With an expanding population in many part of the world and an increasing occupation of ﬂoodplains by dwellings. if dams of all heights are included in the count.1. commercials and industrial facilities . although some failures have been attributed to slope failure. wether the dam is low or high. There are no accurate records of the number of dams that have failed throughout the history of their construction. In embankment dams a common cause of breaching is overtopping of existing spillways or waterlevel control facilities. The cost in human life. More pertinent is the volume of water stored in the reservoir behind the dam.
and intensiﬁcation of eﬀorts to assure the safety of existing dams and reservoirs and those that will be in the future is an increasingly urgent necessity. The engineers constructed a concrete archgravity dam at an approved cost of $174.680 kilograms). however imperfect they may be.000 workers. Once the location was chosen. They made maps of 70 locations. and silting. Where once had been burning desert. built between 1930 and 1935. In spite of these good works. Construction crews also built . 1. which travel on tracks along opposite sides of the site. designers made their plans. It stores the entire ﬂow of the Colorado River for two years.20 CHAPTER 1. INTRODUCTION load of the dam on its foundation. dams continue to fail. The responsibility for the safety of dams and reservoirs no longer belongs only to the designer and builder but must be shared by those who have knowledge and understanding. high water level. bored holes to test the rock for a sound foundation. Seepage of water through or beneath the dam may produce slow deterioration that may promote eventual failure. engineers built Boulder City to house about 5. The engineers made geologic and topographic surveys to select the site. Each of the ﬁve cableways could carry 25 tons (22. Construction gangs built railroads and highways for transporting great quantities of equipment and materials.000. Modern technology and social responsibility require that safe construction and maintenance of dams and reservoir shall be the shared responsibility of engineers.4 HOW DAMS ARE BUILT The methods of building dams can be envisioned by following the construction of Hoover Dam. Workmen strung cables across the canyon from pairs of towers. Recognition of the need for worldwide surveillance of dams and reservoirs with emphasis on their safety has resulted in the formation of the International Commission on Large Dams (lCOLD) a unit of the World Power Conference within individual countries increasing eﬀorts are being made to regulate and maintain continued safety of dams and reservoirs through the close cooperation of engineering organizations and government agencies. 660 feet (201 meters). Two of them had spans of nearly half a mile. Much preliminary work had to be done. and studied the river’s speed.000. geologists. and rock mechanical expert. They then made models to test their design. Its crest is 45 feet (14 meters) thick and its base. It is as tall as a 60story skyscraper. of the expected behavior of natural materials under the condition superimposed by the loads of dams and reservoirs.
or loose rock and muck. and Aswan High. One of the oldest great dams is Egypt’s Aswan for irrigation on the Nile.000 feet (300 meters) each. in Uganda. were drilled and blasted from the rock of the canyon walls. SayanoShushensk. was one of the earliest major Soviet dams. completed in 1932. Three of the foreign dams that are constructed of the largest volume of materials are Tarbela. was next forced into the foundation to ﬁll seams and holes. Grout. Concrete was poured into the forms from eightcubicyard buckets traveling on the cableways.5 FAMOUS DAMS OF THE WORLD Rogun and Nurek. and Guri. Another problem was the freezing of control gates in winter. Forms were made. As each block of concrete dried. in Switzerland. in Nigeria. making the dam into one solid piece. in Tajikistan. in Canada. In addition the concrete would have shrunk and cracked. in Venezuela. Dams that form the largest reservoirs in capacity are Owen Falls (Lake Victoria). grout was pumped between the blocks. coﬀerdams of earth and rock were built upstream and downstream from the dam site to block the river. in Georgia. in Russia. ”High scalers” stripped tons of loose and projecting rock from canyon walls. Mica. Bratsk. When construction was completed the tunnels served as spillway outlets and penstocks for the power plant. on the Cauvery River . bordering Zambia and Zimbabwe. was dug out to expose the bedrock. Cold water circulating through 528 miles (850 kilometers) of oneinch (25mm) pipes embedded in the concrete carried oﬀ the heat. Some of the world’s other major dams are Grande Dixence and Mauvoisin. They rise to a height of more than 1. Lower Usuma. Refrigerating pipes were also used to freeze landslides of wet earth at Grand Coulee Dam. These tunnels were used to divert the river around the site. Kariba Gorge. Inguri. in Russia. each 50 feet (15 meters) in diameter. Next. Another historical dam is Mettur. 1. in Pakistan. A huge electrical heating apparatus was installed in the spillway and gates.1. in Colombia. Chicoasen. are among the world’s highest dams. a thin mortar of cement and water. The overburden. FAMOUS DAMS OF THE WORLD 21 a great gravel screening plant and two huge concrete mixing plants. and Guavio.5. Allowing so gigantic a structure to cool naturally would have taken a century because of the heat given oﬀ by the setting cement. The Dnepr. These would be used for building the dam in enormous blocks. ﬁnished in 1902. On each side of the river two tunnels. in Egypt. in Mexico.
22 CHAPTER 1. Mmaterial in dam (cu yd). RR—River regulation. UC—Under construction. M—Mining. E—Rolled earth ﬁll. RCReservoir Capacity (billions of gal) A—Arch. C—Concrete. G—Gravity. The world’s ﬁrst dam to furnish hydroelectric power from ocean tides was completed in 1966. RP—Recreation purposes. N—Navigation. HHeight (ft). FC—Flood control. B—Slab and buttress. near StMalo. It is located on the Rance River. I—Irrigation. P—Power. The following tables show some of the famous dams of the world with their speciﬁcations. S—Stone masonry. H—Hydraulic earth ﬁll. R—Rock ﬁll. INTRODUCTION in India. WS—Water supply. in France. . In the tables the followings notation was used for simplicity. LLength (ft). M—Multiple arch.
1 8 8 .0 0 0 1 1 . FAMOUS DAMS OF THE WORLD T able 1.7 4 5 106 261 9 3 5 .0 0 0 9 7 1 .9 0 0 .7 0 0 1 .9 9 5 .5 0 0 .0 0 0 1 .2 2 3 .5.0 0 0 .1 2 6 .3 9 0 2 .0 0 0 4 6 0 .1 5 3 734 2 .0 0 0 2 .0 0 0 1 .0 0 0 1 .0 9 1 2 .5 1 7 8 .9 7 0 .0 0 0 5 .1 4 8 3 7 .6 0 7 9 .2 4 6 879 3 .0 0 0 2 0 .1 2 5 8 .5 1 3 624 1 .1 8 2 882 700 53 766 557 782 .0 0 0 8 6 1 .4 7 3 23 8 .1 0 9 1 .0 0 0 4 2 .8 6 0 2 .0 0 0 23 R 3 .0 0 0 1 5 .0 5 4 1 .3 1 1 1 .0 0 0 4 .9 5 0 .4 0 0 .7 0 0 .0 0 0 7 .0 0 0 1 5 .8 8 1 696 1 .2 7 1 269 6 .6 0 0 3 .5 6 8 919 1 .0 0 0 2 .0 0 0 2 .9 2 0 1 .2 4 7 2 .0 0 0 7 8 .7 9 2 .6 0 0 M 8 1 .2 6 1 439 215 48 1 .0 0 0 2 .0 0 0 4 .7 3 9 .7 9 8 5 .0 0 0 2 3 .2 0 0 1 .0 4 0 935 892 858 856 807 794 794 787 778 777 770 764 742 726 722 722 717 710 705 703 682 679 666 662 656 656 637 625 L 2 .0 0 0 4 .7 0 6 6 .0 0 0 1 4 .0 6 6 1 .0 0 0 6 .2 8 0 2 .1.6 5 5 .0 0 0 4 .7 3 8 3 .0 0 8 .9 0 1 .1 Highest dams of the world Nam e R ogun N u re k G ra n d e D ix en ce In g u ri Va io nt Teh ri G u av io M ic a S aya n oS hu sh en sh aya C h ico a se n C h ivo r M a u vo isin O rov ille C h irke y skaya B h a k ra H o ove r C o ntra M ra tin je D w o rsh a k G le n ca nyo n Tok to g u l D a n iel J o h n so n L u zz o n e K eb a n D ez A lm e n d ra K o lnb re in K a ru n N ew B u llard ’s B a r N ew M elon es L o ca tio n Ta jik ista n Tu rk m en ista n S w itze rla n d R u ssia Ita ly In d ia C olom b ia C an a d a R u ssia M ex ic o C olom b ia S w itze rla n d C alifo rn ia R u ssia In d ia USA S w itze rla n d Yu g o slav ia Id a h o USA K y rg y z sta n C an a d a S w itze rla n d Tu rkey Ira n S p ain A u stria Ira n USA USA Typ e E E C G A A E R E R E R A E R E R C A E A C G C A G A A G C A G A C M C A E R G C A A G A A A E R P u rp o se P I P I P P IFC P FC P N P P P P IFC W S P IFC W S P I P IFC R R P P P FC R P P P I P P P P I P P P I P IFC R P P IFC R P Ye a r 1987 1980 1962 1980 1961 UC 1989 1974 1980 1981 1977 1957 1968 1975 1963 1936 1965 1975 1972 1964 1977 1968 1963 1974 1963 1970 1978 1975 1970 1975 H 1 .5 0 4 1 .0 0 0 1 .0 0 0 7 5 .2 8 7 1 .4 7 6 4 .5 0 6 2 .2 4 4 1 .5 6 0 1 .5 7 0 .4 0 0 .2 8 0 2 .6 9 6 23 232 1 .1 8 6 .0 0 0 6 4 7 .6 0 2 .8 6 4 .9 6 7 .7 0 0 .0 9 6 .9 1 6 .
4 9 9 .6 1 7 6 .7 0 0 1 3 .6 4 2 1 8 .0 0 0 8 5 .2 0 5 .5 0 0 2 .2 L arg est dams of the world Nam e N ew C o rn e lia Ta ilin g s Ta rb e laN a m e Fo rt Peck L ow er U su m a A ta tu rk G u ri Ya c y reta A p ip e O ahe M a n g la G ard in er A fslu itd ijk R o gu n O rov ille S a n L u is N u re k G arriso n O o ste rsch eld e C o ch iti Ta b ka K ie v A sw a n H ig h B en n ett Tu cu ru i M issio n Tra ilin g # 2 Fo rt R an d a ll K anev Itu m b ia ra L o ca tio n U SA P a k ista n U SA N ig eria Tu rkey Ven ez u ela A rge ntin a Pa ra g u ay U SA P a k ista n C anada T h e N eth e rla n d Ta jik ista n U SA U SA Tu rk m e n ista n U SA T h e N eth e rla n d U SA S y ria U k ra in e Egypt C anada B ra zil U SA U SA U k ra in e B ra zil Typ e E E R H E E R E R G E G E E E E E E E E E E G E E E G E R E E G E E E E G P u rp o se M P I P IFC N P P IW S R P P IFC N P I P IW S FC W S P I P IFC W S P I P I P IFC N FC W S FC R P P IFC P N P IR P P P N W S P IFC N P Yea r 1973 1976 1940 1990 1990 1986 UC 1963 1967 1968 1932 1987 1968 1967 1980 1956 1986 1957 1975 1964 1970 1967 1984 1973 1956 1974 1980 H 98 470 250 528 604 531 108 245 380 223 62 1 .1 6 8 .6 6 6 .0 0 8 .1 5 3 664 2 . INTRODUCTION T able 1.5 0 6 .0 0 0 1 2 5 .9 8 1 M 2 7 4 .0 2 6 1 4 .0 0 0 9 .0 6 6 770 382 1 .0 0 0 5 6 .6 7 8 2 .0 0 0 5 2 .0 6 3 .0 0 0 1 1 0 .9 2 0 1 8 .0 0 8 .3 9 0 1 1 .4 4 8 1 2 .3 0 0 2 9 .0 0 0 R.9 2 5 167 3 .8 9 1 1 4 .3 0 0 1 1 .5 0 0 9 .6 0 0 2 .5 3 1 5 .0 8 8 .0 0 0 1 0 1 .0 9 6 .2 0 3 .0 0 0 7 7 .0 0 0 6 5 .0 0 0 6 4 .8 5 3 1 6 4 .5 6 5 6 .0 9 1 1 .5 8 5 3 .2 0 3 .0 0 0 8 2 .0 0 0 5 7 .3 9 7 .6 3 1 .4 6 4 7 .2 3 4 26 3 6 .4 3 5 5 0 .7 7 9 1 0 .0 0 0 5 7 .0 0 0 6 0 .0 4 0 203 164 253 197 72 364 600 282 128 165 82 328 L 3 6 .7 0 0 1 0 .0 0 0 6 6 .9 7 1 3 0 .5 7 5 8 .6 0 7 1 .0 0 0 4 7 .9 8 2 15 1 .7 4 5 7 .0 0 0 7 5 .8 5 8 692 4 .5 2 2 .0 0 0 5 7 .6 4 0 .5 2 0 .0 0 0 8 5 .6 8 7 1 .7 4 3 .6 1 2 .5 0 6 6 .8 6 4 .0 0 0 2 1 .9 5 0 2 1 .5 2 8 2 6 .0 0 0 9 2 .7 0 0 5 2 .0 0 0 4 9 .8 7 2 .0 0 0 1 6 .0 0 0 9 5 .0 2 6 .0 0 0 8 1 .8 1 9 .6 9 8 984 4 4 .2 0 0 .24 CHAPTER 1.0 0 0 7 8 .5 5 2 .0 0 0 1 2 1 .7 2 0 4 .7 6 4 1 7 7 . 6 3 .0 0 0 1 5 9 .2 4 4 .9 2 7 .
6 8 5 . POWER GENERATOR. Flowing water rotated the waterwheel. it can do useful work for man. Today.3 5 0 .0 0 0 1 0 . for only a fraction of the total energy used for mechanical power.7 1 3 4 4 . It accounts. FLOOD CONTROL AND IRRIGATION DAMS25 T able 1.0 9 1 4 4 .7 0 0 2 .0 6 9 1.0 0 0 3 . He ﬁrst made water work for him with the waterwheel—a wheel with paddles around its rim. For example.3 6 4 1 8 . The roar of a waterfall suggests the power of water. Coal. heat.8 5 3 3 . or nuclear energy.8 6 4 1 2 . Electric energy may also be obtained from waterpower.0 0 0 1 0 1 .2 0 3 .6 4 2 4 2 .5 7 5 1 8 . while other energy sources are destroyed when they are used.1 0 0 4 .0 0 0 5 .1. Electricity from waterturned generators is called hydroelectric power. Since ancient times man has put to work the energy in the ﬂow of water.4 0 0 . and natural gas supply most of the rest.1 3 9 . new kinds of waterwheels spin generators that produce electricity.8 1 9 .7 2 0 1 9 .7 2 5 1 6 . and refrigeration. however.2 0 3 .3 6 1 3 9 .6.1 0 2 3 7 .2 8 3 . petroleum. Waterpower produces about 8 percent of the electricity used in the United States. When the power of water is harnessed. oil.0 0 0 2 . Among water’s virtues as a source of power is the fact that it can be used again after it supplies energy.6.0 2 5 2 .5 6 5 2 . Water passes downward through a hydraulic .9 5 0 . which in turn ran machinery that was linked to it. Rampaging ﬂoodwaters can uproot strong trees and twist railroad tracks.3 4 3 M 1 8 . light.0 0 0 R 5 4 .1 POWER GENERATOR. gas. a dam built to provide a head for water turbines usually creates a reservoir that can supply water for irrigation and drinking.0 0 0 5 7 . FLOOD CONTROL AND IRRIGATION DAMS Power Generator Dams Most electric power is generated in large plants that use coal.0 0 0 5 7 .3 1 1 3 0 .4 9 3 6 .6 1. however.0 0 0 1 .4 7 3 3 6 .3 Dams with greatest reservoir Nam e O w e n Fa lls B ra tsk A sw a n H ig h K a rib a G o rg e A ko so m b o D a n iel J o h n so n G u ri K ra sn oya rsk Bennet Z e y skaya L o ca tio n Uganda R u ssia Egypt Z a m b ia Z im b a bw e G hana C a n ad a Ve n ez u e la R u ssia C a n ad a R u ssia Type C G E C G E R C A R C M E —R G C G E B P u rp o se P I P N W S P IR P P P P P P FC N P P IFC Yea r 1954 1964 1970 1959 1965 1968 1986 1972 1967 1975 H 100 410 364 420 463 703 531 407 600 369 L 2 . Hydroelectric installations also supply added beneﬁts to a region.
are fertile farmlands. some generating only a few hundred kilowatts. There are many small plants on rivers. levees. Each cubic meter of water weighs 1000 kilograms. While the communities were small.000. 1.26 CHAPTER 1.770 MW. With the great population and industrial growth of the cities. reservoirs. the damages suﬀered from ﬂoods were limited. therefore. and power to run their mills and factories.2 Flood Control Dams One way to avoid ﬂoods is to take the obvious precaution of living where there is no danger of high waters. . One watt is the force it takes to raise one kilogram one meter in one second. The power produced by water depends upon the water’s weight and head (height of fall). Then the product is multiplied by 980. and large portions of them were built on land that was subject to periodic ﬂooding. Large plants. American pioneer settlers depended upon the streams for drinking water. INTRODUCTION turbine that is connected to a generator. transportation. or 980. were established along the riverfronts. It has always been convenient and often necessary to build homes and factories on the ﬂoodplains along rivers and streams and on the seacoasts. The power potential of a waterfall is found by multiplying its ﬂow. deep with the silt laid down by overﬂowing rivers. Floodplains. It is part of the Columbia River Basin project. Hoover Dam has an installed capacity of 1.000 MW. The earliest towns and farms. which has a planned capacity of 9. and other engineering works. For example.244 MW. measured in meter. It would press upon each square meter of turbine blade with a force of 1000 (1 x 1000) kilograms.6.025 MW.81(gravitational acceleration). a column of water that is one meter square and one meter high would contain one cubic meter of water. a 100meter high waterfall with a ﬂow of 10 cubic meter per second would develop 100 x 10 x 980. measured in cubic meter per second. by its height. For example. which depend on large volumes of water dammed upstream. Engineers measure waterpower in terms of watts (kilowatts or megawatts). watts. Grand Coulee has an installed capacity of 2. can generate more than 2. Among the largest plants in the United States are Hoover Dam and the Grand Coulee Dam. ﬂood damage has become a serious national problem. One of the basic approaches to ﬂood control is to minimize the extent of ﬂooding by building dams. The output of a hydroelectric plant is usually measured in kilowatts or megawatts of electricity. which is 1000 (the number of kilogram in a cubic meter of water) multiplied by 9.
FLOOD CONTROL AND IRRIGATION DAMS27 Engineers in ancient times built earthen mounds to keep back ﬂoodwater. Some lakes. and marshes that were once drained to make farm lands are being restored to preserve the level of underground water in time of drought. Even if the reservoir is nearly full. navigation. Dams and the reservoirs behind them help control ﬂoods. Thus the prevention and control of ﬂoods are tied with drought measures as well as with waterpower.1. During excessive rains. Furthermore. water collects in the storage reservoirs and is released in controlled amounts to the channel below the dams. and land use. it acts like a safety valve. Thus uniform height and strength are required. held Chinese rivers in check for centuries. Moreover. By means of terracing. which muddies the rivers and is ultimately lost in the ocean. This method was followed in colonial America. Thus less water ﬂows on to swell ﬂoods downstream. contour plowing. Thus ﬂood prevention and erosion control go hand in hand. Because a levee at one point conﬁnes the water there and raises the peak of ﬂood waters upstream and downstream. Floodcontrol dams are built to create big storage capacity and are planned for rapid ﬁlling and emptying. The water carries with it large quantities of the richest topsoil. and silt from above slowly rebuilds the eroded spots. storage space is obtained in which the ﬂood waters can be impounded for gradual release later. The amount of water which would add 3 meter to the height of a river 30 meter wide would add only one meter to a reservoir or lake 90 meter wide. Only a government which controls the river from end to end can safely supervise levee building. called levees. runoﬀ and erosion are checked on gentler slopes. Floodways and spillways divert excess water from the main river channel and carry it oﬀ by a diﬀerent route. carry oﬀ an estimated 280 million cubic meters of solid matter each year. swamps. and a wise choice of plants. Agricultural experts propose to return the steepest hills along the headwaters of American rivers to forest. soil conservation. for example. runoﬀ is slowed. POWER GENERATOR. This program also reduces ﬂoods by increasing evaporation and by the safetyvalve action of wide lakes or swamps on narrow rivers. By emptying a dam before a ﬂood is expected. .6. American rivers. Such artiﬁcial embankments. levees once started usually have to be built at all the low points of a river system. a system of levees is only as strong as its weakest spot. By damming gullies. Steps to lessen the eﬀects of drought also aid in ﬂood control. evaporation from the broad surface of a reservoir or lake is far greater than evaporation from the narrow surface of a river. Preventing soil erosion also aids ﬂood control by slowing the rate at which silt ﬁlls the reservoirs behind ﬂoodcontrol dams.
ﬂow velocity and water levels. 1.3 Irrigation Dams Irrigation is the artiﬁcial supply of water to agricultural land. settlement. or reservoir. uplift. To ensure continuous supply. municipal and industrial water supply. hydroelectric power generation. Generally the need for irrigation water is highest in the dry season when river ﬂows are lowest. Some reservoirs are capable of storing billions of gallons of water.6.7. Irrigation projects must also allow for removal of excess water. stress. seepage. Compilation of results on a computerized data base is done and results are interpreted and reported in the form of recommendations outlining appropriate action need to be taken. Instruments can be used for verifying design veriﬁcation. and recreation. Large dams and the associated reservoirs are often built for multipurpose use—irrigation.28 CHAPTER 1. an artiﬁcial lake. By erecting a dam. piezometric pressures. when it is necessary.1 INSTRUMENTATIONS AND SURVEILLANCE OF DAMS Surveillance External and internal examination of dams are required during their life time. strain.7 1. Long term surveillance program can ensure dams owner of their safe operation. ﬂood control. and alarm systems as required for a particular impoundment. is created from which water can be released as required.7. INTRODUCTION 1. water must be stored on a seasonal and sometimes annual basis.2 Instrumentation Instruments are used to characterize dam site condition as well as dam structure conditions. 1. It is practiced by more than half the farmers in the world because they need more water for their crops than is available from rainfall. Design of monitoring installations is to measure displacement. . Systems of dams and reservoirs along a river and its tributaries are developed with the purpose of providing comprehensive and integrated water resources management in an entire basin. Long term surveillance requires instrument installation and monitoring to assess the performance of structures with respect to design parameters.
The economical designer will specify minimum performance requirements. and presentation of data. Performance Instruments are used to monitor the inservice performance of a structure. Safety Instruments can provide early warning of impending failures. What information is required for the initial design? What information is required for evaluating performance during and after construction? When the parameters are identiﬁed. Safety monitoring requires quick retrieval. and precision. The designer typically speciﬁes the highest values required. Instrument performance is speciﬁed by range. resolution. accuracy. Range is deﬁned by the highest and lowest readings the instrument is expected to produce. it may be suﬃcient to monitor only one parameters but when the problem is more complex. Sometimes the term accuracy is mistakenly substituted for resolution. The Choice of Instruments depends on critical parameters of each project. The designer must identify those parameters and then select instruments to measure them. Another beneﬁt of selecting instruments to monitor complementary parameters is that at least some data will always be available. Design Veriﬁcation Instruments are used to verify design assumptions and to check that performance is as predicted. and deformation can provide an indication of the performance of a dam. performance assessment. Resolution is the smallest change that can be displayed on a readout device. For example. INSTRUMENTATIONS AND SURVEILLANCE OF DAMS 29 safety assessment.7. processing. Thus it is common practice to choose instruments that provide complementary measurements. Resolution typically decreases as range increases. Instrument data from the initial phase of a project may reveal the need (or the opportunity) to modify the design in later phases. allowing time for safe evacuation of the area and time to implement remedial action. and precision. Resolution is usually many times better than accuracy and is never expressed . pore water pressure. For example. even if one instrument fails. so that decisions can be made promptly. inclinometer data indicating increased rate of movement may be correlated with piezometer data that shows increased pore pressures. since the cost of an instrument increases with resolution. monitoring parameters such as leakage. and precision of measurements. accuracy. resolution.1. it is useful to measure a number of parameters and to look for correlation between the measurements. the speciﬁcation for instruments should include the required range. In some cases It may be necessary to have complementary parameters and redundant measurements. In some cases.
(3) there are too many sensors for timely manual readings. the original baseline data is no longer useful.30 CHAPTER 1. It is expensive and sometimes impossible to replace a failed instrument. or ±1% of full scale. but may exhibit excessive drift over the long term. (2) sensors are located at a remote site or in a location that prevents easy access. Accuracy is the degree to which readings match an absolute value. Some instruments used in monitoring of dams are brieﬂy described. the value returned by the instrument is slightly diﬀerent. Do technicians have the skills required to install and read a particular type of instrument? Are adequate support facilities available for maintenance and calibration of the instrument? An automatic data acquisition system may be required when: (1) there is a need for realtime monitoring and automatic alarms. Precision or repeatability is often more important than accuracy. Instruments such as hydraulic piezometers and liquid settlement gauges have limited use in freezing weather. . Consider the personnel and resources at the site when choosing instruments. the choice of instruments should be narrowed to those that can be connected to the system easily and inexpensively. such as ±1mm. or (4) qualiﬁed technicians are not available. The diﬀerence in cost between a highquality instrument and a lesserquality instrument is generally insigniﬁcant when compared to the total cost of installing and monitoring an instrument. less reliable instrument. In tropical heat and humidity. 1. INTRODUCTION as a ± value. since what is usually of interest is a change rather than an absolute value. For example. Accuracy is expressed as a ± value. Temperature and humidity also aﬀect instrument choice. Some instruments are excellent for shortterm applications.3 Instruments In the following section. ±1% of reading.7. It is false economy to install a cheaper. Every time a reading is repeated. Precision is expressed as a ± value representing how close repeated readings approach a mean reading. simple mechanical devices may be more reliable than electrical instruments. the cost of drilling and backﬁlling a borehole is typically 10 to 20 times greater than the cost of the piezometer that goes in it. If a data acquisition system is required. Even when it is possible to replace the instrument.
Design and build for uplift pressures and buoyancy. allowing time for corrective action to be taken or. Piezometer measurements help engineers to Monitor water levels. a direct measure of stability. typically 1 to 3 meters.1. Inclinometers monitor movement. can provide early warning of catastrophic failure. Predict slope stability. particularly inplace inclinometers that are monitored continuously. but the resulting data cannot provide absolute displacement and settlement proﬁles. and Corrective Measures rate of movement. beam sensors can be linked endto. Monitor seepage and verify models of ﬂow. They Provide early warning of threatening deformations. Installed at the proposed site for a dam. Inclinometers may be installed to check that actual movements of a structure correlate to those predicted during the design phase. so changes in tilt can be converted simply and accurately to millimeters of movement (settlement. INSTRUMENTATIONS AND SURVEILLANCE OF DAMS 31 Piezometers measure porewater pressure and ground water levels.end to monitor diﬀerential movements and provide absolute displacement and settlement proﬁles. This information helps engineers determine the need for corrective measures. In general. Beam sensors are used for monitoring settlement. an inclinometer might detect movement at a subsurface shear plane. Tiltmeter is used for monitoring changes in inclination. Two version of beam sensors are horizontal version for monitoring settlement and heave and vertical version for monitoring lateral displacement and convergence. The shear plane could cause problems later when the reservoir behind the dam is ﬁlled and porewater pressure along the shear plane increases. if necessary. or lateral displacement). however.7. Horizontal inclinometers are used to monitor settlement in foundations and embankments. that of monitoring rotation. the beam sensor has a deﬁned gauge length. Inclinometers. Design and build for lateral earth pressures. both types of sensors can be used to evaluate the performance of dams under load. heave. Tiltmeters typically have a more limited function. so they are often used in site investigations. direction. lateral deformation. Beam sensors diﬀer from tiltmeters in two important respects: First. Second. Inclinometers are installed for long term monitoring to detect LongTerm Performance changes in ground conditions or in the structure itself. heave. It is used for monitoring the rotation of concrete dams. Inclinometers are installed to monitor the magnitude. or convergence. for safe evacuation of the area. . convergence. A tiltmeter can be used with a large number of tilt plates to detect diﬀerential movement in a structure.
Typical application include monitoring movement at submerged construction joints in concrete dams and monitoring joints or cracks in tunnels and tanks. Surface Extensometers (Tape Extensometer) is used to determine changes in the distance between reference points anchored in walls or structures of an excavation. Load cells are used to prooftest and measure loads in tie. In general. abutments. . Strain gauge is used for monitoring strain in steel structural members. They can be read from a central location and are particularly useful where access is diﬃcult.32 CHAPTER 1.backs. they are used to verify design assumptions and to warn of soil pressures in excess of those a structure is designed to withstand. heave. and pipes. rock bolts. retaining walls. Pneumatic settlement cells provide a singlepoint measurement of settlement. and embankments. magnitude and directions of total stresses. Data from the extensometer indicate the depths at which settlement has occurred as well as total amount of settlement. Total Pressure Cells measure the combined pressure of eﬀective stress and porewater pressure. and lateral deformation in soil and rock. against piles. INTRODUCTION Borehole Extensometers are used to monitor settlement. Soil strainmeters are used for monitoring horizontal strain in embankments and monitoring tension cracks in earth structures. and direction of total stresses within the clay core. Surfacemount jointmeters are used to monitor movement at joints and cracks in concrete structures under harsh environments and submersion. In Embankment Dam Total pressure cells used to conﬁrm design assumptions and An array of cells provides data to determine distribution. Embeded jointmeters are used for monitoring movement at joints in mass concrete structures. Total pressure cells are installed within ﬁlls to determine the distribution . and struts. convergence. precision measure of convergence. They are used to monitor consolidation in the foundation during construction and monitor longterm settlement in the foundation and ﬁll. reinforced and mass concrete. size. ground anchors. foundations. They are also used for monitoring the performance of anchor systems. It is used in mass concrete structures such as dams. foundations to monitor movement at joints. Typical applications include monitoring settlement or heave in excavations. They can also be installed with one surface against a structure to measure total stresses acting on upstream face of the dam.
peaking versus base load generation. its purpose (e. Generally changes in the biological community are driven by changes in the physical or chemical (physicochemical) dimensions of the habitat (niche). the evidence for external impacts is substantially weaker than that for ac.g. Finally.1. morphometry. its size. genetic diversity. concrete or earth). hydrologic regulation. productivity and feedback.(b)within the water column of the reservoir/impoundment. Here. The Physicochemical parameters consists of temperature. Whether any particular concern applies to any particular dam is a function of a suite of factors that might include the dam’s type (e. The physicochemical factors are: temperature. internal seiches and altered currents.g. species richness. tailings impoundment). The observations are derived from an ecocentric rather than any anthropocentric ethical premise. particulate loading. timing of drawdowns. oxygen content. the Physicochemical eﬀects (physical or chemical factors) have separated from biological eﬀects.g. hydroelectric generation. The community structures are longer migration times. (c) at the sediment/water interface (which may or may not feed into the water column) and (d) a category for impacts that act outside the immediate dam/reservoir system. etc. transparency. ﬂow rates/shear and channel simpliﬁcation. transparency. Downstream impacts could include changes in Physicochemical factors and Community structure. epilimnetic versus hypolimnetic draw. location and operating protocol (e. The community structures are: environmental conditions exceed tolerance limits of original community members. oxygen content.8. Impacts are also organized spatially as those that occur: (a) downstream of the dam. discharge patterns. environmental conditions exceed tolerance limits of original . ECOLOGICAL/ENVIRONMENTAL CONSIDERATION OF DAM OPERATION33 1. In any event. water chemistry. Water column impacts could include changes in Physicochemical parameters and Community structure.). The ecosystem approach can work for dam owners/operators to a greater extent than they may perceive it to be working against them. that is a rather uncompromising ecological view of the relationship between nature and humans. These latter impacts might be viewed as externalities with less relevance for operators.8 ECOLOGICAL/ENVIRONMENTAL CONSIDERATION OF DAM OPERATION The following observations are generalized from the literature at large. water chemistry.
light.9 THE HISTORY OF DAMS DESIGN Throughout the history.34 CHAPTER 1. optimal performance standards might equate ecological integrity with revenue generation). INTRODUCTION community members. productivity. There will have to be a willingness to experiment (e. hydrologic budget. Doing so requires increased communication and trust among engineers. species richness. remineralization. The second one consists of environmental conditions exceed tolerance limits of original community members. The local externalities can cause riparian losses. Recent examples include the installation of a selective water withdrawal system on the Hungry Horse Dam and the (initially seredipitous) drawdown at Glen Canyon. . 1. Engineers need to evaluate the extent to which regard for the environment is codiﬁed as good practice. species diversity and ecological diversity that maintains the ﬂows of energy and materials through ecosystems) is the key mechanism for adapting to changes in environmental conditions through evolutionary change. dams have been designed to fulﬁll a speciﬁc purpose to civilizations. Ecological integrity (the genetic diversity.g.g. oxygen content. The ﬁrst one consists of temperature. Protecting diversity means protecting habitat. demonstration projects) with a commitment to redeﬁning operational technical directives as new knowledge emerges ( e. productivity and feedback Externalities could occur at a variety of scales and might include Local. compression and sedimentation. diversity and feedback. habitat fragmentation. These externalities could lead to loss of biodiversity and decreasing ecological integrity. Regional or Global externalities. Sediment impacts could include changes in Physicochemical parameters and Community structure. ecologists and other professionals in order to properly bound the problem. Dams can be operated in ways that decrease their environmental impacts. not only against loss but against fragmentation. The regional externalities can cause climate change while the global externalities can cause climate change and orbit dynamics. The following historical outline illustrate the main dam purposes and the development of their design.
the ﬁnal failure came about 1. It reached a height of 20m and was about 700m long. Heavy ﬂooding in the wet season and long droughts in the dry season makes farming very diﬃcult. a modern embankment dam with a capacity of 400 million m3 was built in 1986. The Kirsi dam of Iraq was built under the Assyrian king of Sennacherib about 355 km north of Baghdad. It was not repaired and lead to immigration of about 50. if the water is contained during the wet season and released periodically through the dry season. A 50m long spillway.8 with no road crest along the top. instead of the typical horizontal layer conﬁguration. Romans The Romans were highly advanced in hydraulic engineering from their many engineering feast such as aqueducts. One advantage . Ancient Dams The ﬁrst dam designed for irrigation purposes was the Marib dam in Yemen Capital City of Saba. The upstream face was vertical with a board overﬂow crest and a stepped downstream face. The most interesting aspect of this dam is that the ﬁll was placed in layers parallel to the slopes. However. The embankment slopes on each side were 1:1. they did not attempt dam construction until around 150 BC when they annexed Greece.000 people that depended on the water supply.1. This dam also did not have any type of impermeable element in its design. This dam was also a diversion dam. used to direct the river ﬂow into a 15 km long canal for irrigation purposes. However. about 15% of the average annual rainfall for the area. Recently. a constant water supply will be present. THE HISTORY OF DAMS DESIGN 35 1.9. 3 km upstream of the ancient dam site. The sills of both outlets were almost located at the same height from the riverbed.1 Irrigation Dams A steady supply of water for irrigation is extremely important for any civilization to thrive. The Marib was an embankment dam constructed about 510 BC.9. On each end of the dam an outlet structure was built of excellent aslar masonry.300 years after the completion of the dam. the capacity of the Marib dam was 30 milliom m3 . It was designed as a gravity dam constructed rubble masonry weirs. with a sill height was also included in the dam design located in the northern section of the dam.
was that they already possessed fully developed construction technique based on traditional tools such as levers. In many parts of the world. 1. it was essential in the southern countries of Europe like Italy and Spain and required storage reservoirs. which contrast to dams built in northern Europe that were built mainly of soil embankments. INTRODUCTION the Romans had. . The dams in these southern countries were mostly of masonry.9. They added volcanic ash or ground brick into the concrete mix to ensure that it would harden. Irrigation was used throughout Europe. the freshwater supply ﬂuctuates from enormous amounts during the rainy season to long droughts during the dry season. The Romans also used pulleys in multiple conﬁgurations to lift heavy objects vertically and horizontally. the Romans also added concrete to their dams. squares.36 CHAPTER 1. Some of the Romans advance tools included rules. The reason for dams built in Italy and Spain being masonry is most likely an inheritance from Roman civilizations who used embankments primarily for supporting elements. for this reason dams were invented to store large amounts of water from the end of the rainy season until the end of the dry season. The ﬁrst true arch dam built in Europe since the Roman time. was built in 16321640 near Elche.2 Dams Designed for Water Supply It is impossible for a civilization to survive without a constant fresh water supply. The main arch was 75m long and 9m wide at the crest and had a curve radius of 62m and a curve angle of 70◦ . picks and shovels. The design of this dam was later tested using the crowncantilever computer analysis that determined it to be satisfactory. The arch thrust was directed downwards. and spirit levels. The arch stretched across the main section of the gorge and butted into wing walls directed upstream on either side of the gorge. however. Traditional building materials were used in their design however. Medieval and Postmedieval Europe Damming for irrigation purposes was practiced throughout this era. to ensure a constant supply of water. even when submersed under water. since it did not have much support at the crest level. This allowed them to develop new shapes using formwork and concrete. since it is a basic need of all humans. plumblines.
the water supply for Istanbul in Turkey. The ﬁrst arch dam was constructed at Barcinas on Cubillas river 2km north of Granada.000 B. with an earth core that acted as the waterretaining element. The ﬁrst storage dam was built in 1560 on the remains of the Roman Belgrade dam. Medieval and Postmedieval Europe The Roman standard of public fresh water supply continued throughout this period. The dam was later raised by 1m.6m wide. using similar design methods as the original dam. It consisted of 1. but to lengthen the dam crest for easier passage of ﬂoodwater. It consisted of two dry masonry walls. It was massive gravity dam.3 million m3 reservoir capacity. which means that the downstream wall was subjected to the entire load. It is remarkable that the Roman did not use the arch dam earlier. A downstream embankment provided stability of the structure.1. since they had already become masters at employing it into their bridges and buildings.C. and was designed as a gravity dam. during the beginning of the ear. It was not until after the conquest of Istanbul in 1453 that the Ottoman Turks reconstructed some of the old Roman aqueducts. This dam was 12 high and 18m long. This dam also included a couple of unnecessary buttresses to add stability to the structure. Romans The Romans were the ﬁrst civilization to use an arch dam. The earth core between the two walls was about 1. THE HISTORY OF DAMS DESIGN Ancient Dams 37 The oldest dam in the world is the Jawa dam found 100 km northeast of Jordan capital Amman. in southeast Spain. An impervious blanket was also provided in front of the upstream heel. This dam was built around 3.3m upstream wall and a 1m downstream wall made out of masonry. with a radius about 14m and a 73◦ central angle.9. was dealt with by many underground cisterns within the city. with a height of 15m and a base width of 56% of its height. The design of this dam was quite complex considering that it was ﬁrst dam to be built. the arch design for this dam was not actually used to increase stability. The length of the dam was about 85m and had a 1. the ﬁrst true arch dam was built by the Romans in the Vallon de Baume 4km south of SaintRemy de Province in southeastern France. .
4m base width. It was designed as a gravity structure with height of 7m.000 blocks weighting 300kg each were carefully placement over the upstream and downstream walls to protect the dam against erosion. many dams were built to cause ﬂooding. a crest width of 6. It had three distinct sections. the upstream shell also made of rock and a centre core made out of ﬁne silty sand and gravel. injuries and damage are caused by rapid ﬂooding it is important to have ﬂood control methods. 17. Romans the Romans built the Cavdarhisar dam near Kutahya in western Turkey to protect against raging ﬂoodwater.9.38 CHAPTER 1. the lack of knowledge in hydrology and soil mechanics lead to the dam being over. not a ﬂood control measure. . INTRODUCTION 1. however. the downstream shell made of rock. designed. Originally. It was built of solid masonry and contained two outlets 5.7m which allowed a large amount of water to rush out in a short time. Dams have been used for ﬂood control since the ﬁrst civilizations were formed. the dams built for ﬂuming were made from timber. wide of 8m and a length of 80m. Construction problems were most likely encountered since it took longer than one dry season to built. and no evidence of any type of diversion could be found. It was 18m high with a length of 35m. however. An example of a ﬂuming dam is the Belcna dam in western Slovenia built in 1769. few of these dams have any remains. It stands 14m high and has a 113m crest. It also contained a large bottom outlet with a clear section about 11m2 that shows this dam was not designed to be a reservoir. Controlled ﬂooding was used for transporting wood by ﬂuming.3 Flood Control Dams Since large number of natural fatalities. from around the end of the 17th century the timber dam were replaced with masonry ﬂuming dams. Ancient dams The SaddelKafara dam of Egypt was built around 2600 BC.8m and a 12.1m in height and a width of 3. Medieval and Postmedieval Europe Throughout the period few dams were actually built to for ﬂood protection. about the same time as the beginning of the Pyramids.
Romans The Romans were the ﬁrst civilization to harness the power of water however.4m wide. The millhouse was located at one end of the weir and supplied directly with water from the reservoir. By the end of the 8th century.1. THE HISTORY OF DAMS DESIGN 39 1. because they also included two buttresses to increase the stability. One of the interesting dam built in this period for waterpower was the Castellar dam and millhouse in southwestern Spain. however. the height of the dam was 19m high. they were limited to mostly grinding cereals. The outlets were 1. The dam was built using blocks of shist laid horizontally in a lime mortar. about 300 years after the disintegration of the western half of the roman empire technology had gain sprouted new wings. It is thought that the water from two outlets was used to drive waterwheels further down the river. the dam’s reinforcement consisted of the fact that the curved central part at each end was embedded into the wing walls without any abutment blocks to absorb the horizontal arch thrust. The invention of the dam was an essential development in the rise of waterpower because it diversiﬁed the applications of both the vertical and horizontal water wheels. Dams were built in order to remove the waterwheels from rivers where debris.3 milliom m3 ). Medieval and Postmedieval Europe It took several centuries for Europe to recover from impacts of Christianization and Germanization.7m high. The dam’s reservoir capacity is very small(0. It was a 5. This large height gave a very high potential head of the water in the reservoir.9. a method of design that is used in modern lowhead dam power plants.4 Power Dams Waterpower was a technology that did not really take oﬀ until after the 11th century.9. The designers of this dam must have been unsure of their sophisticated design. which could be transferred into waterwheel power. Another equally modern design characteristic was the location .2m and 1. with a height of 1m through the dam’s base. One of the most sophisticated dams built by the Romans was the Monte Novo dam located 15km east of Evora in southeast Portugal. 52m long arch dam with a radius of 19m and a central angle of 90◦ . ice jams and ﬂoods played havoc with the waterwheels and their structures.
The following is a summary of 7th and 8th century dams in central Arabia.0 113 Ardab 5. Adhaim dam was built on the Adhaim river. T able 1. which was near the toe of the 19m high reservoir. Adhaim Dam had an .0 205 All of these dams were of the gravity type. Strong inclinations were applied to dams with relatively wide base. although it was built around 1500 AD. Whereas.40 CHAPTER 1. They all have two outer walls of dry masonry and an earth or rubble core in between.0 130 Qusaybah 30. Many of the concepts built into this dam have direct links to modern methods. the other had inclined walls as much as 60%. INTRODUCTION of the millhouse. South Western Asia The Moslems conquered southwestern Asia in Seventh and Eighth centuries as far as the Indus river and Uzbakistan.4 Summary of 7th and 8th century dams in central Arabia Nearest City Name Height (m) Length (m) Mecca Agrab 4. one third of them like Saammallagi. were built. most of these dams are already destroyed. Although.0 225 Medina Hashquq 2. like Darwaish.0 150 Sammallaagi 11. Several irrigation dams around the new power centres of Mecca and Madina.5 Arabia The Moslem World Arabs quickly developed their country within one century after the impact of new Islamic religious founded by prophet Mohammed. During this time. This dam was located 150 km north of Baghdad. The dam was 130 meters long and 12 meters high. 1.5 315 Darwaish 10.9. The walls of some of these dams were vertical. are operative and more or less in their original shape. An equal number of vertical shafts lead the water to the horizontal water wheels. The sturdy end and intermediate walls of the mill house act as buttress because the width of the dam was only 36% of its height.
6 Ghaznavid Gravity Dams Name Gishtbank Khan Sheshtaraz Soltan Mahmud Nearest city Samarkand Saamarkand Mashhad Kabul Height (m) Length (m) 8 25 15. The Amir dam is a large power dam including 30 water wheel being driven by a retained water.7 shows the summary of the data. Afghanistan and Uzbakestan. Upstream face was vertical and downstream one was stepped to 43% average inclination. several large gravity dams were built under Ghaznavid rule. T able 1. Table 1. Mongolians initiated the construction of several dams in Iran. T able 1. In the 10th century.9.2 52 25 35 32 220 Genghis khan from Mongolia started to conquer the world in 1205.5 Dimensions and name of the dams on the Kur river Name Amir Feizband Tilkan Mawan Djahanabad Height (m) Length (m) 9 103 7 222 6 162 6 66 5 50 Towards the end of the 10th century.1. These dams were built in countries such as Iran. Table 1.5 shows . Table 1.6 contains some information about Ghaznavid gravity dams. east of Shiraz in southern of Iran. a whole series of irrigation and power dams were built on Kur river. This dam has a trapezoidal cross section with a relatively steep down stream face and a base width of more than twice its height. dimensions and the name of the dams on the Kur river.7 Mongolians Dams . THE HISTORY OF DAMS DESIGN 41 equal width all along its length. T able 1. Khan dam was built of granite ashlar masonry and was barely stable.
the construction of arch dams discontinued and only gravity dams were built. INTRODUCTION Year of completion Name Nearest city Type Height (m) Length (m) 1285 Saveh Tehran Gravity 25 65 ~1300 Kebar Arch 26 55 ~1350 Kalat Mashhad Gravity 26 74 1400 Abbas Tabas Arch 20 ? ~1450 Golestan Mashhad Gravity 16 The Saveh dam was the ﬁrst structure built during this time. This dam is 150 km northwest of samarkand. It was a gravity dam.6 Development of the Modern Dams Unlike past periods in history. His experiments determined the natural slope response of soils. Most of the well known dams during this time. deLaplace’s formula of how water ﬂows in a soil. the modern dam is designed using speciﬁc principles. professor of hydraulics in Austria. As time went on. The design of this dam was not successful.42 CHAPTER 1. were located in northeastern Iran and Uzbakistan. 1. This graphical method is commonly known as a ﬂownet. increasing the advancements of both dam principles and construction methods. The river crushed the dam after the ﬁrst impoundment. to increase the height of embankment dam designs. engineers were then able to choose diﬀerent soils based on their internal angle of friction. Henri Gautier was one of the ﬁrst to study slope stability of diﬀerent soils. The height of this dam is 15 meters and it has a length of 85 meters.9. Khajoo dam is one of the most beautiful dams not only in the Moslem. The modern embankment dam has improved greatly with the increase in research on hydrology and soil mechanics. In 16th and 17th century. Studies were conducted on how soil permeability changed for diﬀerent soils. developed a graphical method of Pierre S. Abdullahkan dam was built in the beginning of the 16th century. has been the formation of engineering schools. but in the entire world which is located in Isfahan of Iran. In 1186 Philipp Forchheimer. Engineers were subdivided into speciﬁc disciplines. This changed the individualistic craftsmanship of engineering into a profession based on scientiﬁc principles. The single most important theoretical . Its downstream face was stepped to an average inclination of 69%. The largest change in the design of a modern dam compared to the earlier dams.
and the upper ones had to partially support the cantilever. These new construction techniques have spread throughout the world at a rapid pace. H. the arch dam took on many new shapes. bulldozers and vibrators all grew in size. by determining that the consolidation was caused by the dissipation of the water pressure in the soil pores. The analysis of stress in structures in their elastic condition. Throughout this period. This trend will not be able to continue as the world population continues to increase and fresh water. These concepts were the beginning of understanding how the force in gravity dam designs acted and helped to increase the height.9.1. This concept is still considered the single most important concept of modern soil mechanics. Scraper. the Nurek earth ﬁlled embankment dam in Tajikistan was completed. It also considered the median vertical section of the dam to be a cantilever ﬁxed at the base. The distribution of the loads resulted in a considerable relief for the lower arches. In the late 1970’s early 1980’s vibrations rollers replaced the tractor mounted immersed vibrating system. It stretched to a record height of 300m and had an embankment volume of 59 million m3 . the concepts of the modulus of elasticity. without increasing the amount of materials used. With the modern ear. have played a roll in the construction and design methods for gravity dams. Roller compacted concrete is among the newest techniques implements in gravity dam construction. and vibrators were attached to caterpillar tractors. New techniques for placing concrete. that is equal to the total stress minus the pore water pressure. He found the explanation for why clays consolidate. THE HISTORY OF DAMS DESIGN 43 development of soil mechanics was published in 1925 by Karl von Terzaghi. Engineers need to constantly search for a better solution. increasing both the dimensions of dams and the speed of completion. Navier. This ﬁnding led Terzaghi to the concept of eﬀective stress. taking all aspects . In 1980. improved mechanical construction method have caused the greatest impact on all types of modern dam designs. In the early 1950’s Switzerland introduced bulldozers to spread concrete. The number of new dams being built has decreased considerably. irrigation and ﬂood control needs increase as well. This new method did not merely design the arch dams as astach of individual arches but as an interdependent system of arches. Breakthrough advancements in arch dam design came in 1880’s when Hubert Visher and Luther Wagnore developed a new method of arch dam design. resulting in the modern dams being completed in record time. and safety stresses was introduced in 1819 by Louis M.
This warm. brown underfur. It is important that proper monitoring programs are setup to ensure that failure do not occur. One method of monitoring is a system of alarms downstream of the dam to warn of potential failures. lodges. nose. The past is a collection of design ideas. The beaver’s facial features also allow for its aquatic lifestyle. dams continue to grow in height and capacity. and is covered with scaly skin. The tail is shaped like a paddle. thus keeping water out of its mouth. their impacts on society and the environments also increases. toes. The toes on the beaver’s large hind feet are webbed for swimming. As a danger signal to other beavers. INTRODUCTION of design and the repercussions of that design into account. tail. waterproof coat allows the beaver to swim in icy water in the wintertime without discomfort. the normally placid animal makes a loud noise by slapping its tail on the water’s surface. The animal carries objects underwater in its mouth by closing loose lips behind prominent front teeth. and lips are so constructed that the animal is well equipped for life in the water as well as on land. ears. both good and bad. or gnawing animals. 1. The second toe on each hind foot ends in a double claw with which the animal combs its fur. and canals. the beaver has welldeveloped front teeth. As time progress onwards. The animal is also known for its aquatic lifestyle as well as for its beautiful fur. These teeth . storehouses.10 BEAVERS A mammal belonging to the order of rodents. A beaver may remain submerged for up to 15 minutes. By using teeth and paws. With the growing size and capacity of modern dams. because their eﬀects will be devastating. The short. The beaver has a thick body covered with a coat of long. The tail serves as a prop when the beaver sits upright and as a rudder and scull when it swims. The front feet are small and handlike and are used for picking up and carrying various objects. As they become larger. that should never be overlooked when faced with a new design challenge. thick head has small rounded ears and a nose which are equipped with valves that close when the animal swims underwater. the beaver has been recognized as a master engineer. Most of the beaver’s physical characteristics of the fur. the environment is an aspect that needs to be carefully controlled.44 CHAPTER 1. dense. beavers construct dams. broad and ﬂat. reddishbrown outer hairs and soft. Like other rodents. the environmental impact and awareness therefore increases as well.
The beaver has a total of 20 teeth. A mother will sometimes raise not only her own oﬀspring. Newborn kits weigh from 8 to 24 ounces (225 to 680 grams) and are about 15 inches (38 centimeters) long. It is in these artiﬁcially created ponds that beavers build their lodges and storehouses. Some dams more than 1. rivers. and lakes bordered by timberland. Beavers live most of their lives in or near water. A mated pair locates a fairly deep. They settle along banks of streams. They gnaw down a number of aspen. They dig a burrow into the bank. Into this foundation the beavers ﬁt and pile more saplings. The animal’s teeth are always growing to make up for wear. Normally a family of beavers can build a dam 35 feet (10 meters) long in about a week. Beaver dams come in all shapes and sizes. BEAVERS 45 have a very hard layer on the front surface and a softer backing.1. with tails that measure 3 1/2 inches (9 centimeters). They usually mate for life. Large beaver populations have been credited with reducing ﬂooding because of the dams they build across streams. starting below the surface of the water and slanting upward to a small room above the highwater mark. shallow site in the water as a place to build a dam. The life span of the animal may be as long as 19 years. The adult beavers select a narrow. Since this softer part wears away quickly. A female ﬁrst breeds when it is by their third summer the young beavers are mature and ready to mate. These they drag to the site and bury in mud with the butt ends pointing upstream. and they are weaned by six months. slowmoving stream. They are out learning to swim when they are only a month old. The reservoirs created by the dams are places where the beavers feel secure. it leaves the thin chisel edge of the front layer exposed.000 feet (300 meters) long have been found.10. On the other hand. Their eyes are open at birth. Constant ﬂooding can also damage valuable timber and block routes of migrating salmon. or kits. adding mud and stones until a strong barrier is completed. This structure allows enough seepage or overﬂow to keep the water in the reservoir fresh. but these are . but also the young of another female that has died. they may also cause the ﬂooding of roads and woodlands because of the reservoirs of water that build up behind the dams. They are born in the spring. or willow saplings. about four months after conception. A beaver begins its life in a litter of from two to eight young. Beavers are social animals. Not until the following autumn does the couple set about building their permanent home—the lodge. Four is the usual number. This is only a temporary residence in which the ﬁrst litter will be born in the spring. birch. they live in colonies and work together. A family usually consists of a mature pair of beavers and two sets of oﬀspring.
It consists of a platform of carefully interlaced branches held together by clay and dead leaves.46 CHAPTER 1.000 feet (300 meters) in length and from 2 to 3 feet (60 to 90 centimeters) in width. When they live in ﬂat areas they sometimes build canals. the beavers gnaw whole groves of trees and sink the wood in the pond near their lodge. This result is obtained because the beaver adds material to the dam from the upstream side and. The dam is quite steep on the downstream side but slopes gently into the pond on the upstrean side. in North America. Another entrance is used for the transportation of wood for winter food. . lodges. all in a perfect safety from predators. After an autumn of toil. and up to ﬁve. These canals allow the beavers to more easily transport the logs that are too heavy to drag overland. by doing so. unintentionally achieves a dam with the strongest possible conﬁguration and therefore provides maximum security for the colony. These underwater entrances help protect the beavers from attacks by predators which include. INTRODUCTION the work of generations of beavers. Entrances to the beaver lodge often open underwater. From a distance. willow. must be plastered with mud. the beavers fashion a domeshaped roof over it. berries. Beavers live in a structure called a lodge. Beavers do most of their building and food gathering at night. In the summer they also eat water plants. Beavers’ dam is usually a high ﬂat dam and having unwittingly created the lake which aﬀords them safety and food. A canal may measure more than 1. dry lodges only to pluck a twig or branch from the storehouse. Beavers move awkwardly on land. the lodge resembles a heap of tree branches and mud. The lodge is built in the riverbank or in the pond created by the dam. such entrances.5 meters) high. the lynx. A steep and narrow entrance is used by the beavers for entry and exit. and canals is ﬁnished.” Their diet consists primarily of fresh green bark and wood such as poplar. and birch. they prefer to swim. it does not freeze to the bottom and the beaver can swim under the ice in winter time to get sticks from the food pile accumulated the previous fall. This collection forms the underwater winter ”storehouse. beavers spend the winter resting. Before the coming of winter. swampwood. They swim out of their warm. return to the feeding chamber of the lodge and leisurely eat the food. There are at least two. When the platform has been built up a few inches above the water. When construction on dams. and fruit. and the wolf. the wolverine. Because the water in the pond is a meter deep or more. so that the animals may pass in and out below the winter ice. the entire structure. which may enclose a room more than 5 feet (1.
It is the quest for the beautiful pelt that has most drastically reduced the beaver population. beginning early in the 1600s and continuing through the early 19th century. which is used in perfumes. At one time beaver pelts were a medium of exchange. which is at its best in late winter and spring. and their musk glands. Almost too late. beavers had become an endangered species by the mid19th century. In fact. . made into caps and capes. It was also a popular medicine in the Middle Ages. was prompted by the search for beaver fur. were a staple of the world’s fur trade. conservation laws were passed throughout the world. and the beaver was saved and was resettled in some areas where it had once been common. possibly because of various natural causes as well as trapping. apparently used to heal ailments ranging from headaches to dropsy. much of the exploration of North America. The healing ability of castoreum comes from its salicylic acid—a basic ingredient of aspirin—which the beaver acquires by eating willow bark. During the 17th.10. Beaver populations have been diminishing for centuries. Beaver ﬂesh in general was highly esteemed during the Middle Ages. has long been highly valued. their tails. beaver skins. Both sexes possess scent glands at the rear of the body that produce a liquid called castoreum. BEAVERS 47 Although once plentiful throughout the wooded parts of the Northern Hemisphere. and early 19th centuries. The soft. 18th. They have been hunted for their fur.1. The beaver has in the past been hunted for its scaly tail. thick underfur of the beaver. The beaver population in Russia and in North America has recovered to the extent that bans on hunting imposed earlier in the 20th century have been slightly modiﬁed. which was considered a culinary delicacy.
INTRODUCTION .48 CHAPTER 1.
They are vital for hydropower generation. In this chapter. The capacity of the reservoir may vary depends on the shape of the valley and topography of the dam site. Dams diﬀer from other structure because of their size and their containment of water. In some locations. navigation. The ultimate fate of all dams and reservoirs. drinking water.Chapter 2 RESERVOIR 2. Every reservoir that impounds water behind a dam is a real or potential threat to those who live and work at the downstream side of the dam. The more capacity of the reservoir. Special attention must be taken to understand their behaviors. The reservoir provides an extra eﬀect of dynamic action into the dam response. we will establish reservoir’s governing equation of the motion and associated boundary conditions of the damreservoir system. is deterioration and failure or ﬁlling by sedimentation. Modern technology combining the knowledge of construction and accurate design to reduce the risk that is inherent in dam and reservoir system. 49 . Today’s dams are a critical part of the nations infrastructure. the eﬀects of a sever earthquake may be dangerous to the integrity of the damreservoir structure and may tend to the destruction of the system. recreation and ﬂood control.1 INTRODUCTION The ﬁrst dams built by early man were low earth or rock structures designed to impound and divert water for agricultural use. the more threat to the area in case of a failure.. irrigation.
For steady ﬂow. z).2 GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION Fluids are composed of molecules in constant motion and collisions. pressure and temperature which are their macroscopic structure. vz in x. the value of the velocity at a position remains invariant with time. velocity v. We can stipulate ﬁx coordinates x1 . In order to deﬁne the velocity of a particle we must deﬁne the position of the particles. the velocity of all particles can be as v(x. may be diﬃcult for engineering purposes. Consider some variable as a = V a0 dV . Therefore for a big V we get diﬀerent values of a for a range of V . RESERVOIR 2.1 Velocity Field In a deformable system there are an inﬁnite number of particles. v = v(x. There must be a volume V such that a range of V value exist over which a is constant. respectively. The restriction V 3 ¿ L is for a to be measurable at a point while 1 the criteria δ ¿ V 3 is for a to be statistically signiﬁcant and deterministic (always get same answer if we start from same initial condition and boundary conditions). y. Instead we are interested in average measuring of the molecule manifestations such as density. To take account of each molecule in a ﬂow. we can express velocity of function .1). in which a0 has a random characteristic(ﬁgure 2. y and z direction of coordinates. spatial coordinates. The continuum concept oﬀers a great deal of simpliﬁcation in analysis. two procedures may be used to study ﬂuid particles. A ﬂuid point (particle) represent a spatial average over some small volume R 1 V . A ﬂuid has no structure. LAGRANGIAN AND EULERIAN VIEWPOINTS In investigation of the ﬂuid motion. 1 We want δ ¿ V 3 ¿ L. v is called ﬁeld velocity vector.2. The variable a can be density. ρ. etc. y. vy . 2. Using this. at the speciﬁc time. in which δ is the largest length scale of the ﬂuid structure (say mean free path) and L is the length scale of the domain of the 1 problem..50 CHAPTER 2. y1 and z1 in the velocity ﬁeld functions and letting time pass. z. However if we make V successfully smaller we get same value of a. They are not distinguished by their microscopic (molecular) structure. t) which is consisted of three components of velocity vx .
The x. we can study the motion of a single particle in the ﬂow by following the particle in its path. z). The change of ﬁsh concentration with respect to time reﬂects the motion of the boat as well as time variation. y and z can be expressed as a function of time x = x(t).1: Fluid point at any time passing this position as v(x1 . This approach is called the Lagrangian viewpoint where v = v(x.2. from the Eulerian point of view. holding x. y and z to locate the particle. This viewpoint is called the Eulerian viewpoint. y = y(t) and z = z(t) (ﬁgure 2. suppose we are standing on a bridge and note how the concentration of ﬁsh (c) just below us changes with time.2. y. y1 . The Eulerian viewpoint is easier to use in continuum mechanics because it is concerned with the description of motion at a ﬁxed position. y and z constant. In the partial time derivative. This approach means continuous variation of the x. In continuum mechanics this method requires the description of motion of an inﬁnite number of particles and thus becomes extremely cumbersome. We ∂t can also assume ourselves on a boat and moving with around the river. On the other hand. z1 .2). Using this approach we study the motion of a continuous string of particles which pass the ﬁxed point. therefore by ∂c we mean partial of c with respect to t. The Lagrangian viewpoint is used mainly in particle mechanics. In this case. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 51 Figure 2. This is called total time variation and can be . t). the position is ﬁxed in space.
dy and dz are the components of the velocity of the boat. Now if dt dt dt we turn oﬀ the engine of the boat and ﬂoat along counting ﬁsh (the lagrangian viewpoint).52 written as: CHAPTER 2. position changes or temperature changes.2 System and Control Volume An identiﬁed quantity of matter is called system. but entailed the same . vy and vz are the components of the local ﬂuid velocity v. the velocity of the boat would be simply the velocity of the streams.2. The change of ﬁsh concentration in this case is called substantial time derivative and is written as follows: Dc ∂c ∂c ∂c ∂c = + vx + vy + vz Dt ∂t ∂x ∂y ∂z in which vx . It may undergo shape changes. Figure 2.2: a)Lagrangian viewpoint b)Eulerian viewpoint 2. RESERVOIR Dc ∂c ∂c dx ∂c dy ∂c dz = + + + Dt ∂t ∂x dt ∂y dt ∂z dt in which dx .
2. control volume is a deﬁnite volume in space and its boundary called control surface. We are trying to establish a relation between the rate of change of N in system with variation of this property inside the control volume.1) ∆t→0 lim ηρdV ∆t ¢ t+∆t − lim ∆t→0 ¡RRR ηρdV I ∆t ¢ t In the ﬁrst term on the right hand side of equation.3 Reynold’s Transport Equation Consider N as properties of a substance whose measure depends on the amount of the substance presented in a system. The amount of matter inside control volume may change but its shape is ﬁxed. 2. we can say that. The control volume is considered with system at time t and is ﬁxed.2. as ∆t goes to zero the region II becomes that of control volume . The streamlines correspond to those at time t. Rate of change of N with respect to time for the system can be written as: µ ¶ dN DN = = dt system Dt ¡RRR ηρdV + III RRR ηρdV II ¢ − t+∆t ∆t ¡RRR ηρdV + I RRR ηρdV II ¢ t ∆t→0 lim The above equation can be arranged as following: Ã ¡RRR ¢ ¡RRR ¢! ηρdV t+∆t − ηρdV t DN II II = lim + Dt ∆t→0 ∆t ¡RRR III (2.2.cv. the concept of system and control volume are related to the Lagrangian and Eulerian viewpoints. Then: . GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 53 matter. y. An arbitrary ﬂow ﬁeld of v = v(x.3). RRRThe distribution of N per unit mass will be given as η. Region II is common to the system at both time t and t + ∆t. t)is observed from some reference xyz.2. On the other hand. such that N = ηρdV with dV representing an element of volume with mass density of ρ. z.1.. The system at time t and t+∆t can be divided into three regions. We consider a system of ﬂuid with ﬁnite mass at time t and t + ∆t (ﬁgure 2.
3: Moving system . RESERVOIR Figure 2.54 CHAPTER 2.
Thus.3).dA ηρv.2)and second.4) − lim I t ∆t→0 = Net eff lux rate of N f rom cs which is the volume of ﬂuid that has crossed the cs in time dt.dAdt is negative while for the ﬂuid going out of the control volume the expression is positive. the third term approximates the amount of N that has passed into the cv during ∆t through remaining of the cs (inﬂux). Similarly. As ∆t goes to zero this becomes exact rate of eﬄux of N through the cs.dAdt (2. the term dV can be written as: dV = v.1 shows the amount of the property N that has crossed part of the control surface. rate of change of N inside cv having the shape of the system at time t (equation 2.2.2) The second term in equation 2.3) In conclusion.dA+ ZZ ηρv. the rate of change of N for system at time t are sum of ﬁrst.2. Therefore. Multiplying (equation 2. the lase two integrals give the net rate of eﬄux of N from cv at time t as: ¡RRR ηρdV ∆t ¢ ¡RRR ηρdV ∆t ¢ ∆t→0 lim III t+∆t (2.dA leaving the control volume through the indicated area dA.3.dA Inf lux rate through cs ≈ − ZZ N et ef f lux rate through cs ≈ ARB ZZ ARB ALB ALB .3 can be written as follows: Ef f lux rate through cs ≈ ZZ ηρv. In equation 2. the second term is the average of eﬄux of N across ARB during time interval of ∆t. shown as ARB. cs. the net rate of eﬄux of N through the cs at time t (equation 2. dA is the normal outward vector on control surface.dA ηρv.4) by ρ and dividing by dt then gives the instantaneous rate of mass ﬂow of ﬂuid ρv. Therefore equation 2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 55 ∆t→0 lim Ã ¡RRR ηρdV II ¢ − t+∆t ∆t ¡RRR ηρdV II ¢! t ∂ = ∂t ZZZ cv ηρdV (2. For the ﬂuid entering the control volume the expression dV = v.
the time rate of change of N is in eﬀect observed from the control volume and velocities and time rates of changes are those seen from control volume.2. To go from system approach to control approach. Since we could use a reference xyz having an arbitrary motion.6 which is the ﬁnal form of the continuity equation expresses that the net eﬄux rate of mass through the control surface equals the rate of . we can write the Reynold transport equation as following: ZZ ZZZ DM ∂ =0= (ρv.56 CHAPTER 2.dA)+ ρdV Dt ∂t cs cv The velocity v and the time derivative are measured relative to the control volume. η = 1. RESERVOIR In the limit as ∆t → 0. Thus for a constant mass in a system at any time t. the mass of a RRR ﬂuid system and it is M = ρdV .dA) = − ρdV (2. Thus the Reynold transport equation will then instantaneously be correct if we measure the time derivatives and velocities relative to the control volume.dA). Now the equation 2. we use Reynold transport equation in which N is for our case M. In the Reynole transport equation v is measured relative to some reference xyz and the control volume was ﬁxed in this reference. Above equation can be written in the following form: ZZZ ZZ ∂ (ρv. Therefore. so we can express RR the right side of the above equation as cs η(ρv.4 Continuity Equation A system always entails the same quantity of matter. no matter what the motion of the control volume may be.6) ∂t cs cv Equation 2.1 can be written as: ZZ ZZZ DN ∂ = η(ρv. Thus. the mass M would be constant.dA)+ ηρdV (2. it means that the control volume can have any motion. As a consequences. where the integral is the closed integral over the entire control surface.5) Dt ∂t cs cv This is called the Reynold Transport equation and it is a change from system approach to control volume approach. 2. Therefore. the approximation becomes exact. v in the above equation is measured relative to control volume.
it can be written as: .The continuity equation can be written as: ZZ v.7) = Dt ∂t cs cv The equation of linear momentum is a vector equation and can be divided into its three components. P = v(ρdV ). Since.dA) = 0 cs The other case is the case of incompressible ﬂuid having constant ρ. or simply v. The case of steady ﬂow in which all ﬂuid properties at any ﬁxed position in the reference must remain invariant. 2. Two special cases of continuity can be considered. F if the resultant of the all external forces acting on the system and v the time derivative are taken from the inertial references. we have N = P = mv. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 57 decrease of mass inside the control volume. the second law of Newton is based on the Eulerian viewpoint. we are following a system. Then the Reynold transport equation can be written as followings: ZZ ZZZ DP ∂ v(ρv.dA = 0 cs A more detailed deﬁnition of compressibility is brought in the proceeding sections.5 Linear Momentum Equation If we take linear momentum as parameters N which is used as a general term (vector and scalar) in the Reynold transport equation. the continuity equation can be written as: ZZ (ρv.2. The term ηRRRthis case would become momentum per unit mass. Thus.2.dA)+ vρdV (2. Newton’s second law states that: ·Z Z Z ¸ X d dP F= vdm = dt system dt system P In above equation.2. in then.
then the derivatives in the right hand side is taken from a inertial references and we may use the Newton’s second law (equation 2. Surface forces. the positive direction of the reference axis xyz are the same as the positive direction of the velocity components as well as the surface and body force components. z. t) given as forces per unit area on the boundary surface.9 means that: Sum of surface forces acting on the control surface and body forces acting on the control volume are equal with the sum of the rate of the eﬄux of the linear momentum across the control surface and the rate of increase of linear momentum inside the control volume. T(x. y.8) to replace it in the Reynold equation and then we get: ZZ cs TdA + ZZZ cv BρdV = ZZ cs ∂ v(ρv. For example in x direction we can write it as follows: ZZ cs Tx dA + ZZZ cv Bx ρdV = ZZ cs ∂ vx (ρv.9) Equation 2. y. RESERVOIR X D F= Dt ·Z Z Z ¸ DP vdm = Dt There are two types of the external forces.dA)+ ∂t ZZZ cv vρdV (2. Thus.dA is independent of the sign consideration. The above equation can be written as followings: ZZ bs TdA + ZZZ ib BρdV = DP Dt (2.58 CHAPTER 2.dA)+ ∂t ZZZ cv vx ρdV (2. Force acting on the material inside the boundary. t) given as force per unit mass. Sign of v. if we ﬁx the control volume in inertial space. B(x.8 is the Newton’s second law for a ﬁnite system. z.8) Equation 2.10) In this book for simplicity. . ib are called body force. Equation 9 is a vector equation and can be written for each components. bs. Knowing that the control volume and system are coincident at time t.
ρvx x+∆x ∆y∆z − ρvx x ∆y∆z (ﬁgure 2. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 59 2.4: Volume element The net eﬄux rate along the plane of control volume that are perpendicular to the x axis (x planes of the element) is.2. we are trying to ﬁnd out diﬀerential equations valid at any point of ﬂuid.2. y. They don’t give any detail information of the ﬂow everywhere inside the control volume.6 The Equation of the Motion The continuity equation and the linear momentum equation was developed for ﬁnite systems and their were related to control volume. we consider an inﬁnitesimal control volume in the shape of a rectangular parallelepiped ﬁxed in xyz for a general ﬂow v(x. Here. Writing continuity equation in the form of diﬀerential equation. The equations give an average values of quantities or only components of resultant forces. z) measured relative to xyz.4) If we write the similar equation for the axis in y and z direction and add them up we get the net eﬄux rate as following: .2. Figure 2.
∂t Equating the net eﬄux rate and the rate of decrease of mass inside the control volume and dividing all resulting equation by ∆x∆y∆z and taking the limit as ∆x. RESERVOIR Net eﬄux rate = [ρvx x+∆x ∆y∆z − ρvx x ∆y∆z] + [ρvy y+∆y ∆x∆z − ρvy y ∆x∆z] + [ρvz z+∆z ∆x∆y − ρvz z ∆x∆y] The rate of mass accumulation inside the control volume would be ∂ρ ∆x∆y∆z. The surface stress on each side of .60 CHAPTER 2.5 ). we consider an inﬁnitesimal system in the shape of a rectangular parallelepiped ﬁxed in xyz (ﬁgure 2. The linear momentum equation is a form of Newton’s second law which has an Eulerian viewpoint.5. ∆y and ∆z approach zero. we get: ∂(ρvx ) ∂(ρvy ) ∂(ρvz ) ∂ρ + + =− (2.v = ∂vx ∂vy ∂vz + + ∂x ∂y ∂z Then we can write the general form of the continuity equation 2.11 in the following form: ∇.11) ∂x ∂y ∂z ∂t This is called diﬀerentia continuity equation and one special cases of it would be case of steady ﬂow which results in: ∂(ρvx ) ∂(ρvy ) ∂(ρvz ) + + =0 ∂x ∂y ∂z The other case of incompressible ﬂow would be as following: ∂vx ∂vy ∂vz + + =0 ∂x ∂y ∂z Introducing the divergence operator which for a vector ﬁeld v can be deﬁned as: ∇. Consider a volume element ∆x∆y∆z as shown in ﬁgure2.(ρv) = − ∂ρ ∂t Writing the linear momentum equation in the form of diﬀerential equation.
in which the stress is located on and j is the direction of the stress. If we add all of the surface forces we will get: [Txx x+∆x ∆y∆z − Txx x ∆y∆z] + [Tyx y+∆y ∆x∆z − Tyx y ∆x∆z] + . j = x.2. plane. i. Here.10). we start by considering the x direction(equation 2. Thus we can write for each components of it. As it was already mentioned equation of linear momentum is a vector equation (equation 2.9). z where i is an indication of the side. For example Txy is the components of the stress on a plane normal to x axis and directed toward y axis. y.10 is all of the surface forces acting on the element at the x direction. Similarly.5: Rectangular Parallelpiped element a cubic element cab be written as Tij . The resultant forces acting on the x planes along the x directions is Txx x+∆x ∆y∆z − Txx x ∆y∆z. The ﬁrst term on the left hand side of the equation 2. The resultant of forces acting on the y plane along the x direction is Tyx y+∆y ∆x∆z − Tyx y ∆x∆z. It can be shown that for equilibrium we must have Tij = Tji . GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 61 Figure 2. for the forces on the z plane along the x direction is Tzx z+∆z ∆x∆z − Tzx z ∆x∆y.2.
62 CHAPTER 2. The same can be written for the y and z planes. The ﬁrst term on the right hand side of equation 2.10 is the body force and is: ρgx ∆x∆y∆z where gx is the component of the gravitational force along the x axis.14) ∂x ∂y ∂z ∂x ∂y ∂z ∂t . we obtain the xcomponent of the equation of the motion as: ∂Txx ∂Txy ∂Txz ∂(vx ρvx ) ∂(vx ρvy ) ∂(vx ρvz ) ∂(ρvx ) + + +ρgx = + + + (2.12) ∂x ∂y ∂z ∂x ∂y ∂z ∂t Similarly. Thus if we add them up we get: [vx ρvx x+∆x −vx ρvx x ] ∆y∆z + [vx ρvy y+∆y −vx ρvy y ] ∆x∆z + [vx ρvz z+∆z −vx ρvz z ] ∆x∆y The second term on the right hand side of the equation 2. ∆y and ∆z approaches zero.10 is the accumulation of the momentum inside the element and is: ∂(ρvx ) ∆x∆y∆z ∂t If we substitute all the terms into the equation 2.13) ∂x ∂y ∂z ∂x ∂y ∂z ∂t ∂Txz ∂Tyz ∂Tzz ∂(vz ρvx ) ∂(vz ρvy ) ∂(vz ρvz ) ∂(ρvz ) + + +ρgz = + + + (2. for the components in the xx and y direction it can be written as: ∂(vy ρvx ) ∂(vy ρvy ) ∂(vy ρvz ) ∂(ρvy ) ∂Txy ∂Tyy ∂Tzy + + +ρgy = + + + (2.10 is the net rate of eﬄux. the rate at which x component of momentum enters at x is vx ρvx x ∆y∆z and the rate at which it leaves at x + ∆x is vx ρvx x+∆x ∆y∆z. RESERVOIR [Tzx z+∆z ∆x∆z − Tzx z ∆x∆y] The second term on the left hand side of the equation 2.10 from the above equations and dividing the entire resulting equation by ∆x∆y∆z and taking limit as ∆x.
When three components added together vectorially. GENERAL FORM OF RESERVOIR’S EQUATION OF MOTION 63 The quantity ρvx . . If we use the equation of continuity to substitute it in the equation 2. The terms ρvx vx .ρvv and ∇.mass velocity vector. ρvy vz .T + ρg = ∇.ρvv + (2. we get: Dv = ∇. which is the dyadic product of ρv and v. ρvx vy . Tyx . It can be seen that momentum balance is equivalent to Newton’s second law of motion. Txz . ρvx vz .T. ρvy and ρvz are the components of the ρv . Txx . The above equation is valid for any continuous medium.15) There should be noticed that ∇. etc.12.16 we can insert expressions for various stresses in forms of velocity gradient and pressure. are the nine components(six components because of symmetry) of the momentum ﬂux ρvv.. are the nine components( six components because of symmetry) of the T. Similarly.T are not simple divergence because of tensorial nature of the ∇. The above equation is the general form of the linear momentum equation.16) This is an statement of the Newton’s second law which was developed for a volume of ﬂuid element moving with acceleration because of the forces acting upon it.T + ρg ρ {z } {z} Dt  {z } surf ace f orces on element body f orces on element mass per unit volume per unit volume per unit volume time acceleration (2. known as the stress tensor. Txy . the above three equations and write them in a vector format as followings: ∂ρv ∂t ∇. etc. Therefore. we need to ﬁnd a relation between various stress and velocity gradient.ρvv and ∇.2. In equation 2. We can combine. we will get: ∂Txx ∂Tyx ∂Tzx Dvx + + + ρgx = ρ ∂x ∂y ∂z Dt The same can be written for y and z components.2.
All gases and simple liquids are described by the later formula. Txy is not proportional to the change of rate of velocity in normal direction of it. Laminar ﬂow can be described as a wellordered pattern whereby ﬂuid layers are assumed to slide over one another. which as in liquids the molecules travel only very short distances between collisions.64 CHAPTER 2. For a laminar ﬂow. or inviscid ﬂuid. the principal mechanism for momentum transfer is the actual colliding of the molecules.6). well ordered ﬂow.3 VISCOSITY A ﬂuid with zero viscosity is called a nonviscous. Experiments show that in some ﬂuids. whereby ﬂuid particles move in straight. In Newtonian law of viscosity for a given temperature and pressure µ is constant.002 centipois for air. In this case while ﬂuid has irregular molecular motion. Txy ∝ ∂vx or Txy = µ ∂vx . Whereas for liquids the viscosity usually decreases with increasing temperature. At room temperature. molecules travel long distance between collisions and the momentum is transported by the molecules in free ﬂight. the shear stress on an interface tangent to direction of ﬂow is proportional to the distance rate of change of velocity. For gases at low density viscosity decrease as temperature increases. parallel lines. µ is about 1 centipoise for water and about 0. RESERVOIR 2. called Newtonian ﬂuids. Where µ is called ∂y ∂y the coeﬃcient of viscosity with the dimension of (F/L2 )t (ﬁgure 2. in gases. wherein diﬀerential is taken in a direction normal to the interface. Figure 2. These few industrial −1 . The reason may be that. is macroscopically. This is wellknown Newton’s viscosity law.6: Wellordered parallel ﬂow This unit in a system of centimeter (cm)gram (g)second(sec)is g − cm − sec−1 and is called the poise.
v) − p ∂x 3 ∂vy 2 µ(2 − ∇. there are more general relations between the stress ﬁeld and the velocity ﬁeld( namely constitutive law).14) for a newtonian ﬂuid with varying density and viscosity. 2. In addition each normal stress is directly related to pressure p.12. 2. we get: . ∂y In more general ﬂows.2. . .4. Here. Knowing that most Newtonian ﬂuids have ﬂow properties which are independent of direction of coordinates and also considering that for most ﬂuid p = − σ= − 1 (Txx + Tyy + Tzz ). j = x. This is based on assumption that each stress is linearly related through a set of constants to each of the six strains rates . z ²ij =²ji ). that are not described by the Newtonian law of viscosity are referred to nonnewtonian ﬂuids. NAVIERSTOKES AND EULER EQUATIONS 65 important materials. i. we reach the following relationships: 3 Txx = µ(2 Tyy = Tzz = Txy = Txz = Tyz = ∂vx 2 − ∇. (²ij . 2. In this case µ may be described as a function of ∂vx or Txy . The Stoke’s viscosity law degenerates to Newton’s viscosity law for the special case of parallel ﬂow.4 NAVIERSTOKES AND EULER EQUATIONS Substituting the Stoke’s viscosity law into the equations of the linear momentum (equations 2. y. we consider Stoke’s viscosity law. The constants are called viscosity coeﬃcients and ﬂuids behaving accordingly to this relation are called Newtonian ﬂuid.v) − p ∂z 3 ∂vx ∂vy + ) µ( ∂y ∂x ∂vx ∂vz + ) µ( ∂z ∂x ∂vy ∂vz + ) µ( ∂z ∂y This is a common form of Stoke’s viscosity law.13.v) − p ∂y 3 ∂vz 2 µ(2 − ∇.
17) · ¸ · ¸ · ¸ ∂p ∂ ∂vx 2 ∂ ∂vx ∂vy ∂ ∂vz ∂vx − + 2µ − µ(∇. the equations 2. The above equation is called ”NavierStokes equation which ﬁrst developed by Navier in France.v) +Bz − + ∂z ∂x ∂x ∂z ∂y ∂y ∂z ∂z ∂z 3 These equations.18 and 2.17. For constant ρ and constant µ the equation may be simpliﬁed by using the equation of continuity (∇.v) + µ( + ) +By − + ∂y ∂x ∂x ∂y ∂y ∂y 3 ∂z ∂y ∂z ρ Dvz = Dt (2.18) · ¸ · ¸ · ¸ ∂vy ∂vx ∂ ∂vy 2 ∂ ∂vz ∂vy ∂p ∂ µ( + ) + 2µ − µ(∇. µ = µ(ρ).v = 0) to results in: ρ 2 Dv = −∇p + µ∇2 v + B Dt 2 2 ∂ ∂ ∂ The operator ∇2 = ∂x2 + ∂y2 + ∂z2 is called the Laplacian operator. and the boundary and initial conditions determine completely the pressure.19) · ¸ · ¸ · ¸ ∂vz ∂vx ∂ ∂vz ∂vx ∂ ∂vz 2 ∂p ∂ µ( + ) + µ( + ) + 2µ − µ(∇. For inviscid ﬂuids.19 will result in: ρ Dv = −∇p + B Dt This is Euler equation which is widely used for describing ﬂow systems in which viscous eﬀects are relatively unimportant. RESERVOIR ρ Dvx = Dt (2.v) + µ( + ) + µ( + ) +Bx ∂x ∂x ∂x 3 ∂y ∂y ∂x ∂z ∂x ∂z ρ Dvy = Dt (2.66 CHAPTER 2. . 2. along with the equation of continuity. the equation of p = p(ρ). density and velocity component in a ﬂowing isothermal ﬂuid.
the pressure will increase to the left of A − A and this increase in pressure will move in the tube at high speed C. Thus. as shown in ﬁgure 2. Thus we need equation of energy and equation of state(ρ = ρ(p. .7. By tapping the piston we may cause a pressure increase dp on the right of A − A. Due to molecular action. According to Newton’s second law. Sound is the propagation of compressible an expansion wave of ﬁnite but small amplitude such that the ear can detect them. In the incompressible ﬂows. 000 Hertz while the magnitude is typically less than 10 P a. COMPRESSIBLE FLUID 67 2.5. but also because of the wave propagation phenomena that are predominant at ﬂow speed higher than the speed of sound. Consider a long tube ﬁlled with motionless ﬂuid and having a piston at one end. Water needs a pressure change of 20. Fluids of constant density are termed incompressible ﬂuids and assumed during computation the density is constant. A ﬂowing ﬂuid is said to be compressible when appreciable density changes are brought about the motion. ﬂuids undergo very little change in density.2. Up to now. For compressible ﬂow. 000kP a to have 1 percent change in its density. we needed four scalar equations (the equation of continuity and three components of the equation of linear momentum) to describe fully a ﬂow ﬁeld. The variation of density is usually accompanied by temperature changes as well as heat transfer. The general theory of compressible ﬂow is very complicated. not only because of large number of the equations involved. In compressible media a small shift of ﬂuid element will induce similar small movements in adjacent elements and by this a disturbance called an acoustic wave propagates at a relatively high speed through the medium. the density and pressure changes are also accompanied by temperature changes. We thus have a pressure wave of speed V moving to the left due to microscopic action. compressibility means admission of elastic waves having ﬁnite velocity. Frequencies range from 20 to 20. T )). The speed of sound is deﬁned as the rate of propagation of an inﬁnitesimal pressure disturbance(wave) through a continuous medium. The variation of density means that a group of ﬂuid elements can spread out into a larger region of space without requiring a simultaneous shift to be made of all ﬂuid elements in the ﬂuid. Now two things will happen. these propagation have inﬁnite speed which adjustment (disturbance) took place simultaneously through the entire ﬂow and there are now wave to be considered. The second eﬀect is on the macroscopic level.5 COMPRESSIBLE FLUID Despite the existence of large pressure.
68
CHAPTER 2. RESERVOIR
the ﬂuid just to the left of wavefront described above must accelerate as a result of the pressure diﬀerence dp to a velocity dv. Once the pressure rise dp has been established in the ﬂuid, there is no further change in velocity, so it remains at dv. Behind the wavefront, the ﬂuid is thus moving to the left at speed dv. During an interval of dt the wave has progressed a distance Cdt and is shown at position B in ﬁgure 2.7. Meanwhile, ﬂuid particle at A move a distance dvdt to position A1 . At an intermediate position, such as halfway between A and B , shown in diagram as D, the ﬂuid velocity dv has persisted for a time interval dt . Consequently, ﬂuid initially at D has moved 2 a distance dvdt to position D1 . 2
Figure 2.7: Wave front movement and ﬂuid movement By tapping, as inﬁnitesimal pressure disturbance which will move down the tube at a constant speed, the ﬂuid behind the wave is slightly compressed, while the ﬂuid ahead of the wave remains undisturbed. This is an unsteady state problem. However, if we assume an inﬁnitesimal control volume around the wave, travelling with the wave velocity C, we can apply a steady state analysis. The wave is thus stationary while the ﬂuid ﬂows with an approach velocity C, as shown in ﬁgure 2.8. Neglecting friction eﬀects, the velocity proﬁle can be assumed ﬂat. The continuity equation may be written as: ρAC = (ρ + dρ)(C − dv)A Neglecting inﬁnitesimal quantities of higher order (dρdv) gives: Cdρ − ρdv = 0 (2.20)
2.5. COMPRESSIBLE FLUID
69
Figure 2.8: Moving pressure disturbance in a motionless ﬂuid and ﬁxed wave in a moving ﬂuid
70
CHAPTER 2. RESERVOIR
Applying linear momentum balance to the control volume, we get: 0 = C(ρAC) − (C − dv)ρAC + pA − (p + dp)A Which further reduces to: dp = ρCdv By combining equations 2.20 and 2.21 we get: C2 = dp dρ (2.22) (2.21)
This expression gives the velocity of propagation of a compressible wave. Since the propagation of inﬁnitesimal expansion and compression waves is called sound, C is then the speed of sound. the above equation is valid for any continuum, be it a solid, a liquid, or a gas. Because of the high speed of travel of the wave there is very little time for any signiﬁcant heat transfer to take place, so the process is very nearly adiabatic. It should be noted that C is measured relative to the ﬂuid in which the front is propagating. We also assumed a constant value of C, that is we used an inertial control volume with steady ﬂow relative to this control volume. The wave involve inﬁnitesimal pressure variation. Wave with comparatively large pressure variation over a very narrow front are called shock waves. these waves move relative to ﬂuid at speed in excess of the acoustic speed. Then we may consider acoustic wave as limiting cases of shock waves where the change in pressure across wave become inﬁnitesimal. Therefore what we have developed is valid for any weak spherical and cylindrical waves which moves caused by any small disturbance of pressure. If a continuum were incompressible, equation 2.22 would give an inﬁnite speed of sound. However, no actual liquid or solid can be perfectly incompressible. All materials have a ﬁnite speed of sound. For example the speed of sound in water, air, ice and steel at 15◦c and a pressure of 101.325 kP a are 1490,340, 3200 and 5059 m/s, respectively. For liquids and solids it is customary to deﬁne the bulk modulus as a parameter relating the volume (or density) change to the applied pressure change: ∆p ∆p dp K = − ∆V = ∆ρ = ρ dρ V ρ (2.23)
6. ∞ The transition depends partly on the Reynold number( ρvµ x ) where x is the distance downstream from the leading edge.9. In the turbulent region we ﬁnd.6 BOUNDARYLAYER THEORY For ﬂuids with small viscosity such as air and water with a high degree of accuracy. Outside these solid boundary layers the ﬂow may be considered inviscid (µ = 0) and can. According to Prandti’s boundary layer concept viscous eﬀects at high Reynolds number (Re = ρvL = µ Inertia F orces v2 L = µv/ρL2 ) are conﬁned in thin layers adjacent to solid boundV iscous F orces aries. We consider qualitatively the boundary layer ﬂow over a ﬂat plate. 3 By combining equations 2. A transition region is reached where the ﬂow changes from laminar to turbulent. Transition occurs in the range Rex = 3 × 105 to Rex = 106 . that as we get . Within the boundary layer the velocity component in the main ﬂow direction (x) changes from vx = 0 (at the solid boundary) to vx = v∞ (the free stream velocity at the edge of the boundary layer). with a consequent thickening of the boundary layer. In 1904 Ludwing Prandti introduced the concept of boundary layer and showed how NavierStokes equations could be simpliﬁed. we could not properly neglect frictions (Newton’s viscosity law) so we consider these regions apart from the main ﬂow. The bulk modulus related to young’s modulus of elasticity E by the expression: K = 3(1 − 2ν) E Where ν is the Poisson’s ratio. a laminar region begins at the leading edge and grows in thickness. This concept literally revolutionized the science of ﬂuid mechanics. we can consider frictionless ﬂow over entire ﬂuid except for their regions around the contact areas. Here. BOUNDARYLAYER THEORY 71 Water( at 20◦c and atmospheric pressure) has a bulk modulus of about 2. because of high velocity gradients.2.2 × 106 P a while steel about 200 × 106 P a. As it is shown in ﬁgure 2.22 and 2. terming the boundary layers.23 we get: s K C= ρ 2. thus be described by Euler equation. For many common metals such as steel and aluminum ν ≈ 1 and E = K.
72
CHAPTER 2. RESERVOIR
Figure 2.9: Details of boundary layers near the boundary the turbulence becomes suppressed to such a degree that viscous eﬀects predominate, leading us to formulate the concept of a viscous sublayer shown darkened in the diagram. There is actually a smooth variation from boundary layer region to the region of constant velocity. The velocity proﬁle merges smoothly into the mainstream proﬁle. There are several deﬁnitions of boundarylayer thickness that are quite useful. One can consider that the thickness is the distance δ from the wall out to where the ﬂuid velocity is 99 percent of the mainstream velocity. The experimental determination of the boundary layer thickness as deﬁned above is very diﬃcult because the velocity approaches asymptotically the free stream value v∞ . The edge of the boundary layer is poorly deﬁned. For this reason alternative thicknesses which can be measured more accurately are often used. The displacement thickness δ ∗ deﬁned as distance by which the boundary would have to be displaced if the entire ﬂow were imagined to be frictionless and the same mass ﬂow maintains at any section (ﬁgure 2.10). Thus, considering a unit width along the z across an inﬁnite ﬂat plate at zero angle of attack for incompressible ﬂuid:
∞ Z 0 ∞ Z vx dy = q = vm dy δ∗
Hence, changing the lower limit on the second integral and solving for δ ∗ we get:
2.6. BOUNDARYLAYER THEORY
73
∞ Z vx )dy δ ∗ = (1 − vm 0
Figure 2.10: Displacement thickness in boundary layer The motivation for using displacement thickness is to permit the use of a displaced body in place of the actual body, such that the frictionless mass ﬂow around the displaced body is the same as the actual mass ﬂow around the real body. Use is made of the displacement thickness in the design of wind tunnels, air intakes for airplane jet engine. As the boundary layer thickness growth, the ﬂuid is pushed away from the plate, which means that the velocity has also a component in the y direction , vy . we would expect that vy ¿ vx because the boundary layer thickness is small compared to any signiﬁcant body dimension. The boundary layer thickness, that is the value of y for which vx = 0.99v∞ , can be obtained as following: r µx 5x =√ δ ≈ 5.0 ρvm Rex A rough comparison between the various thicknesses gives: δ δ∗ = 3 The eﬀects of viscosity is inﬂuential as δ ∗ increases. In the case of concrete dams since δ ∗ /h ratio is very small, the viscous eﬀect can be neglected, specially for the sudden ground acceleration on the base of the reservoir.
74
CHAPTER 2. RESERVOIR
2.7
IRROTATIONAL FLOW
A ﬂuid element can be subjected to three types of ﬂow, namely: translation, rotation and deformation. These types of elemental motions are depicted in ﬁgure 2.11. The concept of translational motion is selfevident.. The deformational motion will occur when relative orientation of the axis changes as shown in ﬁgure 2.11. Deformational ﬂow also exists if one or more axis are stretched or compressed. Here we focus our attention on rotational ﬂow, which is depicted by the ﬂuid element of ﬁgure 2.11 turning around.
Figure 2.11: Three types of ﬂuid motion Let us examine the ﬂow in a ﬂuid with circular streamlines as shown in ﬁgure 2.12. The ﬂuid rotates like a rigid body. each element turns around at a certain angular velocity. The arrows shown rotate at the same rate. This is unquestionably a case of rotational ﬂow. Now let us consider the ﬂow between the two horizontal ﬂat plates, where the bottom plate is stationary while the top one moves at velocity v0 as shown in ﬁgure 2.13. we note that the horizontal arrow is simply translated, while
2.12: Fluid rotating like a rigid body Figure 2. IRROTATIONAL FLOW 75 Figure 2.13: Shearing ﬂow between two ﬂat plates .7.
14: Change of relative positions in an arbitrary ﬂow ﬁeld We deﬁne the angular velocity ωz about the axis z as the average rate of counterclockwise rotation of the two lines: 1 dα dβ − ) ωz = ( 2 dt dt In the limit of small angles we would have: d( ∆yα ) d( ∆yα ) ∂vy dα ∆x ∆t = = = dt dt dx ∂x ∆x d( ∆yβ ) d( ∆tβ ) dβ ∂vx = = = dt dt dy ∂y ∆x . if not the ﬂow is rotational. For generality we refer to ﬁgure 2. To determine that we propose to use the average rate at rotation of the two arrows as a measure. Figure 2. It is not clear wether this is a case of rotational ﬂow or not. RESERVOIR the vertical one turns. If the average rate of rotation is zero the ﬂow is said to be irrotational.14.76 CHAPTER 2.
then: ζ =2ω = ∇ × v The vector ω is thus onehalf of the vorticity vector:The vector ω is thus onehalf of the vorticity vector: i j k 1 1 ∂ ∂ ∂ ω = ∇ × v = ∂x ∂y ∂z 2 2 vx vy vz Referring back to simple ﬂow conﬁguration of ﬁgure 2. It represent the rigid body angular velocity ω z of the line segments about the z axis. Rotational ﬂows are those where ω 6= 0 at points in the ﬂow. . the ﬂow is rotational.2.13 we note that vy = vz = 0 and vx = v0 y .7. 1 ∂vy ∂vx − ) ωz = ( 2 ∂x ∂y Similarly: 1 ∂vz ∂vy − ) ωx = ( 2 ∂y ∂z 1 ∂vx ∂vz ωy = ( − ) 2 ∂z ∂x 77 To avoid the factor 1 we deﬁne the vorticity ζ as equal to twice the 2 rotational vector ω. then we have: 1 1 ω z = (0 − v0 ) = − v0 2 2 Thus. Thus the expression 1 ( ∂vy − ∂vx ) is actually more than just 2 ∂x ∂y average rotation of the line segments about the z axis. IRROTATIONAL FLOW Thus. What’s left of the relative motion must be rigid body motion.14. At this time we deﬁne irrotational ﬂows as those where ω = 0 at each point in the ﬂow. The normal strain rates give the rate of stretching or shrinking of the two lines shown in ﬁgure 2. while the shear strain rates give rate of change of angularity of the two lines.
if we divide the pressure into static and dynamic pressure we can write for static pressure −∇ps + B =0. Thus the shear stress(Newton’sss viscosity law) in such ﬂows and in more general ﬂows will depend on the viscosity of the ﬂuid and the rate of variation of the velocity in region. . therefore for only hydrodynamic pressure we can write the following: ρ Dv = −∇p Dt The p is the dynamic pressure in excess of static pressure. In the Euler equation.78 CHAPTER 2. For ﬂuids with small viscosity such as water and air. .8 RESERVOIR’S EQUATION OF MOTION Based on the information presented in the previous sections. Irrotational analysis may be carried out if the boundary layer thickness is small in comparison with the scale of the ﬂuid. assumption of irrotational ﬂow is valid for great part of the ﬂow except region with large velocity gradient. we deﬁne the following: ρ p = −K(²xx + ²yy + ²zz ) . thus we can write: ∂v = −∇p ∂t For a linearly compressible ﬂuid. 2.v) v + ρ ∂v = −∇p ∂t For small amplitude motion the eﬀect of convective term of the left hand side of the above equation is negligible.25) Derivating of whole equation with respect to time: The ²xx can be written as following: . The above equation can be written as: ρ(∇.24) (2. the Euler equation is valid for the case of dam’s reservoir. RESERVOIR A development of rotation in a ﬂuid particle in an initially irrotational ﬂow would require shear stress to be present on the particle surface. p= −K(²xx + ²yy + ²zz ) . (2. . Thus the boundary layer regions are the places that element of ﬂuid rotates.
24 to take the divergence of the whole equation we get: ∇.26 and substitute it in equation 2. C = ∞and then we have: ∇2 p = 0 The above equation is the Laplace equation of motion for incompressible ﬂuid. in which is the velocity of pressure wave in ﬂuid. Thus we can write: p= −K(∇. equation 2. therefore.26) If we use equation 2.9 RESERVOIR BOUNDARY CONDITIONS The formulation of the boundary conditions associated with the reservoir boundaries is simply the expression in mathematical terms of the physical situations. µ = 0 b)small amplitude motion.28) 2 K ∂t pρ We deﬁne c = K .25 2. RESERVOIR BOUNDARY CONDITIONS 79 ∂ ∂ (²xx ) = ²xx = ∂t ∂t .v) ∂v ) = ∇.27) ρ ∂ 2p = ∇2 p (2.(ρ Or we can write the following: ∂ (∇.2. The equation of motion for reservoir (equation 2. For incompressible ﬂuid (K = ∞).v) = −∇2 p ∂t Using equation 2.28) obtained based on the following assumption: a) inviscid ﬂuid. convective terms are neglected c)linear compressible ﬂuid.9. There are an inﬁnite number of solutions to the diﬀerential . µ ∂ux ∂x ¶ ∂ = ∂x µ ∂ux ∂t ¶ = ∂vx ∂x In which ux is the displacement in the xdirection.( − ∇p) ∂t .27 we get: ρ (2. (2.
reservoirfoundation. This results into the condition that at the normal to the boundary there is no relative velocity or another word we can write it mathematically: s vn = vn in which n is the unit normal vector to the boundary at the damreservoir s interface and vn and vn are the velocity of the structure (dam) and ﬂuid along the n respectively (ﬁgure 2.1 DamReservoir Boundary Condition At the surface of ﬂuidstructure.9.80 CHAPTER 2. This temporal condition is called an ”initial condition”. The task is to ﬁnd the solution that is relevant to the boundary condition.15: Boundaries of the damreservoir system 2. It should be noted that in addition to the spatial (or geometric) boundary conditions. RESERVOIR equation.16). it is clear that there must be no ﬂow across the interface. free surface. The above equation can be rewritten as following: s vn = v.15). In case of damreservoir we have four spatial (geometric) boundary conditions namely. damreservoir. and reservoir farend boundary conditions (ﬁgure 2.n . there are temporal boundary conditions which specify the state of the variable of the interest at some points at time. This is based on the fact that face of the concrete dams are impermeable. Figure 2.
n n ρ or: ∂p = −ρas n ∂n Figure 2.2 ReservoirFoundation Boundary Condition If there is no absorption or penetration of water into the reservoir bottom.2.9.n ∂t ∂t 81 It must be noted that n is normal to the surface and it direction is constant ass well as its value at a point. the same damreservoir boundary condition that was obtained previously.9.16: Damreservoir interface 2. can be . RESERVOIR BOUNDARY CONDITIONS Diﬀerentiating the above equation with respect to time will give: s ∂vn ∂v = . From the Euler equation and knowing that ∂vn is normal acceleration of the dam at the interface we can write: ∂t 1 as = − ∇p.
= −ρ an (t)+ κ (0. we can account it in a very simpliﬁed manner. t) in the water is governed by the onedimensional wave equation: ∂ 2p 1 ∂ 2p = 2 2 ∂n2 C ∂t n≥0 (2. t) = Cr gr (Cr t) and κ (0.82 CHAPTER 2.. t) is the acceleration of the reservoir bottom due to interaction between the impounded water and the reservoir bottom materials. Er is the Young’s modulus of elasticity and ρr is the density ρr of the reservoir bottom materials. the D’alembert solution to equation 2. t) ∂n (2. t) in the layer of reservoir bottom materials is governed by: 1 ∂ 2κ ∂ 2κ = 2 2 ∂n2 Cr ∂t n≤0 (2. Here. At the reservoir bottom.30 is κ = gr (n + Cr t) where gr is the wavefront of the refracted wave propagating vertically downward in the reservoir bottom materials.29) Similarly. t) . Equilibrium at the surface of the reservoir bottom materials requires that: p(0.30) Where Cr = Er . An upward propagating wave does not exist because of the radiation condition for the assumed inﬁnitely thick layer of . the interaction displacement κ(n. the accelerative boundary condition states that the normal pressure gradient is proportional to the total acceleration: ¤ £ ∂p(0. t) ∂y (2. 0 reservoir bottom materials. t) = −Er ∂κ (0. The hydrodynamic pressure p(n..31) q Where κ (0. RESERVOIR used for reservoirfoundation boundary. we consider only interaction in the normal direction due to the assumption that the hydrodynamic pressure waves incident on the reservoir bottom only excite vertically propagating dilatational waves in the reservoir bottom materials.. t) = ∂t . In case of reservoir with sediment at the bottom. Note that ∂κ (0.32) . The boundary condition at the reservoir bottom relates the hydrodynamic pressure to the sum of the normal acceleration and acceleration due to interaction between impounded water and the reservoir bottom material.
gives: C r r ∂p(0.33. (2. In general.2.34) ∂n ∂t Where q = ρ ρ r .. RESERVOIR BOUNDARY CONDITIONS 00 83 2 Cr gr (Cr t). so rC the second term on the right hand side of the above equation is zero giving the boundary condition for a rigid reservoir bottom. The wave reﬂection coeﬃcient κ can be obtained by considering the reﬂection of a harmonic pressure wave in the impounded water impinging vertically on the reservoir bottom. t) κ (0. while the resulting upward £ ω ¤ propagating wave is pr = κ exp i C (−n + Ct) . t) ∂p(0. when substituted into equation C 2. where the prime indicates the derivative of gr with respect to argument (n + Cr t). Cr = ∞ and q = 0. ∂p(0.32 with respect to t gives: ∂p(0. Diﬀerentiating equation 2. The wave reﬂection coeﬃcient may range within the limiting . t) 00 = Er Cr gr (Cr t) ∂t or: Er . The downward vertically propagating incident £ ω ¤ wave of unit amplitude is pi = exp i C (n + Ct) .33) The solution for κ (0.29) The substitution of the sum pi + pr of the two hydrodynamic pressure waves into the boundary condition at the reservoir bottom. which is the ratio of the amplitude of the reﬂected hydrodynamic pressure wave to the amplitude of a vertically propagating pressure wave incident on the reservoir bottom can be related to the damping coeﬃcient.34. The fundamental parameter that characterizes the eﬀects of absorption of hydrodynamic pressure waves at the reservoir bottom is the admittance or damping coeﬃcient q.31 with the identity Er = ρ 1 r . with no normal acceleration results in an equation which can be solved to results in: κ= 1 − qC 1 + qC The wave reﬂection coeﬃcient is independent of excitation frequency ω. The wave reﬂection coeﬃcient κ. For rigid reservoir bottom materials. it depends on the angle of incident of the pressure wave at the reservoir bottom. t) −q = −ρan (t) (2.9.. Both pi and pr satisfy the wave equation (equation 2. t) =− ∂t Cr . t) from equation 2. equation 2.
t) = 0 (2. resulting in κ = 1. y. In other word. RESERVOIR values of 1 and −1.35) where ϕ is the displacement of the free surface above the horizontal plane. In the case of surface wave of negligible surface tension. as would the water surface.84 CHAPTER 2.3 Free Surface Boundary Condition The geometry of free surface boundary is not known a priori. then the total derivative of the surface with respect to time would be zero on the surface. t) = 0 . Cr approaches zero and q = ∞. 2. ∂ϕ ∂F ∂F ∂F DF (x. For rigid reservoir bottom Cr = ∞ and q = 0. t) = z − ϕ(x. which means we have a very diﬃcult boundary condition to cope with.17: Free surface wave say z = 0. y. z. z. y. This shape is part of the solution. t) =0= +vx +vy +vz Dt ∂t ∂x ∂y ∂z Or: on F (x. z. Figure 2. y. resulting in κ = −1. z. The material properties of the reservoir bottom medium are highly variable and depend upon many factors. If the surface varies with time. For very soft material. It is believed the κ values from 1 to 0 would cover the wide range of materials encountered at the bottom of actual reservoirs. it does not changes.9. if we move with the surface. we call them gravity wave.. The free surface of a wave can be described as: F (x.
2.35 is: v. z.37 gives: 1 ∂p = vz ρg ∂t ∂ϕ ∂t at z = 0 at z = 0 (2.36 with respect to t and cancelling ϕ with equation 2.36) (2. Rearranging the above equation.37) .n = q¡ ¢ ∂ϕ 2 ∂x ¡ ∂ϕ ¢2 ∂z on z = ϕ(x.38) ∂t ρ ∂z Diﬀerentiating equation 2.n ∇F  ∂t ∇F Where the unit vector normal to the surface has been introduced as n = ∇F  . t) ∂ϕ ∂ϕ ∂ϕ + vx + vz ∂t ∂x ∂z Neglecting the convection terms and approximating pressure we get: p = ρgϕ vz = We also know: ∂vz 1 ∂p =− +g (2. t) +1 + Where: Carrying out the dot product: vz = − ∂ϕ i− ∂ϕ j + k q¡ ∂x ∂y ¢ n= ¡ ¢ 2 ∂ϕ 2 + ∂ϕ + 1 ∂x ∂z on z = ϕ(x. t) = 0 The boundary condition at the free surface using equation 2. we get: − ∂F v.∇F = v. RESERVOIR BOUNDARY CONDITIONS 85 − ∂F = v. y.9.n = ∂t ∇F  ∂F ∂t on F (x.
In the solution.38 then substitution of ptotal = p + ρgϕ. we get: 1 ∂ 2 p ∂p =0 + g ∂t2 ∂z at z = 0 This is an approximate to the surface boundary condition. ux = ag cos( 2πt ) at x = 0 T p = 0when x → ∞ . The above boundary condition is usually replaced with the boundary condition that: p=0 at z = 0 This assumption of no surface wave is common assumption in concrete dams and is valid. 2.10 SOLUTION OF THE RESERVOIR EQUATION Westergard’s classic work (1933) on the hydrodynamic water pressure on dams during the earthquake started a new area for the researcher in this ﬁeld. Westergaard’s solution to wave equation for rigid dams during earthquakes was obtained based on the assumptions of dam with rectangular reservoir subjected to horizontal earthquake. The more complicated free surface boundary condition can be established in which we can assume pressure distribution due to interaction of wind and surface.. These cases are above the scope of the book and can be found in appropriate references of water wave mechanics. RESERVOIR A second diﬀerentiation with respect to t and elimination of vz with equation 2. The pressure equation for the system shown in ﬁgure 2. the reservoir extended to inﬁnity in the upstream direction and the eﬀect of surface wave was neglected. in which p represent the excess pressure due to motion and ρgϕstands for hydrostatic pressure.86 CHAPTER 2. The system was subjected to the horizontal ground acceleration which was perpendicular to the dam axis. The motion was assumed to be small amplitude motion and water was taken inviscid ﬂuid.18 is found under the following boundary conditions: p = 0 at y = h uy = 0 at y = 0 . It was shown that the surface waves are negligible.
5. . y.. t) = ) ) cos( e sin( π2 T n=1. T. and can be written as: 8aρgh pmax (x.3..3... The shape of the pressure diagram is such that the curve has a horizontal tangent at the top and a vertical tangent at the bottom. µ ¶ 1 nπ(h − y) ) sin( n2 cn 2h (2.39) cn = 16ρh2 1− 2 = n KT 2 1− The maximum pressure at x = 0 occurs when t = 0...18: Rigid daminﬁnite reservoir system The hydrodynamic pressure for the above mentioned boundary conditions is as follows: ¶ ∞ X µ 1 8aρgh 2πt nπ(h − y) −qn p(x..2. n2 cn 2h in equation 2. 2T.40) The maximum pressure at a given time on the upstream face (x = 0) happens at the bottom of the reservoir.10. . y. SOLUTION OF THE RESERVOIR EQUATION 87 Figure 2. t) = π2 ∞ X n=1..39 qn and cn are deﬁned as followings: qn = r nπcn x 2h r 16h2 n2 C 2 T 2 (2.5.
Also we imposed the boundary condition at farend of the reservoir that as x → ∞ then p → 0.03125ton − f t3 the width of water along the xdirection is as follows(ﬁgure 2. 2. is greater than the . Tm . the westergaard’s solution is valid if the period of loading. Therefore. The maximum hydrodynamic pressure to be added to the hydrostatic pressure is as follows: p 0. The Westergaard solution is valid if the value of cn is real.. 16h2 ≥0 n2 C 2 T 2 4h The term.f t the width of concrete would be: p b0 = 0.72( 1000T f t ) The mass of water that moves with the dam body can be transferred in to extra mass of concrete attached to the dam body. we need to ﬁnd the mass of water that attached to the dam and moves with it. (2m−1)c is the mth period of the reservoir..38 h(h − y) As it was already mentioned the Westergaard approach gives a pressure distribution which is in phase with ground acceleration. Another word.39) has same phase compared to ground acceleration. the maximum pressure (equation 2. For a dam concrete with the unit mass of 144lb. Thus. m = 1.0255 ton − f t3 a h(h − y) q pmax (0. t) = h sec 1 − 0.40 ) can be replaced by a parabolic curve. 3.. RESERVOIR To further simplify the hydrodynamic pressure on the dam. one may consider the hydrodynamic pressure as a certain body of water moves with dam.percu.88 CHAPTER 2. This is called as added mass approach. T . .19): b= 7p h(h − y) 8 In an approximation. This is possible because of the fact that the hydrodynamic pressure obtained (equation 2. we can write: 1− This will result in: T > 4h nc n = 2m − 1. This boundary condition is an indication of energy dissipation at the farend. For water with unit weight of 0. y.
19: Added mass approach .10. SOLUTION OF THE RESERVOIR EQUATION 89 Figure 2.2.
In his solution for the horizontal earthquake. . Another solution to the wave equation presented by Chopra. If T = Tm . Chopra (1967) presented a closed form solution for hydrodynamic pressure in the case of rigid dam with the vertical upstream face under the horizontal and vertical ground motion. They can be written as follows: λm = ωm C ω= 2π T ωm = 2π (2m − 1)πC = Tm 2h m = 1. The Westergaard pressure equation at the upstream face of the dam can be written as followings: ∞ m−1 X 4aρg (−1) q p(0. then p = ∞ and we have c resonance.. Also it was found that the response to vertical ground motion is a real valued function and there is no decay with time so that the system is truly undamped in this case. For the excitation frequency less than the fundamental frequency of the reservoir. y... RESERVOIR ﬁrst period of the reservoir. t) = cos λm y cos ωt π 2 ω2 m=1 (2m − 1) λm − 2 C in which ω m and Tn are the circular frequency and period of the mth mode of the reservoir and ω is the circular frequency of the earthquake. T1 = 4h . The greater excitation frequency causes an outofphase pressure compared to Westergard’s solution.. The hydrodynamic pressure for a harmonic ground excitation of ug = eiwt would be: .. In this case the hydrodynamic pressures cannot be represented by inertia eﬀect on an added mass moving with the dam..90 CHAPTER 2. 2. 3. ∂p = −ρ ug at x = 0 ∂x . the wave equation for a rigid dam with rectangular reservoir (ﬁgure 2.18)presented subjected to the following boundary conditions: p = 0 at y = h ∂p = 0 at y = 0 ∂y . both Westergaard and Chopra’s solution are the same.
then the term involving sin ωt vanishes and the solution of the wave equation is same as Westergaard’s solution.41) and would be as: p(0. SOLUTION OF THE RESERVOIR EQUATION 91 (2. Therefore. For m1 > 1 we have pressure at inﬁnite distance. t) = π m=1 (2m − 1) λ2 − m ω2 C2 ω2 exp −x λ2 − 2 m C Ã r ! cos λm y eiwt ω2 C2 cos λm y ω in which m1 is the minimum value of m such that λm > C . For ω > ω 1 the term involving sin ωt does not vanish and represent existence of pressure at inﬁnite distance up the reservoir. (ug = ag cos( 2πt )).. the hydrodynamic pressure cannot be represented by added mass approach. To compare the above solution with Westergaard’s solution.10. The pressure response has two terms one is in phase with ground motion and the other is in opposite phase of ground motion. t) = m1 −1 X (−1)m−1 4aρg q cos λm y + sin ωt π 2 ω2 m=1 (2m − 1) − λm C2 ∞ X (−1)m−1 4aρg q cos ωt π 2 m=m1 (2m − 1) λm − ∞ (−1)m−1 4ρ X q p(x. y.2.41) It can be seen that as x → ∞ the pressure does not approach zero if ω λm < C which was the fourth boundary condition imposed in Westergaard solution. This means that for the case of ω m < ω < ω m+1 the pressure would be zero at x = ∞ for ωm+1 and higher modes. . y. again it can be noticed that the Westergaard’s solution is valid if the frequency of the excitation is less than the fundamental frequency of the reservoir system. While for ωm and lower modes we expect to have pressure at very farend of the reservoir. If m1 = 1 which ω is an indication of λ1 > C or ω < ω 1 . we can consider the real part of the ground motion as well as coeﬃcient ag .The Solution in this case consist of the real part of the T pressure solution (equation 2. Here.
For compressible ﬂuid. However. g The magnitude of ωC for harmonics of signiﬁcance is therefore considerably g less than 6. If we ignore the free surface wave and change the second boundary condition by: p = 0 at y = h then the solution would be: ω agρC sin C (h − y) p(x. such low ωC frequency harmonics contribute little to the response in the case of typical reservoir depths encountered.. Therefore.2 h T . y. except possibly near the free surface. varies as follows: √ h e < 0. RESERVOIR The solution of wave equation subjected to the vertical ground motion for a rigid dam with rectangular reservoir is investigated. Thus.82 × 10−3 (ω = 1. Most investigator have ignored the waves that may be generated at the free surface of water. The reservoir have the following boundary conditions: .82×10−3 . y.72 radians per sec. t) = g ω ω agρC sin C (h − y) − ωC cos C (h − y) cos(ωt) g ω ω ω cos C h + ωC sin C h ∂p 1 ∂2p + ∂y g ∂t2 ∂p = 0 at ∂x The solution is independent of the x−coordinate.42 is suﬃciently accurate. C = 4720f ps). ∂p = −ρ uy at y = 0 ∂y = 0 at y = h x=0 The pressure solution when the reservoir is subjected to the vertical .42) The diﬀerence between the two solutions depends on the magnitude of the g quantity ωC .92 CHAPTER 2. and harmonic excitation it was found that the error .. because natural frequencies for the reservoir are rather high: e.05 if > 4.e . the largest value of g may be taken as 6. ground motion of uy = ag cos(ωt) would be as follows: p(x. Analysis of strong ground motions of the past earthquakes seems to indicate that frequencies of most of the signiﬁcant harmonic components lie in the range 1 < ω < 120 rad/sec. and errors introduced by dropping terms involving ωC will be small. equation2.g. t) = cos(ωt) ω ω cos C h (2. introduced by ignoring surface wave. for a 300 ft deep reservoir ω1 = 24.
5m. Similar conclusions can be derived for other depth between 100 ft and 520 ft. e < 0. Because the natural periods of the system are very small (ﬁrst natural period for 100 ft depth of water is 0.05. Hence.05 if T < 1. On this basis. the error e < 0. the contributions to the total response from harmonic components of ground motion with the periods longer than 1.085 sec). Under this condition .6 h < < 4.05 < e < 0.20 if √ √ h 2.05 if h > 158. This is particularly so because we are dealing with an ﬂuid. Thus for T = 3 sec. 2.43) ∂x ∂n C ∂t in which n is the normal at the truncation boundary.20 if Most of signiﬁcant harmonic in typical strong ground motion have a period below 3 sec.11.11 RESERVOIR FAREND TRUNCATED BOUNDARY CONDITION The truncated boundary for the ﬁnite element and boundary element modeling of the inﬁnite reservoir was worked by so many researchers. A plane wave can be represented by an equation of the form: p = F (x − Ct) This equation leads to the following condition: ∂p ∂p 1 ∂p = =− (2.2 h T √ h < 2. it seems that the eﬀect of surface waves can be ignored with little loss of accuracy. It introduces a damping in the system and models the loss of energy in the outgoing waves. RESERVOIR FAREND TRUNCATED BOUNDARY CONDITION93 0.6 h T e < 0. This represents wellknown Sommerfeld radiation condition.2. Sommerfeld boundary condition is the most common one that is based on the assumption that at long distance from the dam face.32 sec .32 sec will be small. water wave can be considered as plane wave. For a reservoir depth 100 ft. it may be concluded that errors introduced by neglecting surface wave is on the order of 0.
the classical solution for the hydrodynamic pressure was obtained in which there are inphase and outofphase pressures. The solution to the wave equation can be investigated for a rigid dam with rectangular reservoir subjected to horizontal ground motion and the following boundary conditions: p = 0 at y = h ∂p = 0 at y = 0 ∂y . RESERVOIR for the far end reservoir.44) p(x.i = C qm + i C (2..41 when ug = eiωt can be represented as following: m1 −1 ∞ X Am X A q m p(x. Their radiation condition derived only for the horizontal ground motion in the case of dam with vertical upstream face.94 CHAPTER 2. y. y. As a rule for the truncated boundary there is no reﬂection for the outgoing wave and all of our eﬀort is in modeling the energy loss in outgoing wave such that all energy can be absorbed on the truncated boundaries. The hydrodynamic pressure for a harmonic ground excitation of ug = eiwt would be: ∞ 4ρ X (−1)m−1 eqm (L−x) + Zm e−qm (L−x) cos λm y eiwt (2. Zm = −1 ω . The .45) It was found that the Sommerfeld radiation damping condition does not yield good approximation for the inﬁnite reservoir for excitation frequencies between the ﬁrst and second natural frequencies of the reservoir.. Within this range of frequencies an increase in the length of reservoir does not lead to any increase in the accuracy(Humar and Roufaiel. 1983). t) = π m=1 (2m − 1)qm eqm L − Zm e−qm L Where: qm = r λ2 m ω √ qm − i C ω2 − 2 . Humar and Roufaiel (1983) used a radiation condition which adequately models the loss over a wide range of excitation frequencies. ∂p = −ρ u at x = 0 ∂x ∂p ∂p 1 = ∂n = − C ∂p at x = L ∂x ∂t . t) = e−qn x cos λm y + cos λm y eiωt 2 qn m=m m=m λ2 − ω 1 1 m C2 .. also it was shown that the conditions give much better results as compared to the plane wave Sommerfeld condition. Solution of the equation 2.
47) ∂p 1 =− ∂n C ω1 ´2 1− ω ω > ω1 (2.44.11.46) For the case of m = 2 (when ω lies between ω 1 and ω 2 ) equation 2.2. y.48 and 2.49) Apparently. t) = Am e cos λm y eiωt = m1 −1 X m=1 pm (2. For large value of ω equation 2.47 becomes: ! Ãr ³ ω ´2 ∂p 1 ∂p 1 =− 1− ∂x C ω ∂t Therefore. RESERVOIR FAREND TRUNCATED BOUNDARY CONDITION95 It was already shown that at large distance from the dam.49 reduces to equation 2.49 will be given by an expression similar to equation 2.43. and the pressure is given as: m1 −1 X m=1 ¶ µq 2 − λ2 − ω 2 x m C p(x. except that Zm will now be as follows: Zm = 1 q ¡ ¢2 qm − 1 − ω1 ω q Zm = ¡ ω1 ¢2 ω qm + i C 1 − ω ω iC ω < ω1 ω > ω1 . the second term in the equation vanishes.48) From the above equation we can write: Ãr ! m1 −1 ω2 ∂p ∂p 1 X 2 − λm pm p =− 2 ∂x ω m=1 C ∂t (2.46 gives p = p1 and equation 2. The hydrodynamic pressure obtained by applying the modiﬁed boundary condition at x = L represented by equations 2. it can be said that the following condition should be applied at the truncated boundary: ∂p =0 ∂n Ãr ³ ω < ω1 ! ∂p ∂t (2. the preceding condition is not exact when ω is greater than ω 2 .
It was found that the modiﬁed radiation boundary condition gives much better results as compared to the Sommerfeld one for excitation frequencies between 0. the governing equation of the pressure would be the Laplace equation. Sharan (1985) used the radiation condition for the submerged structure surrounded by unbounded extent of compressible ﬂuid as following: ∂p = −Ψp ∂n Deferent geometry of the solidﬂuid interface was used to ﬁnd the accuracy of the proposed radiation boundary condition. A large extent of the ﬂuid domain is required to be included in the analysis. RESERVOIR The modiﬁed boundary condition at the far end was obtained with ignoring the second term in the pressure equation at the large distance from the dam. ∂p Sharan (1984) showed that the condition of ∂n = 0 for incompressible ﬂuid at the far end is the form of that for a rigid stationary boundary and the behavior of the ﬂuid motion at the truncated boundary is not truly presented. The structure submerged in unbounded ﬂuid in the upstream direction.0 and 2. the results for an inclined face are relatively more accurate than those for a vertical face. The above condition is same as Sommerfeld condition for the incompressible ﬂuid (C = ∞). Under the assumption of incompressible ﬂuid in the rectangular reservoir and rigid dam. Knowing the pressure equation.96 CHAPTER 2. Sharan (1987) proposed a damper radiation boundary condition for the time domain analysis of the compressible ﬂuid with small amplitude.0 ∂p therefore ∂n = 0 which is same condition at the inﬁnity for the inﬁnite reservoir. He found that under the assumption of the rigid dam and rigid rectangular reservoir bottom. he found that at large distance from upstream face condition would be: ∂p π =− p ∂n 2h It can be observed that at very large distance away from the dam face p = 0. the truncated boundary can be located very close to the structure. It was found that although the proposed boundary condition was derived for the dam with the vertical upstream face. The proposed radiation condition was: ∂p π 1 ∂p =− p− ∂n 2h C ∂t . for the horizontal vibration.
β = 0 and α = 1 − ( 2Ω )2 . therefore Sommerfeld radiation 2 ∂t boundary condition is not justiﬁed for Ω < π . For the limiting case ∂p → 0 when T → ∞. the magnitude of . As a general. For these range of frequencies 2 ∂p x values of p. It is not an important problem because in high frequency excitation the hydrodynamic forces are not comparable so that its error in hydrodynamic pressure is not signiﬁcant.2.and ∂x approach zeroes as the ratio h is increases. Sharan (1985) expressed the condition at the truncated face as: ∂p α ∂p βω =− + p ∂x C ∂t C For (Ω = ωh ) less than 3π which is for excitation frequency less than the C 2 x second natural frequency of the reservoir if h is suﬃciently large. pC . For Ω > 3π even for large value of h . it was found that proposed boundary condition at the truncated surface is more eﬃcient and needs less C computational time. This ∂t explains why satisfactory results may be obtained even with use of improper ∂p ∂p 1 boundary condition such as ∂n = 0 and ∂n = − C ∂p by considering a very ∂t x large extent of the reservoir. for Th À 4. RESERVOIR FAREND TRUNCATED BOUNDARY CONDITION97 The above boundary condition was found to be very eﬀective and eﬃcient for a wide range of the excitation frequency. In the case of Th = 1 (high excitation frequency) both methods shows discrepancy with the case of L = ∞. In ∂t this case approximated boundary condition for the truncated surface may be ∂p written as ∂n = 0 . the magnitude of pand ∂p become inﬁnitely large. then α = 0 p p π π and β = − ( 2Ω )2 − 1 and for π < Ω < 3π . . If the value of T be near the natural period of vibration of the ﬂuid ω C domain ( Th → 4).11. one h ∂t of the terms on the right hand side of proposed equation of damping would be small compare to the other. Thus in both case. In the case of Th ¿ 4. 2 2 ∂p For Ω < π . becomes small compared to that of ∂p . ∂n is function of p instead of ∂p . For the ﬁnite value of c and h. Therefore for such a case. the eﬀectiveness of the proposed damper depends on the period of excitation T . On the other hand. For the limiting case p → 0 h ∂t C when T → 0. neither Sommerfeld damper nor C the proposed one would be eﬀective. ∂p . ( 2π ). the 2 parameters α and β are sensitive to x and y coordinate point on the truncated surface. ∂p is much less than that ∂t of pC .
RESERVOIR .98 CHAPTER 2.
3. if we try to apply a dynamic force f (t). the equation of the motion can be written as: 99 . x or θ ) is required to deﬁne the position of the particle at each second.1. In mathematical modelling. Having formulated the equation of the motion. we then procced to solve the equation of the motion.1 FINITE ELEMENT MODELLING OF THE STRUCTURE SingleDegreeOfFreedom Systems A singledegreeoffreedom system (SDF) is deﬁned as a system whose motion can deﬁned by a single parameter. Here we will ﬁnd the structure’s equation of the motion and its ﬁnite element equation in addition to reservoir ﬁnite element equation. we need to formulate the equation of the motion.1 are systems of SDF. In previous Chapter. reservoir’s equation of the motion was formulated. For a system shown in 3.1 3. As shown in the ﬁgure only one parameter (u.Chapter 3 FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM The ﬁrst step in dynamic analysis of a system is to properly model the actual system.1−a. Systems shown in ﬁgure 3.
.. in ﬁgure3. FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM Figure 3.2 u is the acceleration of the mass M relative to the base and ug is the ground acceleration. Thus the equation of the motion for a SDF system can be written as: M(u + ug ) + C u +Ku = f (t) This is a diﬀerential equation of second order. . . The inertial force is equivalent to the total acceleration of the . Other form of spring may be proposed which result in nonlinear equation.. FD and FS are inertial. the inertial force can be written as M(u + ug )..100CHAPTER 3. If there is .. damping and spring (stiﬀness) forces..1: Systems of single degree of freedom f (t) = FI + FD + FS in which FI . . For high velocity. a base motion. The equation can be rewritten into the following form: M u +C u +Ku = f (t) − M ug . . As is shown . The damping force is viscous damping force . .The spring force can be written as Ku. respectively. . mass M and can be written as M u when there is no base motion. This is a valid assumption when velocity is small.. damping force is proportional to the square of the velocity. .. which is a linear spring or linear stiﬀness. and is given by FD = C u which C is the damping coeﬃcient.. .
the base motion can be interpreted as a force acting on the mass in opposite direction of its motion.2: Forces on a single degree of freedom It can be seen that the base acceleration is equivallent of a force acting on the mass in the opposite direction of its motion.2 MultiDegreeOfFreedom System In a multidegreeoffreedom (MDF) sytem.3.1. it is required to know the values of the displacement in more than one point in order to deﬁne its motion at any instant of time.1.3: An example of multidegreeoffreedom (MDF) sytem with degrees of freedom in y direction . 3. FINITE ELEMENT MODELLING OF THE STRUCTURE 101 Figure 3. Consider a system with n degrees of freedom which accelerate in the y direction as shown in ﬁgure 3. Figure 3. In the other word.3.
........ .................... = FD = .... ..................... ........................ FD2 C21 C21 ........... . ..... FIn ...... FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM The inertial forces can be written as: M1 ......... ..................... FI(n−1) ....... = FI = ..................................... ....... ...... C11 C12 ........ un FDn Cn1 Cn2 ...102CHAPTER 3.....K2n u2 ....... FSn Kn1 Kn2 ... u2 ..C1n u1 FD1 ...............M2 ....... un−1 ...... The forces due to elasticity can be written as: K11 K12 .............. u1 .. ...MIn or it can be write as: FI = [M]{U }t The mass matrix [M] is a diagonal matrix and in general it can be a nondiagonal symmetric matrix............... ....K1n u1 FS1 FS2 K21 K21 ...... un g or it can be written as: ... A detailed description of the stiﬀness matrix can be found in ﬁnite element books........ FI1 FI2 ...................... ................. .. . = FS = .............Cnn or it can be written as: .. ... .......... ............Knn un FS = [K]{U} where [K] is the stiﬀness matrix............ ........... .................... .................. The damping matrix can be written as : ... ......C2n u2 ................. .MI(n−1).....
. . u gy .3. If there is a base displacement like ground acceleration along y direction as shown in ﬁgure 3.... . {f (t)} = ..1.. 1 1 =ugy ... . The vector of ground acceleration can be 1 1 . written as: . the equation of motion for a multidegreeoffreedom system can be written as: [M]{U t } + [C]{U } + [K]{U} = {f (t)} . fn (t) . =ugy {I} . FINITE ELEMENT MODELLING OF THE STRUCTURE 103 FD = [C]{U } where [C] is the damping matrix.. . Thus. ugy . {Ug }= .. Equating of the external forces and internal forces will results in: [M]{U t } + [C]{U } + [K]{U} = {f (t)} which {f (t)} is the external forces act on the structures and can be written as: f1 (t) f2 (t) . the equation of motion can be written as: [M]{U } + [C]{U } + [K]{U} = {f (t)} − [M}{U gy } ... . . ugy . Where ugy is the ground acceleration at the base along y direction. .. .3.. ugy . . . .. . . in which {U t } = {Ugy } +{U }.
4: An example of MDF system with two degrees of freedom at each mass For a system of MDF system when subjected to ground acceleration in two directions and there are two degree of freedoms for each mass. . FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM Figure 3. [M]{U } + [C]{U } + [K]{U } = {f (t)} − [M]{U g } In which {U g } can be written as: .4. The same can be written for {U } and {U }.. . .For this system. .. un vn . .as shown in ﬁgure 3..104CHAPTER 3. we can write the displacement vector as: {U} = u1 v1 u2 v2 . The equation of motion can be written as following: ..we can write the equation of motion based on the both degrees of freedom at each mass. .
1 0 1 0 + ugy .. that act on the structure. In the case of concrete dams. . {f (t)} = . fnx (t) fny (t) .. Usually in the case of concrete gravity dams isoparametric plane stress or plane strain are used.. a system is discretized into its degrees of freedom. ..3. {f1 }.. . . {f (t)} can be seperated into hydrodynamic pressure {f } and resultant of all the other forces. . the ﬁnal form of the equation of the motion can be written as: [M]{U } + [C]{U } + [K]{U} = {f } + {f1 } − [M]{U g } In a MDF sytem.. 0 1 0 1 When ugx and ugy are the ground acceleration along the x and y direction. . In the above equation of motion {f (t)} can be written as: f1x (t) f1y (t) f2x (t) f2y (t) . FINITE ELEMENT MODELLING OF THE STRUCTURE 105 {U g } =ugx . The number of degrees of freedom depend on the number of elements and the degree of freedom for each element. 0 1 0 1 .1. respectively. In some cases triangular . .. These types of elements are used for two dimensional problems. 1 0 1 0 . . Which fix (t) and fiy (t) are the forces acting on the ith mass along x and y direction. Thus. respectively.
ui and vi (f xi and f yi ) are the displacements (forces) at node i of the interface element along the global X and Y coordinates. Based on the nature of the problem each node may have 3 diﬀerent components u. In the case of concrete gravity dam. For more complex type of element we may add rotational degrees of freedom in to each node. In 3D analysis the simplest case is the case of tetrahedron element. The superscript and subscript ’e’ refer to the element on the damreservoir interface. v and w in the direction of X.2 COUPLING MATRIX OF THE DAMRESERVOIR The coupling matrix relates the pressure of the reservoir and the forces on the damreservoir interface as following: [Q]{p} = {f } where {f } is the force vector acting on the structure due to the hydrodynamic pressure. FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM elements may be used. In both plane stress or plane strain the displacement ﬁeld is given by the u and v in directions of cartisian x and y axis. assumption of 2D analysis is a valid assumption. Figure ?? shows a line element on the interaction boundary of the damreservoir. The work done by the hydrodynamic pressure on the interaction surface of the structure must be equal to the work of the equivalent nodal forces on the interface boundary of an element. for unit thickness elements as shown in ﬁgure ??.1) pUn ds = {f } {δ} = f x1 f y1 f x2 f y2 u2 se v2 where p and Un are the values of the hydrodynamic pressure and normal displacement along the element interface. {δ} and {f }e are the displacement and force vector of an interface element. the following expression can be written: u1 Z ª v1 © eT (3. respectively. Thus. .106CHAPTER 3. The integration is performed along each element on the damreservoir interface. Three dimensional analysis of concrete gravity dams requires a large number of DOF. respectively. Writing u and v. Y and Z. 3.
4. [Q]e is written as: Z e s [Q] = {Nn }{N f }T ds se e e For an interface element as shown in ﬁgure 3.4) where [Q] and {p} are the coupling matrix and hydrodynamic pressure vector of an element on the damreservoir interface.5. respectively.3. Combining equations 3. we have: Un = un + vn = ηu1 N1 + ηu2 N2 + βv1 N1 + βv2 N2 (3. From equation 3.2 can be written in the following form: Un = © ηN1 βN1 ηN2 βN2 ª s {δ} = {Nn }T {δ} (3. Un . 3. ηand βare the absolute values of the normal vector on the boundary in the global directions of X and Y .2) In equation 3.5) {f } = {Nn }{N f }T ds {p}e = [Q]e {p}e se ª {p}e (3.3 and 3. then: u = u1 N1 + u2 N2 v = v1 N1 + v2 N2 where Ni is the structure shape function at nodei of the interface element. Equation 3. For the normal displacement along the element surface.1. there is obtained: Z e s (3.5. COUPLING MATRIX OF THE DAMRESERVOIR 107 displacements along the global X and Y coordinates of the interface element.3) The hydrodynamic pressure can be expressed as shape function of the ﬂuid in the form: p = {N f }T {p}e = © f f N1 N2 where Nif is the ﬂuid shape function at node iof the interface element. in terms of structure shape functions.2. The total coupling matrix[Q] is obtained by assembling all element coupling matrices.2. then: f f ηN1 N1 ηN1 N2 Z f f βN1 N1 βN1 N2 e [Q] = ηN2 N f ηN2 N f 1 2 se f f βN2 N1 βN2 N2 .
t) = . The motion of these boundaries is related ∇2 p(x. y. the small amplitude irrotational motion of water is governed by the twodimensional wave equation: 1 .6) C2 where p(x. t) (3.. y. Assuming that water is linearly compressible and neglecting its viscosity. FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM Figure 3. is due to the horizontal and the vertical accelerations of the upstream face of the dam. p (x. The hydrodynamic pressure in the impounded water governed by equation 3.5: Interface element on the damreservoir interaction boundary 3.6. the reservoir bottom as well as the far end of the reservoir in the case of ﬁnite reservoir length. C is the velocity of pressure wave in water andxand y are the coordinate axes. the hydrodynamic pressure distribution in the reservoir is governed by the pressure wave equation.108CHAPTER 3. t) is the hydrodynamic pressure in excess of hydrostatic pressure.3 FINITE ELEMENT MODELLING OF THE RESERVOIR As it was discussed in previous Chapter. y.
matrices [H] and [G] are constant during the analysis while the force vector {F } and the pressure vector {p} and its derivatives are the variable quantities in equation 3. y. the wave equation can be written in the following matrix form: . Hij = Hij and F = Fi . . [G]{p} + [H]{p} = {F } (3. For earthquake excitation.7) P e P e P e Gij . t) = 0 whereh is the height of the reservoir. y. The coeﬃcient Ge .7. Using ﬁnite element discretization of the ﬂuid domain and the discretized formulation of equation 3. In the above formulation. the boundary condition at the free surface is written as: p(x. Ae is the element area and se is the prescribed length along the boundary of the elements.3. FINITE ELEMENT MODELLING OF THE RESERVOIR 109 to the hydrodynamic pressure by the boundary conditions. where Gij = ij e Hij and Fie for an individual element are determined using the following expressions: Z 1 e Ni Nj dA Gij = 2 C Ae e Hij Z ∂Ni ∂Nj ∂Ni ∂Nj + )dA = ( ∂x ∂x ∂y ∂y Ae Fie Z ∂p = Ni ds ∂n se where Ni is the element shape function. t) is the component of acceleration on the boundary along the direction of the inward normal n.3. Neglecting the free surface wave. y. No wave absorption is considered at the boundaries of the reservoir. the condition at the boundaries of the damreservoir.6. reservoirfoundation and the reservoirfarend are governed by the equation: ∂p(x.. t) = −ρan (x. t) ∂n whereρ is density of water and an (x. y.
11) C 2h In general form. . FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM 3. .7 results in: .1 Truncated Boundary of the Reservoir’s FarEnd The Sharan boundary condition.. `e is the side of the element on the truncated boundary.9) (3. . can be done by separating the force vector {F } in equation 3.. T Substituting equations 3.3.. at the farend truncated boundary can be written as: ∂p π 1 .10) `e T In equation 3.110CHAPTER 3. 1 π [D]{p} + ([H] + [D]){p} = {F F1 } (3. which was already described.11 in the format of the above equation.10. π 1 [D]{p} − [D]{p} 2h C e e where Dij = Dij and Dij is deﬁned as: Z e Dij = Ni Nj d`T {F F2 } = − (3. the following relationships are obtained: [C 0 ] = 1 [D] C ..9 into equation 3.8 and 3. we can write the ﬁnite element form of the equation of the reservoir as: [G]{p} + .8) where {F F1 } is the component of the force due to acceleration at the boundaries of the damreservoir and reservoirfoundation while {F F2 } is due to truncation at the far boundary and can be written as: . =− p− p ∂n 2h C Implementation of the truncated boundary condition in the ﬁnite element model. . .7 into two components: {F } = {F F1 } + {F F2 } (3.. [G]{p} + [C 0 ]{p} + [K 0 }{p} = {F } − ρ[Q]T ({U } + {U g }) = {F2 } − ρ[Q]T {U } Putting equation 3.
The dot represents the time derivative.[Q] is the coupling matrix. 3. π [D] 2h {F2 } − ρ[Q]T {U } = {F F1 } Where {F2 } is forces due to ground acceleration on the damreservoir boundaries and total acceleration on the rest of the boundaries.. Detailed deﬁnitions of the[G]...[C 0 ] and [K 0 ] matrices and vector {F }. {Ug } is the ground acceleration and ρ is the density of the ﬂuid. The equations of the dam structure and the reservoir can be written in the following form: [M]{U } + [C]{U } + [K]{U } = {f1 } − [M]{U g } + [Q]{p} = {F1 } + [Q]{p} [G]{p} + [C 0 ]{p} + [K 0 }{p} = {F } − ρ[Q]T ({U } + {U g }) = {F2 } − ρ[Q]T {U } where[M]. it can be seen that the damreservoir interaction is a classic coupled problem which contains two diﬀerential equations of the second order. . damping and stiﬀness of the reservoir. and {p} and {U} are the vectors of hydrodynamic pressures and displacements.4. . respectively. . . EQUATION OF THE COUPLED DAMRESERVOIR SYSTEM 111 [K 0 ] = [H] + .. . ..3. {f1 } is the vector of body force and hydrostatic force.. . damping and stiﬀness matrices of the structure and [G]. are presented in the previous sections.. . . [C 0 ] and [K 0 ]are matrices representing mass. . [C] and [K] are mass..4 EQUATION OF THE COUPLED DAMRESERVOIR SYSTEM As we discussed.
112CHAPTER 3. FINITE ELEMENT MODELLING OF THE DAMRESERVOIR SYSTEM .
Chapter 4 DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM 4. The reduced system of equations has high order derivatives which cause some diﬃculties in applying the initial conditions. Diﬀerent approaches to the solution of the coupled ﬁeld problem exist. The simultaneous solution is time consuming and involves many operations. Field elimination. This method contains matrices with a large bandwidth and consequently requires a large amount of memory especially for the cases when the existing matrices are not symmetric. The advantages and disadvantages of each method were addressed by Felippa and Park (1980). especially when a large number of elements is used. the main disadvantage of the ﬁrst two classes of solution arises from the diﬃculties encountered in using available software while the partitioned solution has the capability of using existing software for each subsystem. In such a problem. the presence of interaction implies that the time response of both subsystems must be evaluated simultaneously. The ﬁeld elimination approach is not feasible in the case of nonlinear problems. simultaneous solution and partitioned solution are the three classes of solutions for the coupled ﬁeld system. Staggered solution was described by Felippa and Park (1980) as a partitioned solution procedure that can be organized in terms of sequential execution of singleﬁeld analyser. 113 .1 INTRODUCTION The damreservoir system can be categorized as a coupled ﬁeld system in which two physical domains of ﬂuid and structure interact only at their interface. In addition.
when a linear system is subjected to harmonic excitation. local and global stability. frequency dependence of amplitude of free oscilation. Zienkiewicz and Chan (1989) proposed an unconditionally stable method for staggered solution of soilpore ﬂuid interaction problem. Lack of superposition property. Their method was proved to be unconditionally stable when no damping term was included in the equations of the ﬂuid and the structure. There are two method of solution for dynamic analysis of a system. Most of the staggered solution applications in the ﬁeld of ﬂuidstructure interaction were conducted using a method which is not unconditionally stable. Park and Felippa (1984). Coupled problems and their numerical solutions were addressed by Felippa and Park (1980). Huang (1995) proposed two unconditionally stable methods for the analysis of soilpore ﬂuid problem. have been investigated by several researchers. multiplicity of equilibria and domains of attraction. subharmonic and superharmonic generation are some of characteristics of a nonlinear system. In particular. This character of nonlinear systems make them unable to be solved using frequency domain method. when the damping term is included in the equation of the subsystems. two methods of staggered solution procedure are applied . Methods of solution vary depending on the governing diﬀerential equations of the subystems and may lead to diﬀerent degrees of accuracy and stability of the solution (Park 1980). However. In this Chapter. The methods were named pressure correction method and displacement correction method. These physical systems which are referred to as coupled systems. DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM Most of the physical systems are made of subsystems which interact with each other. A nonlinear system may behave in a diﬀerent manner. Time domain and frequency domain solutions are two methods of solutions and have diﬀerent application depending on the nature of the system. The problem of solutions instability when the damping term is included in the diﬀerential equation.114CHAPTER 4. the proposed method may not be unconditionally stable. the jump response. For a linear system both method can be used while for a nonlinear system only time domain solution can be used to evaluate the dynamic response of the system. was recognized by researchers. Zienkiewicz and Chan (1989) presented an unconditionally stable method for staggered solution procedure for the ﬂuidstructure interaction problem. The behaviour of a nonlinear system is diﬀerent than the linear system. the steady state response will be another harmonic with the same frequency and will be independent of the initial conditions.
[C] and [K] are mass. Detailed deﬁnitions of the[G]. .[Q] is the coupling matrix.2. . . 4. The dot represents the time derivative.4..[C 0 ] and [K 0 ] matrices and vector {F }.3 DIRECT INTEGRATION OF THE EQUATION OF MOTION In direct integration the equation of motion are integrated using a numerical stepbystep procedure.2) where[M]. are presented in the following sections. The accuracy of the solution using both of the proposed methods. and {p} and {U} are the vectors of hydrodynamic pressures and displacements. . THE COUPLED DAMRESERVOIR PROBLEM 115 to the damreservoir interaction problem. Both methods are shown to be unconditionally stable when the two diﬀerential equations of the ﬂuid and structure include damping terms. . 4.1) [G]{p} + [C 0 ]{p} + [K 0 }{p} = {F } − ρ[Q]T ({U } + {U g }) = {F2 } − ρ[Q]T {U } (4... is investigated. . The damreservoir interaction is a classic coupled problem which contains two diﬀerential equations of the second order.2 THE COUPLED DAMRESERVOIR PROBLEM As it was discussed in previous Chapter. damping and stiﬀness matrices of the structure and [G].. . {f1 } is the vector of body force and hydrostatic force. damping and stiﬀness of the reservoir. {Ug } is the ground acceleration and ρ is the density of the ﬂuid. [C 0 ] and [K 0 ]are matrices representing mass.. The term direct means that no transformation of the .. respectively.. . Two diﬀerent conﬁgurations of concrete gravity dams are analysed to illustrate the application of the proposed procedure and to compare the solution with available ﬁnite element solutions. The equations of the dam structure and the reservoir can be written in the following form: [M]{U }+[C]{U }+[K]{U} = {f1 }−[M]{U g }+[Q]{p} = {F1 }+[Q]{p} (4. .
The equation of motion written for the time interval under consideration provides the third expression necessary to determine the three unknown parameters. These speciﬁed or computed values permit the marching scheme to be begun so that response can be computed at as many subsequent points as desired. However. solution does not blow up choosing a big time step. In general two independent expressions of this nature must be speciﬁed. While. Therefore. Newmark method.116CHAPTER 4. Wilson. To achieve an accurate solution a small time step must be chosen based on the natural period of the system as well as the characteristics of the exciting force. in the couples damreservoir equation we are dealing with a couple of equilibrium equation. The stability of each method depends on the time steps taken and some of the methods are unconditionally stable. There are several method of direct integration. DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM equation of motion into the diﬀerent form is carried out. velocitiy aand acceleration are computed. Houbolt are implicit integration methods while Central Diﬀerence Method is the explicit integration method. In time integration scheme. the time integration scheme is called implicite integration method.The direct integration of the equations of motion provides the response of the system at discrete intervals of time which are usually spaced. All the above integration schemes are for the solution of a single equation of motion. In this procedure three basic parameters of displacement. The integration algorithms are based on appropriately selected expressions that relate the response parameters at a given interval of time to their values at one or more previous time points. the accuracy of the solution might be aﬀected. In explicit integration method. The stability of an integration scheme for a single equation does not guarantee the stability of a coupled equation. . The accuracy and stability of the scheme depend on the magnitude of the time interval and marching algorithm. If the equation of motion is written at a time step which the three parameters are unknown(t + ∆t). In an unconditionaly stable method. the equilibrium equation is written at the time step which all the three basis parameters are known (t). the basic parameters are known at the begining of the integration or any points preceding the integration time. It is needed to develope an approach for solution of coupled equation using the above methods.
. (4. The procedure can be started by guessing {P }i+1 ..The governing ﬁeld equations at time i + 1 can be written as follows: [M]{U }i+1 + [C]{U }i+1 + [K]{U }i+1 = {F1 }i+1 + [Q]{p}i+1 [G]{p}i+1 + [C 0 ]{p}i+1 + [K 0 ]{p}i+1 = {F2 }i+1 − ρ[Q]T {U }i+1 . p}p p}i + (1 − γ)∆t{p}i { i+1 = { {p}i+1 = {p}p + β∆t2 {p}i+1 i+1 .. {p}i+1 = {p}p + γ∆t{p}i+1 i+1 . (4.. .. {P }i+1 and {P }i+1 can be written as follows: {U }i+1 = {U }p + γ∆t{U }i+1 i+1 . .. p {p}i+1 = {p}i + ∆t{p}i + (0.3) {U }p i+1 = {U }i + (1 − γ)∆t{U }i {U}i+1 = {U }p + β∆t2 {U }i+1 i+1 .. .7) (4. . {U}p i+1 = {U }i + ∆t{U }i + (0.4) 2 . .4 USING NEWMARKβ METHOD FOR THE COUPLED EQUATIONS Direct integration scheme is used to ﬁnd the displacement and hydrodynamic pressure at the end of the time increment i + 1 given the displacement and hydrodynamic pressure at time i... .7 and 4.. {U }i+1 . . The Newmarkβ method is used for discretization of both equations (implicitimplicit method). . {U }i+1 . ..5) (4. .6) where γand β are the integration parameters.8) The coupled ﬁeld equations 4. In this method . USING NEWMARKβ METHOD FOR THE COUPLED EQUATIONS117 4. (4. .. .4. . (4. .4.5 − β)∆t {U }i . .8 can be solved using the staggered solution scheme.5 − β)∆t2 {p}i .
4. .10 can be modiﬁed as: [M]{U }i+1 = [M]{U }∗ + β∆t2 [Q]{p}i+1 i+1 Substituting equation 4. {P }i+1 can be substituted in equation 4. Substituting {P }i+1 in equation 4. Similarly. Then equation 4. This method can not guarantee the uncondi. .8 can be solved to ﬁnd {P }i+1 .7 to calculate {U}i+1 and its derivatives. thus. tional stability of the solution.9 to calculate {U }∗ ..118CHAPTER 4.. equation 4.5 STAGGERED DISPLACEMENT METHOD In this method. ∗ 2. {P }i+1 can be obtained. guessing{U }i+1 at ﬁrst to calculate {P }i+1 from equation 4.7 gives: [M]{U }i+1 = [M]{U }∗ + β∆t2 [Q]{p}i+1 − γ∆t[C]{U }i+1 − β∆t2 [K]{U }i+1 i+1 (4.7 to calculate {U }i+1 and its derivatives.11 into equation 4.. ... two methods of staggered solution are proposed which are shown to be unconditionally stable.8 and then calculating {U }i+1 from equation 4. . . Substituting {U }i+1 in equation 4. . Therefore. .7 can be approximated as following: [M]{U }∗ = {F1 }i+1 + [Q]{p}p − [C]{U }p − [K]{U }p i+1 i+1 i+1 i+1 Combining equations 4.. the right hand side terms are known. i+1 ..9) (4.. the procedure of the staggered displacement method can be summarized by the following steps: .12.11.8. . 3... . DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM in equation 4.. . .10) Taking advantage of the lumped mass which results in a diagonal mass matrix.11) ([G]+ρβ∆t2 [Q]T [M]−1 [Q]){p}i+1 +[C 0 ]{p}i+1 +[K 0 ]{p}i+1 = {F2 }i+1 −ρ[Q]T {U }∗ i+1 (4.12 to calculate {P }i+1 . ..12) In equation 4. Solving equation 4.. .7 can not provide unconditionally stable procedure. In the following sections. In order to correct the approximation made in equation 4.7 to solve for {U}i+1 and its derivatives. then: . equation 4.9 and 4. (4.. 1.
4.16) . . equations 4. instability can be attributed to that of structure.... While in a conditionally stable method. 4z 2 2z {p}i+1 {p}i+1 {p}i+1 = 2 ∆t ∆t . 2z {U}i+1 ∆t (4.6 become: .1 Stability of the Staggered Displacement Method In an unconditionally stable solution method. ..13) (4.7 without the force term. 4... For β = 0.4. (4. . the condition for stability requires 1−z that the real part of z is negative (Re(z) ≤ 0 ) and that the RouthHurwitz criterion is satisﬁed. .5 and 4. .. . consider a modally decomposed system with scalar values.12 and substituting them into equation 4. [M]{U }i+1 + [C]{U }i+1 + [K]{U}i+1 − [Q]{p}i+1 = 0 . . . then: {p}i+1 = Combining equations 4. the instability may be due to numerical or structural instability.15) {U }p = (1 − z 2 ){U}i+1 i+1 (4.18) . To show that the described method of staggered displacement is unconditionally stable..25 and γ = 0.5. gives: [G]{p}i+1 +[C 0 ]{p}i+1 +[K 0 ]{p}i+1 +ρ[Q]T [M]−1 ([M]+γ∆t[C]+β∆t2 [K]){U }i+1 = 0 (4.17) . . In such a system.5. 2z − z 2 = {p}p = (1 − z 2 ){p}i+1 {p}i+1 {p}p i+1 i+1 ∆t Rewriting equation 4. . Thus forµ < 1 we have: {U }i+1 = µ{U}i {p}i+1 = µ{p}i {U }i+1 = µ{U }i {p}i+1 = µ{p}i . displacement and the pressure must not grow.3. 4. 4z 2 {U}i+1 ∆t2 2z − z 2 {U}i+1 = ∆t {U }i+1 = . {U }i+1 = µ{U }i {p}i+1 = µ{p}i . {U }i+1 = {U }p i+1 .. (4. STAGGERED DISPLACEMENT METHOD 119 4.8 without the force term.5.10 and 4.14) Using ztransformation of µ = 1+z .
15 and 4. Therefore.a3 and a4 are always positive. damping and stiﬀness 0 values m.17 and 4. c 0 0 0 andk instead of [G].21) 0 For the structural system of dam and reservoir. c and k instead of [M]. The coupling matrix [Q] would be represented by scalar quantity q. a4 ≥ 0 ¯ ¯ ¯ 0 a1 a3 ¯ (4.21 are given as: ¯ ¯ ¯ a1 a3 ¯ ¯ ¯ ¯ a0 a2 ¯ > 0 a1 a2 − a3 a0 = (4.19) ¯ ρq (m + ∆t c + ∆t2 k) 4z2 g 4z2 + c0 2z + k0 ¯ = 0 m 2 4 ∆t2 ∆t2 ∆t or a0 z 4 + a1 z 3 + a2 z 2 + a1 z + a4 = 0 where: a0 = a2 a3 16mg 8mc0 8gc a1 = + 3 ∆t2 ∆t3 ∆t 0 0 2 4mk 4cc 4gk 4ρq 2ρq 2 c ρq 2 k + = + 2+ 2+ + ∆t2 ∆t ∆t ∆t2 m∆t m 2c0 k 2ck0 + a4 = kk0 = ∆t ∆t (4.17 and 4.18. a1 . DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM The modally decomposed system is represented by a single degree of freedom equation.16 into equations 4. g.18 as follows: ¯ ¯ ¯ m 4z2 + c 2z + k ¯ −q ∆t ¯ ∆t2 ¯ (4. The single degree of freedom equivalent of equations 4.22) 32 32ρ 16ρ 32 2 0 0 m c k + 5 mc02 c + 5 mc0 q2 + 4 c0 q2 c ∆t5 ∆t ∆t ∆t 32 2 0 32 2 8ρ 0 2 + 3 c q k + 5 gc c + 5 g ck ∆t ∆t ∆t .m. a2 . The characteristic equation of the coupled ﬁeld can be written by substituting equations 4.20) The RouthHurwitz conditions for stability are: ¯ ¯ ¯ a1 a3 0 ¯ ¯ ¯ ¯ a0 a2 a4 ¯ > 0 a0 > 0 a1 . c andk0 are positive quantities.17 and g. k.c. The values of the two determinants in equation 4. a2 . a3 . [C ] and [K ] in equation 4.120CHAPTER 4.18 will be obtained by substituting the mass. [C] and [K] in equation 4. a0 .
STAGGERED PRESSURE METHOD + a1 a2 a3 − a2 a4 − a2 a0 = 1 3 32ρ 16ρ 8ρ gcq2 + 4 gc2 q2 + 3 gcq 2 k ∆t5 ∆t ∆t 121 (4. Recalling the condition of stability (equation 4.22 and 4.6..24 into equation 4. 4...26 into equation 4..21). . (4.24) Substituting equation 4.26) Substituting equation 4. (4. . .25) or: ([G] + β∆t2 [K 0 ] + γ∆t[C 0 ]){p}i+1 = [G]{p}∗ − ρ[Q]T {U }i+1 i+1 .7 with [H] = [G] + β∆t2 [K 0 ] + γ∆t[C 0 ]... there is obtained: [G]{p}i+1 = [G]{p}∗ − ρ[Q]T {U }i+1 − γ∆t[C 0 ]{p}i+1 − β∆t2 [K 0 ]{p}i+1 i+1 (4.8 as following: [G]{p}∗ = {F2 }i+1 − [C 0 ]{p}p − [K 0 ]{p}p i+1 i+1 i+1 . .23) √ √ 2 64 64 64 (mk 0 cc0 − gk cc0 ) + 6 mc03 ck + 6 mc02 c2 k0 6 ∆t ∆t ∆t 64ρ 02 2 64ρ 0 2 0 32ρ 02 2 + 6 mc q k + 6 mc q ck + 5 c q ck ∆t ∆t ∆t 32ρ 0 2 2 0 16ρ 02 2 2 16ρ 0 2 + 5 c q c k + 4 c q k + 4 c q kck0 ∆t ∆t ∆t 64 2 02 64 3 0 0 64ρ + 6 gc c k + 6 gc c k + 6 gcq2 c0 k ∆t ∆t ∆t 64ρ 2 2 0 32ρ 2 2 0 32ρ + 6 gc q k + 5 gc q c k + 5 gc3 q2 k0 ∆t ∆t ∆t 16ρ 16ρ 2 2 0 2 2 0 + 4 gcq k c + 4 gc q kk ∆t ∆t All the terms in equation 4. . .4.8... gives: . the pressure can be approximated using equation 4.6 STAGGERED PRESSURE METHOD In this method. then the method of staggered displacement is unconditionally stable. .23 are positive.. .
(4. {U }i+1 into equation 4.29) The characteristic equation of the coupled ﬁeld for a modally decomposed system with scalar values.15 and 4.. 1. .29: ¯ 4z2 ¯ ¯ m 2 + c 2z + k −q ¯ ∆t ¯ ∆t 2 ¯=0 2 4z 2z ¯ ρq 4z g ∆t2 + c0 ∆t + k0 ¯ ∆t2 or a0 z 4 + a1 z 3 + a2 z 2 + a1 z + a4 = 0 (4. can be written by substituting equations 4.. then: [M]{U }i+1 + [C]{U }i+1 + [K]{U}i+1 − [Q]{p}i+1 = 0 . . the variable {U }i+1 can be calculated.26 to calculate {p}i+1 and its derivatives. .6. similar procedure as that used in the displacement method can be applied.. i+1 .27) Using equation 4.. .. Therefore.16 into equations 4. .24 to calculate {p}∗ . 2. the procedure of the staggered pressure method can be summarized by the following steps: .1 Stability of the Staggered Pressure Method For stability check... (4. i+1 . .8 without the force terms..122CHAPTER 4.28 and 4.. [G]{p}i+1 + [C 0 ]{p}i+1 + [K 0 ]{p}i+1 + ρ[Q]T {U }i+1 = 0 .7 gives {p}i+1 and its derivatives. 4.30) where: a0 = a2 a3 16mg 8mc0 8gc a1 = + 3 4 3 ∆t ∆t ∆t 4mk0 4cc0 4gk 4ρq 2 = + 2+ 2+ ∆t2 ∆t ∆t ∆t2 0 0 2c k 2ck = + a4 = kk0 ∆t ∆t .27 to calculate {U }i+1 . ..27. Substituting .7 and 4.. 3. Substituting {U }i+1 in equation 4.28) (4. Solving equation 4. DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM ([M] + ρβ∆t2 [Q][H]−1 [Q]T ){U }i+1 + [C]{U }i+1 + [K]{U }i+1 = {F1 }i+1 + [Q]({p}p + β∆t2 [H]−1 [G]{p}∗ ) i+1 i+1 . Rewriting equations 4. Substituting {p}∗ in equation 4.
4.31) 32 2 0 0 32 32ρ m c k + 5 mc02 c + 5 mc0 q 2 ∆t5 ∆t ∆t 32 2 0 32 2 32ρ + 5 gc c + 5 g ck + 5 gcq 2 ∆t ∆t ∆t a1 a2 a3 − a2 a4 − a2 a0 = 1 3 (4.4.21). Therefore.34) Therefore.27 which changes the mass matrix from diagonal to a full matrix. ..33) ∆t ∆t ∆t These terms are all positive.. the procedure of the modiﬁed staggered pressure method can be summarized by the following steps: .. The determinants in the RouthHurwitz conditions (equation 4. The staggered displacement method is the most suitable coupled ﬁeld problem solution procedure for the case of nonlinear analysis.21.32) √ √ 2 64 64 64 (mk 0 cc0 − gk cc0 ) + 6 mc03 ck + 6 mc02 c2 k0 6 ∆t ∆t ∆t 64ρ 02 2 64ρ 0 2 0 64 3 0 0 + 6 mc q k + 6 mc q ck + 6 gc c k + ∆t ∆t ∆t 64ρ 2 02 64ρ 2 2 0 64ρ + 6 gc c k + 6 gc q k + 6 gcq2 c0 k (4. In the case of the staggered pressure method some diﬃculties may arise due to added mass eﬀect in equation 4. give: a1 a2 − a3 a0 = (4.7. For this reason the staggered pressure method was modiﬁed to apply to nonlinear analysis. the method of staggered pressure is unconditionally stable. . given the stability condition of equation 4. . MODIFIED STAGGERED PRESSURE METHOD 123 The coeﬃcients of the polynomial are all positive. .27 in the following approximate form: [M]{U }i+1 +[C]{U }i+1 +[K]{U}i+1 = {F1 }i+1 +[Q] ({p}p +β∆t2 [H]−1 ( [G]{p}∗ −ρ[Q]T {U }i ) ) i+1 i+1 (4.7 MODIFIED STAGGERED PRESSURE METHOD Most of the available nonlinear solutions assume a diagonal mass matrix for the purpose of analysis. The staggered pressure method is modiﬁed by rewriting equation 4.
..35 gives: [M]{U }i+1 = [M]{U }∗ +β∆t2 [Q]{p}i+1 −γ∆t[C]{U }i+1 −(1+α)β∆t2 [K]{U }i+1 i+1 (4..37 and 4..124CHAPTER 4. . equation 4.26 to calculate {p}i+1 and its derivatives. i+1 . .. . . 2... The governing ﬁeld equations at time i + 1 can be written as follows: [M]{U }i+1 +[C]{U }i+1 +(1+α)[K]{U}i+1 = {F1 }i+1 +[Q]{p}i+1 +α[K]{U}i (4. .36 can be solved using the staggered displacement solution scheme.38) . The coupled ﬁeld equations 4.. .. 3. . the modiﬁed staggered pressure method is used instead of the staggered pressure method and the results are compared with those obtained using the staggered displacement analysis procedure. {p}i+1 and {p}i+1 can be written same as Newmarkβ method.8. 4.34 to calculate {U }i+1 .. . In this method.24 to calculate {p}∗ . Substituting {p}∗ in equation 4. .. 4.. In the following analysis.. 0 p 0 T [G]{p}i+1 +[C ]{ }i+1 +(1+α)[K ]{p}i+1 = {F2 }i+1 −ρ[Q] {U }i+1 +α[K 0 ]{p}i (4. .35 and 4. . Substituting {U }i+1 in equation 4.35 can be approximated as following: [M]{U }∗ = {F1 }i+1 +[Q]{p}p −[C]{U }p −(1+α)[K]{U}p +α[K]{U}i i+1 i+1 i+1 i+1 (4. Solving equation 4.8 USING αMETHOD FOR THE COUPLED EQUATIONS .1 Staggered Displacement Method In this method.35) .36) where α is the integration parameter which is introduced in the coupled ﬁeld equation. i+1 . {U}i+1 . {U }i+1 . . The modiﬁed staggered pressure method does not guarantee unconditional stability of the solution. DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM 1.37) Combining equations 4.. .
Hilber and Hughes. For the nonlinear equations. As the cracks open. The numerical diﬃculties due to opening and closing of cracks can be overcome by using the αmethod (Hilber et al. Solving equation 4. Substituting {p}i+1 in equation 4.39. .39) (4..35 to calculate {U}i+1 and its derivatives. 3. .. {p}i+1 can be obtained. The damreservoir interaction representation using the staggered solution technique using αmethod is suitable for nonlinear fracture analysis of concrete dams..38 can be modiﬁed as: [M]{U }i+1 = [M]{U }∗ + β∆t2 [Q]{p}i+1 i+1 Substituting equation 4... USING αMETHOD FOR THE COUPLED EQUATIONS 125 Taking advantage of the lumped mass which results in a diagonal mass matrix. 1977. When the cracks are closed.8.39 into equation 4. .4. the numerical solution is based on piecewise linear solution.36. It can be shown that the method of staggered displacement is unconditionally stable for the linear coupled equations of the damreservoir system with structural damping when α = 0. (4.37 to calculate {U }∗ .. cracked elements recover their strength and therefore the structure gains stiﬀness. . i+1 . Substituting {U }i+1 in equation 4. The nonlinear seismic analysis of concrete gravity dams includes opening and closing of the cracks due to the cyclic nature of the earthquake. In order to correct the approximation made in equation 4. The eﬀect of opening and closing of cracks introduce high frequency shock waves into the structure.35 to calculate {U}i+1 and its derivatives... the right hand side terms are known. ∗ 2. the procedure of the staggered displacement method can be summarized by the following steps: . . The solution stability depends on the length of the time steps and the introduced numerical damping. Therefore. thus.40. . {p}i+1 can be substituted in equation 4. 1978). equation 4. then: ([G] + ρβ∆t2 [Q]T [M]−1 [Q]){p}i+1 + [C 0 ]{p}i+1 + (1 + α)[K 0 ]{p}i+1 = {F2 }i+1 − ρ[Q]T U }∗ + α[K 0 ]{p}i i+1 .40) In equation 4. 1.40 to calculate {p}i+1 . the stiﬀness of the structure reduces.
9 SEISMIC ENERGY BALANCE In the design of structure subjected to earthquake loading. T Z ([Q]{p})T {dU } . two approaches can be considered for the energy equation. It is an eﬃcient method that is accurate in lower modes and dissipate energy in the higher modes when compared with other time integration techniques. the energy supply must be larger than the energy demand. 1 .1). Filiatrault et al. DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM The αmethod of time integration algorithms introduces numerical damping to the system. .41) 2 Z or: EK + ED + ER = EP + EQ + EH (4. The αmethod is used for discretization of both equations of the coupled ﬁeld problem (implicitimplicit method). 4. Direct integration is used to determine the displacement and hydrodynamic pressure at the time increment i + 1. T T {U t } [M]{U t } + {U } [C]{dU } + {r}T {dU } = (4. (1994) used energy balance to study the nonlinear behaviour of diﬀerent structures under variable earthquake ground motion. The two approaches to energy formulation were found to give diﬀerent energy responses. the energy equation can be used to study the energy absorbtion of diﬀerent components. In this regard.126CHAPTER 4. can be written as: Z Z . They found that absolute energy formulation is simple and more straightforward.42) {f1 } {dU} + T Z {U t } [M]{dUg } + . The damreservoir interaction representation using the staggered solution technique is introduced to the nonlinear fracture analysis of concrete dams. The energy equation of the dam structure governed by equation (2. Diﬀerent time stepping algorithms were used to investigate the eﬀect of numerical damping. In a satisfactory design. Uang and Bertero (1990) used absolute and relative energy formulations for a single degree of freedom system. using the αmethod ensures that the response of higher modes is damped out.. exact energy balance can be achieved. Thus. Without the numerical damping.
{r} is the vector of the nonlinear restoring force. energy is added to the seismic input energy EQ and kinetic energy EK through the mass added to the structural system. Increasing the number of iterations of the staggering scheme is a time consuming process. The absolute kinetic energy is EK. the hydrodynamic energy EH is excluded from the energy equation. Thus: EK + EE + ED + EF = EI EK and EE contribute to the stored energy while ED and EF represent the dissipated energy. The relative displacement is {U } while {Ut } is the total (absolute) displacement vector {Ut } = {U } + {Ug }. The error in the energy balance represents an excessive amount of damage when numerical damping is introduced.42. ACCURACY OF THE SOLUTION SCHEME 127 In equations ?? and 4. and the energy dissipated due to fracture EF (EF = ER − EE).10. is the sum of the seismic input energy due to the inertial force EQ. {Ug } is the ground displacement vector. the work of preseismic applied force is EP . the nonlinear restoring work is ER. the absolute seismic input energy is EQ and the work done by the hydrodynamic pressure is EH. In the analysis. hydrodynamic force EH and work of preseismic applied load EP . The staggered displacement method and the modiﬁed staggered pressure method are compared with the . EH = 0. ER contributes to the stored elastic energy in a system EE. The energy balance error is computed as: Error = (EP + EQ + EH) − (EK + ED + ER) × 100 (EQ + EH) When the damreservoir interaction eﬀects are represented by added masses. The input energy EI. The restoring energy. the viscous damping energy is ED.4. 4. However.10 ACCURACY OF THE SOLUTION SCHEME The accuracy of the staggered solution scheme can be improved by increasing the number of iterations and/or by decreasing the time step. In all of the following analyses no iterations have been made for the purpose of improving accuracy. The accuracy of the proposed methods is based on the selection of the appropriate time step. the results of the fracture response are presented for the time before the ﬁve percent energy balance error is reached.
128CHAPTER 4. DYNAMIC ANALYSIS OF DAMRESERVOIR SYSTEM ﬁnite element solution of example problems for the purpose of evaluating the accuracy of the analysis.
Chapter 5 NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS
5.1 INTRODUCTION
The cracking behaviour of concrete dams has been the subject of extensive research during the past decade because few dams suﬀered severe cracking during earthquakes. Rescher (1990) indicated that most concrete gravity dams will experience cracking even under operational loading conditions and moderate earthquake ground motions. Therefore, the assumption of linear behavior may not be appropriate in the analysis of the seismic response of concrete gravity dams. Concrete dams are distinguished from other structures because of their size and their interactions with the reservoir and foundation. The results obtained from the nonlinear analysis of concrete dams are strongly dependent on the approach to modelling of these interactions. It is a diﬃcult task to develop a comprehensive analytical model to include both nonlinearity and interaction eﬀects. The size eﬀect can also inﬂuence the properties of the dam concrete. The fracture properties of normal concrete can be determined using laboratory tests. However, the dam concrete diﬀers from the normal weight concrete because of aggregate size and its poor strength. Little information is available on the fracture properties of the dam concrete. The fracture surface of the dam concrete specimens is characterized by mainly aggregate failure. 129
130CHAPTER 5. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS Saouma et al. (1991) attempted to measure fracture toughness of a concrete specimen in the laboratory which was considered to be similar to dam concrete. They concluded that a deﬁnitive decision cannot be made concerning the results and their accuracy. Bruhwiler and Wittmann (1990) carried out a dynamic test to determine the material properties of the dam concrete under high rate of loading and an initially applied compression load. They found that the fracture energy of dam concrete is 2 to 3 times higher than that of ordinary concrete. The reason is related to the tensile strength characteristic of dam concrete. The tensile behavior of concrete can be divided into two stages. In the ﬁrst stage, the behavior is linear until the tensile strength is reached. In the second stage, strain softening behavior is observed. Fracture energy is sensitive to the tensile stress. In addition, increasing the preloading decreases the fracture energy. To understand the nonlinear behaviour of concrete dams, modelling of the cracking and damage process is needed. Bazant and Oh (1983) proposed a fracture mechanics approach as a blunt smeared crack band. The proposed approach represented a signiﬁcant advance in comparison to the linear fracture theory. The strain softening of the material was considered based on the fracture parameters, fracture energy, uniaxial tensile strength and crack band width. Fracture energy can be determined from the complete stressstrain curve. Formulas were derived to give the fracture parameters. Two classes of solutions can be found in the nonlinear study of concrete gravity dams. Discrete crack approach is the ﬁrst class of solutions which is based on the variable mesh approach. Two methods of linear elastic fracture mechanics LEFM, and nonlinear fracture mechanics NLFM, can be used in this approach. The second class of solutions is the continuum model in which a ﬁxed ﬁnite element mesh is used. Smeared crack model and damage mechanics are the two methods of solution in this class. These two families of fracture model have been investigated in parallel. They also called global approach and local approach. In the global approach, the cracks are simulated by discontinuity in the continuum and the stresses are obtained using linear elastic fracture mechanics(LEFM). This method is coupled to a numerical technique such as ﬁnite elements or boundary elements. On the other hand, the local approach of fracture is based on changes in the constitutive law governing the behaviour of concrete. The integrity of the structure is not required during the solution process. Bhattacharjee and Léger (1994) applied NLFM to predict the response of concrete gravity dams. The experimental work done on a model of a concrete
The reservoir and foundation were modelled as a series of dampers and springs such that the same response can be obtained for the linear response of the crest when compared with the case of actual interaction. The eﬀect of hydrodynamic pressure inside the crack in the seismic analysis of concrete gravity dams was investigated by Tinawi and Guizani (1994). when a rigid foundation is assumed with no reservoir bottom absorption.2 in which the analysis was successfully completed. INTRODUCTION 131 gravity dam and a small beam specimen conﬁrmed the applicability of the proposed NLFM approach. the . Under earthquake excitation. the analysis stopped after the ﬁrst few seconds because of energy balance error due to spurious deformation of some elements. The nonlinear response of a concrete gravity dam with an initial distribution of temperature gradient when subjected to the earthquake was studied by Léger and Bhattacharjee (1995). The coaxial rotation crack model gives a better response than the ﬁxed crack model. They used frequencyindependent added mass matrix as a representative of damreservoirfoundation interaction. Léger and Bhattacharjee (1994) studied the energy response of concrete gravity dams. no crack was observed at the top part of the dam. They found that the vertical ground motion acceleration component is not critical in seismic cracking response of dams. cracking response showed that the crack starts from the downstream side and moves toward the upstream side.5. They found that the response is sensitive to time variation of the input motion. It was found that under high frequency content earthquakes.1. Without introducing the numerical damping. They used a stiﬀness proportional damping with method of integration. The seismic analysis of Koyna dam under both horizontal and vertical components of the earthquake was conducted. The pressure inside the crack does not change the response of the dam signiﬁcantly. An energy balance error approach is used as a measure of damage. A crack was formed at the foundation level. Léger and Leclerc (1996) studied the nonlinear response of concrete gravity dams subjected to diﬀerent earthquake ground motions. The reservoir eﬀect was represented by added mass technique. Bhattacharjee and Léger (1993). No discrepancies were found in the results of the analysis before the occurrence of instability. Dissipated fracture energy is negligible in comparison to other sources of energy dissipation. The cracks are either horizontal or they sloped downward. Most of the time. when compared with the case of = 0. This form of cracking does not promote dam instability. NewtonRaphson iteration technique was used to remove the unbalanced load.
The model was capable of initiation. 1993). Reservoir interaction was modelled using a boundary element. The nonlinear response of the dam including reservoir interaction was strongly aﬀected in comparison with the added mass approach. The model was applied to a dam of rigid foundation .132CHAPTER 5. The cracking response of a concrete gravity dam when subjected to earthquake loading can be diﬀerent if nonuniform damping or uniform damping including the damping due to cracking is considered (Barrett et al. Ayari and Saouma (1990) proposed a model for simulation of discrete crack closure. The results were obtained for 5 seconds only of the earthquake in which the numerical damping was less than 10%. The reservoir bottom absorbtion and water compressibility did not change the response signiﬁcantly. A ﬁctitious crack approach was used to model the crack tip. opening. Special treatment was used to model aggregate interlock eﬀect. The impact of cracking surface was modeled as a load pulse. The aggregate interlock eﬀect was found to be important in the ﬁnal cracking conﬁguration of the dam. a noticeable change in the response was observed.. At the base of the dam. Pekau et al. Fracture mechanics analysis using discrete crack approach was applied. The model was checked by a shake table test of cantilever beam made of gypsum. the dam was represented by a small number of elements. closing and reopening of discrete cracks. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS hydrodynamic pressure inside the crack may increase when higher modes are signiﬁcant. Singhal (1991) found that the Westergaard’s added mass approach yields higher values for crest displacement and stress than that obtained using other approaches. Using diﬀerent computer codes. The slope of the reservoir bottom strongly inﬂuenced the nonlinear response. Nonlinear seismic response of concrete gravity dams was studied by Skrikerud and Bachmann (1986). The model was applied in the dynamic analysis of the Koyna dam (India) under both horizontal and vertical components of the earthquake. (1991. 1995) and Pekau and Batta (1991) presented a method to study the cracking of concrete gravity dams using the principle of Linear Elastic Fracture Mechanics (LEFM) and boundary element mode superposition analysis. The nonlinear response of the Pine Flat dam was studied using the discrete approach (Wepf et al. Linear response of the dam was compared with EAGD84 code analysis and good agreement was found. 1991). In the analysis. When the bottom few elements were cracked. the hydrodynamic pressure may be 50% higher than the hydrostatic pressure.
The crack pattern was found to be very sensitive to the parameters chosen for the analysis. Soon after initiation of the top crack. Response of the linear model with and without the reservoir interaction was determined. Nonlinear response of concrete gravity dams was also studied by Feltrin et al. the nonlinear model was applied for the case of preformed base crack. They concluded that the eﬀect of damreservoir interaction must be included in the nonlinear analysis. The nonlinearity of the reservoir was introduced into the proposed method to investigate the reservoir interaction eﬀect. The cracks started at the top part from down stream face of the dam near the slope discontinuity and moved horizontally. Mlakar (1987) studied the nonlinear dynamic behaviour of concrete gravity dams using the ADINA code.1. The ﬁrst crack started at the foundation level and then it followed by a crack at the top part of the dam at the same location of the crack of empty reservoir case. The crest displacement was found to be higher than that of the empty reservoir.5. The analysis stopped due to excessive damage. Despite the diﬃculties involved. INTRODUCTION 133 with empty reservoir.95 sec. During the rest of the analysis. In the case of homogeneous dam.b) investigated the nonlinear response of concrete gravity dams. It was found that the crack ﬁrst started at the base. The ﬁrst four seconds of an artiﬁcially generated time history was used for the purpose of analysis. the top crack initiated at t = 1. no other cracks were observed and only rocking and opening and closing of the crack were observed. The water cavitation in addition to cracking of concrete was considered. Nonlinear response of the empty reservoir was studied by scaling the ground motion until cracking occurred. The nonlinearity in concrete behaviour included the strain softening and aggregate interlock. In this chapter the nonlinear fracture response of concrete gravity dams . Then cracking initiated at the top part near slope discontinuity. Fenves and VargasLoli (1988) proposed a method for damreservoir interaction which resulted in a symmetric matrix representation of the total equation of the system. They found that the eﬀect of cavitation is not signiﬁcant in the response of the structure. it went through the dam body and almost separated the top part from the rest of the dam. A diﬀerent response was observed under the eﬀect of reservoir interaction. top crack and homogeneous dam without any cracks. The cracks near the slope discontinuity propagated instantaneously and passed through the cross section. A rigid foundation was assumed for Pine Flat dam and the reservoir interaction was included. (1990). ElAidi and Hall (1989 a.
Analytical models for twodimensional fracture propagation studies are well developed now.134CHAPTER 5. the structural response of gravity dams is mainly determined by gravity action. Unlike the circumstance in arch dams. Both models have been used over the decades because of the advantages and inconveniences that they bring to the constitutive models for ﬁnite element crack propagation analysis of concrete structures. . but the threedimensional application of the fracture models is still in its infancy. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS due to seismic loading is investigated. where the nonlinear joint behaviour in the arch direction could be a decisive factor in the stability of the structure . but the constitutive behaviour of concrete is the most important part of nonlinear seismic response study of the dams. The concrete gravity dam monoliths. 5. Special emphasis is put on the applications of the models in seismic analyses of the dams and the limitations of past investigations are examined. usually unkeyed or lightly grouted. Hence. A stateoftheart review on the constitutive models for twodimensional ﬁnite element crack propagation analysis of concrete gravity is presented in this part.2. Analytical models for twodimensional fracture propagation studies are well developed now. twodimensional planestress idealizations seem to be appropriate for nonlinear seismic response study of concrete gravity dams.2 A BRIEF STUDY OF NONLINEAR PARAMETERS Modelling the constitutive behaviour of concrete is the most important part of nonlinear seismic response study of the dams.1 Finite element models of crack propagation Two approaches have generally been followed for the spatial representation of tensile crack propagation in ﬁnite element analysis of concrete structures: the discrete crack model and the continuum crack model. 5. are expected to vibrate independently under severe ground excitations. Smeard crack approach based on a nonlinear fracture mechanics crack propagation criterion is used to study the cracking and response of concrete gravity dams. The damreservoir interaction is included in the time domain analysis using the method of staggered displacement.
The tendency of the model to cause diﬀused crack pattern and the directional bias caused by slanted ﬁnite element mesh are still signiﬁcant computational diﬃculties. only the constitutive relationship is updated with propagation of cracks and the ﬁnite element mesh is kept unchanged. The progressive physical discontinuity in the system is reﬂected instantaneously in the ﬁnite element model by modifying the mesh during the analysis.2.2. However.3 Continuum crack propagation models (CCPM) The continuum crack can be divided into Smeared crack propagation model (SCPM) and Damage mechanics crack propagation model.5. It is generally argued that the nonlinear response of concrete dams is dominated by a few discrete long cracks. the discrete crack model may be a sensible choice for dam fracture analysis. After initiation of the fracture process. and the direct estimation of crackopening displacement (COD) proﬁle.2. such as the damfoundation interface and construction joints. In these models the fracture is idealized to propagate as a blunt front smeared over an entire element or a certain band width of the element. determined by a suitable constitutive model. the precrack material stress.strain relationship is replaced by an orthotropic relationship with material reference axis system aligned with the fracture direction. The principal disadvantages in applying this model are the diﬃculty and high computational cost due to continuous change of the ﬁnite element topology during the analysis. the aggregate interlock in the rough cracks. 5. The advantage of the models lies in its simplicity and cost eﬀectiveness. A BRIEF STUDY OF NONLINEAR PARAMETERS 135 5. and the unobjective eﬀects of ﬁnite element mesh and crack length increment. The growth of the crack is determined by strength or fracture mechanics based constitutive models.. Thus. although the physical nature of crack representation is questionable.DCPM.( variable mesh) In the DCPM. From this consideration. The tension stiﬀness across the crack plane is either eliminated suddenly or a gradual stressrelease criterion is applied. global approach.2 Discrete crack propagation model . The speciﬁc advantages of the DCPM are the abilities to consider explicitly the water penetration and uplift pressure on the crackopen surfaces. A special case of discrete crack modelling is the application of special elements to represent the a priori weak joints in the system. a crack is represented as a discrete gap along the interelement boundary. the model has been extensively used in .
A sudden release of stress on the fracture plane is commonly assumed upon reaching the material tensile strength. the fracture initiation and propagation criteria.1 Strengthbased criteria The early investigations on cracking of concrete have mostly applied simple criteria based on the concepts of strength of material (SOM). Major developments in the realms of crack propagation analysis and their relative merits are presented in the following sections. The behaviour of concrete under high compressive loading is predominantly nonlinear. to study the nonlinear behaviour of concrete under compression. However. It is a usual practice in concrete structure analysis to presume linear elastic behaviour before the onset of tensile fracture process. Diversity in various models lies in the deﬁnition of the fracture propagation criteria after the crack has been introduced in the structure.3 Constitutive models for crack propagation To conduct a comprehensive fracture propagation analysis. and the postcrack behaviour. Several models based on the concepts of elastoplasticity and elastoviscoplasticity have been proposed. The gradual release of stress with increasing strain has also been used. the selected constitutive model should describe the prefracture material stressstrain behavior. A crack is assumed to propagate when the predicted stress or strain at the cracktip exceeds the critical value representative of the strength of the material. the compressive stresses in concrete gravity dams are expected to be low even under severe ground excitations. It is almost universal in the constitutive models of fracture analysis to assume the initiation of new cracks in a homogenous structure when the principal tensile stress reaches the tensile strength of concrete. A reasonable assumption of linear elastic behaviour under compressive loading has been applied in almost all previous investigations. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS the seismic response study of concrete gravity dams. mainly for numerical stability reasons. in ﬁnite element analyses of cracking problems. The crack propagation criterion. is identical to the new crack initiation criterion.3. The SOM criterion of crack propagation has been used in discrete and smeared crack propagation ﬁnite element models. Comparison of computed tensile . in this approach. 5. 5.136CHAPTER 5.
The fracture mechanics of concrete has drawn signiﬁcant attention of the research community over the last decade. The lack of ﬁnite element mesh objectivity of the SOM criterion was reported. It has been recognized only recently that the failure mechanism in concrete structures is diﬀerent from the usual strength based concept.dimensional fracture propagation studies. the literature on the subject is voluminous. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 137 stress with the strength of the material is not rational for a cracked structure because spurious results may be obtained depending on the size of the ﬁnite element ahead of the propagating crack. For two.tearing fracture Three elementary modes of failure are recognized in the fracture theory (Figure 5.5. based on the concept of energy dissipation by the structure undergoing fracturing process. (c) mode III . is considered for three. only the principal developments in fracture mechanics. (b) mode II planar shear fracture. due to the progressive growth of a fracture process. Fracture mechanics crack propagation models can be . The third one. are reviewed. as a consequence.1: Modes of failure: (a) mode I .2 Fracture mechanics criteria Fracture mechanics is the theory dealing with propagation of cracks. 5. Here. the tearing mode (mode III).Tensile fracture.3.3. relevant to the seismic analysis of concrete dams.1). modes I and II (the opening mode and the planar shear mode) are usually considered.dimensional idealizations. Figure 5.
determined from linear elastic solutions: KI σ 22 √ KII = lim 2πt σ 12 r→0. Once the stress intensity factors. r and θ are polar coordinates. However. The question of whether the fracture process in concrete can take place at a localized point has been a subject of intense debate for quite long time. Techniques to apply the LEFM criteria in smeared crack propagation ﬁnite element model have also been reported in the literature. (1991) have proposed a numerical procedure to apply the linear elastic fracture mechanics criterion in discrete crack propagation analysis with boundary element model for the dams. The disregard of nonlinear behaviour in the FPZ is an assumption of unknown consequences in inﬂuencing the global response of the structure. a suitable functional relationship is applied for propagation of an existing crack: f (KI .138CHAPTER 5. KIII .1) Several functional forms of equation 5. The stress ﬁeld around the tip of a sharp crack is characterized by the stress intensity factors. a sudden release of stress on the surface is assumed with the extension of the crack. the fracture process zone (FPZ) must have some ﬁnite size (Figure5.θ→0 KIII σ 23 where σ ij are the near cracktip stresses. It is argued that the LEFM could be applied if the FPZ is much smaller than the dimension of the structure under consideration. no rational experimental evidence has ever been put forward appraising the extent of FPZ in dam concrete. K1c . It seems . In reality. and Ki are associated with three fundamental fracture modes.1 have been proposed in the literature. In LEFM models. Most investigators adopt the discrete crack propagation ﬁnite element model (DCPM) with the LEFM constitutive models. According to LEFM. Ki. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS broadly classiﬁed into two categories the linear elastic fracture mechanics (LEFM) models and the nonlinear fracture mechanics (NLFM) models. the fracture process occurs right at the cracktip and the entire volume of the material remains elastic (Figure5. have been numerically (or analytically) computed and the material fracture toughness (sometimes named critical stress intensity factor).2b). K1c ) = 0 (5. Very large concrete structures like dams are usually cited as the possible candidates for application of LEFM models. experimentally determined.2a). Pekau et al. Ki . KII .
3. (a) LEFM. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 139 Figure 5. (b) NLFM .2: Fracture process zone (FPZ).5.
dissipated during fracture Process on unit area: Gf = Z δf σ(δ)dδ 0 Gf is a material property and often referred to as fracture energy or speciﬁc fracture energy. σ. δ. which means that a long fracture process zone may exist (Dungar et al. The choice between LEFM and NLFM models could also be inﬂuenced by the strain rate under consideration. have also been modiﬁed to take account of the FPZ through an eﬀective crack length calculated as the true crack length plus a portion of the ﬁctitious crack. They characterized the existence of FPZ as a ﬁctitious crack lying ahead of the real crack tip (ﬁgure5. The behaviour of concrete in the FPZ was represented by a diminishing stress. (1976). The basic nature of the model has made its extensive applications possible in discrete crack propagation analysis. versus crack opening displacement (COD). The primary characteristic of nonlinear fracture mechanics (NLFM) is the recognition of the strain softening behaviour of concrete in the FPZ. may not be true even for concrete gravity dams. The most referenced work is due to Hillerborg et al. δ. The special feature of the Hillerborg’s ﬁctitious crack model is the dissipation of energy over a discrete line crack. In some recent studies the key assumption of Hillerborg’s model. relationship. (Figure 5. In a gravity dam. the tensile resistance being ceased at a critical COD value. it must involve a ﬁnite . Bazant and Oh (1983) came forward with the argument that the energy dissipation cannot take place in a diminishing volume. crack opening displacements may be very small. in the context of discrete crack propagation. 1991). Hence the argument of small fracture process zone in comparison to the thickness of the structure. a relatively stiﬀ structure.3b). The area under the σ − δ curve represents the energy. The LEFM models.140CHAPTER 5. that the tensile stress at the tip of ﬁctitious crack is equal to the tensile strength of concrete has been modiﬁed using the concept of singular stress distribution at the ﬁctitious crack tip. usually cited to apply LEFM models. Gf . Two apparently diﬀerent models have been proposed in the literature considering only the model I nonlinear fracture propagation in concrete. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS appropriate to consider the nonlinear behaviour in the FPZ if the localization of the crack proﬁle is a primary objective of the ﬁnite element analysis.3a).
CONSTITUTIVE MODELS FOR CRACK PROPAGATION 141 Figure 5.3: Nonlinear fracture mechanics models: (a.(c.5.d) crack band model .3.b) ﬁctitious crack model.
The very small value of wc . The fracture process is thus assumed to propagate as a blunt front (Figure5. is assumed to be a material property in this model.3c). thus. wc . to smeared fracture propagation analysis requires the size of ﬁnite elements to be limited to wc . The values of wc in the range of single aggregate dimension to six times that size gave equally good results. The width of the blunt crack (or the band of microcracks). Bazant and Oh (1983). however.3d) and the fracture energy. Application of the crack band model. Various proposals have been made in the literature about the form of σ − δ and σ − ε softening relationships. attempted to establish the value of wc by ﬁtting their proposed model in a series of stressstrain and fracture energy data collected from the literature. is given by Z εf Gf = wc σ(ε)dε 0 The inherent characteristic of the proposed crack band model is the smeared nature of crack distribution over a band width. the ﬁctitious crack model or the crack band model. in the original presentation. The smeared nature of the crack band model is a tempting feature for application in ﬁnite element analysis when the direction and location of crack propagation are not known a priori. should be used. where the fracture process eﬀects are smeared over a zone of the ﬁnite element and the average stressstrain relationship is adjusted to conserve the fracture energy.142CHAPTER 5. Special ﬁnite element techniques have been proposed to ease this limit on element size. wc . A bilinear relationship is usually applied to interpret the experimental observations (Briihwiler 1990. The strain softening behaviour of concrete in the FPZ is represented by a stressstrain relationship (Figure5. renders any practical ﬁnite element analysis of large concrete gravity dams too expensive. Nomura et al. An important aspect of the nonlinear fracture mechanics models is the shape of the softening branch. Gf . The assumption of fracture energy dissipation over a certain band area of material characteristic dimension. seems to be a numerical speculation. A linear strain softening relationship has been . according to the deﬁnition presented by Bazant and Oh (1983). There has been an unrelenting debate as to which softening model. The precise shape of the softening diagram has been reported to have considerable inﬂuence on the numerical results (ACI 1991). in its original form. 1991). wc . has not been determined by any direct experimental investigation. The crack band width. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS volume of the material.
l2 . CONSTITUTIVE MODELS FOR CRACK PROPAGATION 143 Figure 5. (b) √ characteristic dimension. lc= l0 l00 .3. lc = l1 . (c) characteristic dimension.5.4: (a) average stressstrain curve for smeared crack element.
. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS assumed in most of these models. Et .0MPa. the softening constitutive model degenerates to the traditional elastobrittle failure criterion. the characteristic dimension. becomes stiﬀer for increasing value of lc up to a certain limit. dissipating the stored elastic strain energy instantly upon reaching the tensile strength of material. lc . are known for the material. and σ t = 2. The maximum ﬁnite element size.2.4b.) For oblique crack propagation. lc . A favored approach in the nonlinear smeared fracture models (NSFM) is to adjust the slope of the softening branch.2.2 N/mm.l3 .Gf . denoted by l0 . also has often been considered stringent for large scale ﬁnite element analysis at reasonable cost.4c).4a) . beyond which an unrealistic snapback appears in the tensile stress strain relationship of concrete. Gf .0m. that can be modelled with strainsoftening behaviour is determined from equation 5. is a geometric property of the element. for unit area of crack extension remains independent of the element characteristic dimension. lc = wc (wc is the width of crack band).144CHAPTER 5. in ﬁgure5. wc .2: 2EGf l0 ≤ σ2 t For typical dam concrete properties of E = 30 000 MPa. the characteristic dimension. For cracks parallel to a side of square ﬁnite elements. of Bazant and Oh (1983). In the limit case. σ t .2) if the tensile strength. the nonlinear smeared fracture model reduces to the crack band model of Bazant and Oh (1983). assuming that the softening process initiates when the tensile stress reaches the material tensile strength. as in Figure5. and the strain fracture energy. the limiting value is l0 ≤ 3. the elastic modulus. For a special case of .4a can be derived as: Et = σ2E t σ2 − t 2EGf lc (5. which is much higher than the material characteristic dimension. Et . wc . This limit on maximum dimension. etc. equals the element dimension across the crack plane (lc =l1 . (Figure5. can be determined from equation 5. Gf = 0. The area under the average stressstrain curve for a ﬁnite element undergoing fracture process is adjusted such that the dissipated fracture energy. given by equation 5. The softening modulus. lc . lc . the strain softening modulus for the particular element size. Unlike the crack band width. l2 . The strainsoftening modulus. lc can be taken as the square root of the element area under consideration (Figure5. E.
5.3. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 145 Figure 5.5: Nonlinear fracture mechanics in smeared crack propagation model .
5c). and (ii) when applied with a strength based crack initiation criterion. one proposition is to reduce the fracture initiation stress. as a material property and reduce the fracture initiation stress with increasing size of element (Figure5. and at the same time respect the principle of conservation of energy.3 essentially represents the nonlinear smeared fracture model described earlier. This size reduced strength criterion (or in other words the elastic fracture criterion) can be criticized for two reasons: (i) the size independent critical COD value. of no tensile resistance. and α is the averaging factor deﬁned as the ratio between the crack band area and the gross element area. implied by the conservation of fracture energy in the nonlinear fracture model. Several other models with some variations of the crack band model have also been proposed in the literature.146CHAPTER 5. σ 0 . Gajer . Moreover. The alternate proposition of Bazant (1985) is to take the softening modulus. as a material property has not been justiﬁed from experimental investigations. is violated (Figure5. with increasing ﬁnite element size and assume an elastobrittle failure criterion for element sizes greater than l0 (Figure5. This is the socalled size reduced strength (SRS) criterion proposed by Bazant (1984) and Bazant and Cedolin (1979. Gajer and Dux (1990) decomposed the ﬁnite element strain increment in the following form: ∆ε = ∆εe + α∆εcr (5. An important numerical side eﬀect of the elastobrittle SRS failure criterion is the generation of spurious shock waves in the ﬁnite element model. 1983). taking Et . The limit on the maximum size of ﬁnite elements thus appears to be a requirement to ensure a reliable application of the nonlinear fracture mechanics criteria in smeared crack propagation analysis.3) where the incremental strains ∆εe and ∆εcr correspond to the uncracked concrete and the crack band respectively. the principle of fracture energy conservation is likely to be violated in the interior element as well as in the exterior element (ﬁgure5.5a). This approach may allow numerically stable algorithm as opposed to the elastobrittle model. equation 5. Et . but it seems to be based on a weaker theoretical consideration.5d) for conserving the fracture energy. For decomposition of strain in the fracturing direction. and the questions raised for the elastobrittle criterion are still present. δ f . ElAidi and Hall (1989a) encountered such numerical diﬃculties in the nonlinear seismic analysis of concrete gravity dams. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS To circumvent this limit on the size of ﬁnite elements.5b).
Bazant and Gambarova (1984) proposed a hypothetical nonlinear .6: Closing and reopening of partially formed cracks Application of nonlinear fracture mechanics models in dynamic analysis requires the deﬁnition of unloading and reloading behaviours during the fracture process. where the crack strain is averaged over a characteristic dimension of single element. A crude form of the nonlocal model can be considered to be the nonlinear smeared fracture model. in the global ﬁnite element direction. Figure 5.3. described earlier. CONSTITUTIVE MODELS FOR CRACK PROPAGATION 147 and Dux (1990) apparently ignored the directional property of cracking by applying the scalar factor. Very few studies have been reported in the literature in this respect. In that approach. An enhanced form of the crack band model is the socalled nonlocal cracking model (Bazant and Lin 1988). the crack strain at an integration point is smeared over the neighboring points of the ﬁnite element mesh.5. α.
de Borst and Nauta (1985) and Gajer and Dux (1990) applied the concept of . however.6d). It appears that the techniques applied by de Borst and Nauta (1985) and Gambarova and Valente (1990) are subsets of this generalized model with λ = 0. Gambarova and Valente (1990) retained the initial shear modulus unchanged until the complete fracture had taken place and then applied an aggregate interlock model. 5. and the material stiﬀness matrix was derived for normal strains only.6a. In another proposition. Dahlblom and Ottosen (1990) proposed the following relationship for closing and reopening behaviours of partially fractured concrete: ε = [λ + (1 − λ) σ σ max ]εmax (5.3. This formulation is not compatible with the linear isotropic stiﬀness matrix of initial state. to account for shear in the partially formed crack. propagating as a blunt front of microcracks.6b) to characterize the FPZ behaviour of increasing strain with decreasing stress and a secant formulation (line 20 in Figure5.6c). Numerical simulation of such nonlinear model is very complex.0 and 1.148CHAPTER 5. it is not unlikely to expect the rotation of principal stress directions under varying deformation modes. Bazant and Gambarova (1984) proposed a nonlinear mathematical model.4) where λ is the ratio between the residual strain upon closing of cracks and the strain of open cracks (Figure5.3 Shear resistance of fractured concrete After the initiation of fracture process on a plane perpendicular to the direction of major principal tensile stress. Hence. yet to be investigated rationally.6b) to represent the closing of partially open cracks. Bazant and Oh (1983) ignored the shear deformation on the fracture plane.0 respectively. Gambarova and Valente (1990) have applied an assumption of sudden stress release when the closing of partially open cracks is detected at any instant of the fracture process (Figure5. de Borst and Nauta (1985) applied the linear tangent softening modulus (line 12 in Figure5. The deﬁnition of shear stressstrain behaviour of concrete during the fracture process seems to be diﬃcult. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS stressstrain relationship for closing and reopening behaviours of partially open cracks as depicted in Figure 5. The physical phenomenon of crack closing and reopening taking place before the complete fracture of the material is. shear deformation may take place on the partially formed rough fracture plane. named the crack band microplane model.
proposed by Suidan and Schnobrich (1973). Chen and Schnobrich (1981) Reinhardt and Wairaven (1982). This happens due to the application of continuous shape functions in deriving the ﬁnite element stiﬀness matrix. Dahlblom and Ottosen (1990) assumed a linear relationship between shear resistance and COD. After early opposition to the rotating crack model. to derive the tangent shear modulus of the fracture plane. 5. A special numerical problem associated with the shear retention in smeared crack ﬁnite element model is the spurious tension stiﬀness in the direction across the fracture plane (ElAidi and Hall 1989a). A special numerical technique to represent nonorthogonal multiple crack formation has been perfected by de Borst et al. the orthotropic material reference axis system is rotated when the principal stress direction deviates by a certain amount from the direction that initiated the fracture process. . An additional crack plane is allowed to form only when the stress reaches the tensile strength on the plane orthogonal to the ﬁrst fracture plane. de Borst and Nauta 1985). Discontinuous shape functions proposed by Ortiz et al. In the ”rotating crack” model. This type of model is stated as ﬁxed or stationary crack model . 1991b).4 Postfracture behaviour of concrete The fracture direction is generally ﬁxed based on the principal stress direction that initiates the ﬁrst crack. (1986) for localized failure analysis may be a solution to this problem. Bazant and Lin 1988) adopted the concept based on the argument that the cracks of one direction may close and lock in shear while cracks of another direction may form. The models are very often computationally inconvenient because of the nonsymmetric stiﬀness matrix. (de Borst 1987. (Bazant 1983. proposed by Cope et al. Riggs and Powell (1986). Investigate on the Shear resistance of fractured concrete. Bazant et al. The simpliﬁed approach of applying a shear retention factor ignores the shear dilation and the dependence of crack shear stiﬀness on the crack opening displacement (COD). (1980). The models proposed by Bazant and Gambarova (1980). (1991a.4. β. A comparative study on diﬀerent rough crack models is available in Feenstra et al. POSTFRACTURE BEHAVIOUR OF CONCRETE 149 simple shear retention factor. The rotation of physical crack direction does not seem to be acceptable from common perception. and Wairaven (1981) represent the shear resistance and dilation on crack open surfaces of concrete.5.
150CHAPTER 5. Ideally. and elastic modulus. the inﬂuences of rate sensitive ν may be insigniﬁcant in comparison to the inﬂuences of other material parameters. is the modelling of contactimpact phenomenon occurring upon closing and reopening of the cracks. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS A very special postfracture problem.20 has been applied almost universally in the past studies. such as the one by Briihwiler (1990) and Briihwiler and Wittmann (1990) are revealing signiﬁcant diﬀerences in the mechanical properties of structural concrete and mass concrete. And the consequences of the shockwave phenomenon in smeared crack analysis are not conclusively known yet. a review of literature is presented here to establish a reasonable limit of parametric values. specially under transient conditions. However. ν. Poisson’s ratio. Pekau et al. ElAidi and Hall (1989a) presented a discussion on numerical diﬃculties arising from high velocity closing and reopening of cracks in smeared crack analysis No rigorous procedure has been reported in the literature to deal with shock wave generated by the sudden change in stiﬀness resulting from the closing and reopening of the smeared cracks. A value of 0. However. 5.28.17 and 0. Special numerical techniques to simulate the impact behaviour in discrete crack models have been proposed in the literature (Ayari and Saouma 199 1. 199 1). associated with dynamic analysis. are applied to represent the elastic behaviour of concrete in all analyses irrespective of constitutive models selected for propagation of cracks. A signiﬁcant feature of . E. Briihwiler (1990) observed the reduction of Poisson’s ratio with increasing compressive strain rate applied to the concrete cylinder. the selection of material properties for safety analysis of concrete dams should be dealt with on a casetocase basis. Jansen (1988) suggested the Poisson’s ratio for 1year old dam concrete between 0. because the material properties are widely varying from dam to dam. The static modulus for 1 year old dam concrete was suggested by Jansen (1988) in the range of 28000 − 48000 MPa. Recent experimental investigations.5 Material parameters for fracture propagation analysis The present development of mathematical models is far ahead of the current knowledge of material behaviour. Material parameter data determined from reliable experimental studies is limited in the literature for dam concrete.
324fc2/3 0 0 (5.5.48fc2/3 0 (5. Raphael (1984).6) Briihwiler and Wittmann (1990) observed a dynamic magniﬁcation of up to 80% for the investigated strain rates between 10−5 and 10−2 per second. resulting in .5. Under alternating tensile and compressive loadings.5) where fc and σ t are respectively static compression and tensile strengths of concrete in MPa. From a rigorous study with some 12 000 published test results.level damage may take place in the material. A review of this phenomenon along with the tensile strength property of concrete is presented in the following section. In absence of experimentally determined values. Tensile strength of concrete increases signiﬁcantly with increasing rate of applied loading.1 Strengthofmaterial parameters The governing material parameter in SOMbased fracture propagation models is either critical stress or strain. the dynamic load rate eﬀect has been observed to be higher than that in usual structural concrete. the above equation can be applied as an approximation to the expected static tensile strength of concrete. The selection of dynamic magniﬁcation of the concrete tensile strength is not very evident from the literature. from his study with the published data. and nonlinear fracture mechanics. Raphael (1984) proposed the following relationship between tensile and compressive strengths of concrete under static loading: σ 0t = 0. The shear resistance parameter of fractured concrete is also discussed. but the failure strain remains more or less unchanged under varying load rate.50 resulting in the dynamic tensile strength of concrete: σ t = 0. substantial micro. There is a controversy as to whether the static tensile strength or the dynamic strength should be used in seismic response analysis of concrete dams. In the limited dynamic tests performed on mass concrete. MATERIAL PARAMETERS FOR FRACTURE PROPAGATION ANALYSIS151 concrete constitutive behaviour is the dynamic magniﬁcation of the elastic modulus under rapidly varying loading condition.5. linear elastic fracture mechanics. proposed a dynamic magniﬁcation factor of 1. 5. and this magniﬁcation decreased signiﬁcantly due to compression preloading on the tested specimens. Separate reviews are presented for commonly used material parameters in three major crack propagation criteria strength of material.
Nevertheless. σ a .5. exists about the interpretation of tensile stresses computed from ﬁnite element analysis. the nearpeak stressstrain nonlinearity decreases substantially (Briihwiler 1990. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS reduced dynamic tensile strength. Recognizing the fact that the initial tangent modulus does not increase at . however. NRC 1990). 50% dynamic magniﬁcation proposed by Raphael (1984) seems to have received wide recognition (Chopra 1988. Under dynamic loading. Figure 5. Since the prepeak stressstrain relationship is assumed to be linear elastic in most analyses. Hatano 1960).152CHAPTER 5. Kollgaard 1987.7: Strengthofmaterialbased failure criterion A confusion. some investigators have suggested to compare the predicted tensile stresses with the apparent strength of the material (Figure??a). Experimental evidences seem to support an apparent static tensile strength. about 30% higher than the value given by equation 5.
5.25Ei (5. K1c . And only a handful of experimentally . 5. The behaviour under biaxial loading can be expressed using the standard failure envelopes in principal stress space (Figure??b). and aggregate size.7. Equation 5. MATERIAL PARAMETERS FOR FRACTURE PROPAGATION ANALYSIS153 the same rate as the tensile strength under dynamic loading. above relation gives the apparent dynamic tensile strength.7) Assuming that the failure strain remains the same under static and dynamic loadings.5. Apparently.3 × 0.324)fc2/3 = 0. to be 25% higher than the apparent static strength.8 respects the experimental evidences of rate insensitive failure strain. the tensile strength of concrete under tension. σ a . Thus.Ei . from the static compressive strength. the apparent dynamic tensile strength is only about 80% higher than the actual dynamic strength. In the experimental study by Mlakar et al. the dynamic and static initial tangent moduli respectively: Ei0 = 1. no empirical formula is readily available in the literature for dam concrete to select the initial static tangent modulus.25(1.6 and equation 5. σ 0a . fc0 of concrete. and reduced nearpeak σ −ε nonlinearity under dynamic loading. (1985). elastic modulus.2 Linear elastic fracture mechanics parameters The principal parameter applied in the linear elastic fracture mechanics crack propagation models is the fracture toughness. σ 0a = 1. Many more experimental investigations are required to establish an acceptable mathematical model for the eﬀects of precompression load and the fatigue behaviour under cyclic loading. No deﬁnite relationship is readily available in the literature of dam concrete to determine the fracture toughness from standard material parameters such as strength. of concrete. The apparent tensile strength. the Canadian Electrical Association (1990) report on safety assessment of concrete dams recommended the following relationship between Ei0 and Ei . the failure strain showed the general tendency of rate insensitivity.526fc2/3 0 0 (5. if such joints are not modelled explicitly (NRC 1990).8) Comparing equation 5. however. instead of 30% as proposed by Raphael (1984). may have to be lowered by about 10%o − 20% due to the existence of relatively weak construction lift joints in dams.5.compression loading was observed to increase with increasing rate of loading. εt .
Due to multiaxial conﬁning stresses in the ﬁeld conditions. Those relationships were derived from the results of laboratory experiments performed with small size aggregates speciﬁc of structural concrete behaviour.3 Nonlinear fracture mechanics parameters Fracture energy.5 MPam1/2 . laboratory experiments could be performed on recovered core specimens. Values for σ t and E can be selected according to the guidelines used for strengthofmaterial failure criterion. The fracture toughness can also be estimated from the following wellknown relationship (Irwin 1957): K1c = p Gf E where E is the elastic modulus and Gf the fracture energy. Gf .1 MPam1/2 . Ayari and Saouma (1990) assumed an arbitrary dynamic magniﬁcation factor of 60 for K1c . (1989) found a K1c value of 1. Linsbauer (1990) reported K1c . and tensile strength. from wedge splitting tests (Briihwiler and Wittmann 1990). 5. the in situ values for fracture toughness.5. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS determined results has been reported so far. a value of K1c = 1. σ t . Usually. 1991b). (1990) to select K1c : a zero value as a ﬁrst approximation. K1c . E.154CHAPTER 5. is determined from uniaxial or split cylinder tests and the fracture energy. Oh and Kim 1989). should the response be unacceptable. were determined as three times the unconﬁned laboratory test values (Saouma et al. . The following guideline has been proposed by Saouma et al. beyond which a strain softening process is assumed to take place. values in the range of 2.0 MPa . σ t . to deﬁne the entire constitutive behaviour of concrete in the nonlinear fracture mechanics models. Saouma et al. The strain rate sensitivity of K1c for mass concrete is not well addressed in the literature. is the key parameter that is combined with elastic modulus. Briihwiler (1990) predicted the rate sensitivity of K1c to be lower than that of concrete tensile strength.m 1/2 is used. the tensile strength. which seems too high for concrete. Extrapolation of the test results from structural concrete does not seem to be realistic to establish the behaviour of mass concrete that has distinct features of much larger and weaker aggregates. Gf . 1989. and if this value still results in unacceptable crack lengths. Empirical relationships have been proposed to determine the fracture energy from standard material parameters (Bazant and Oh 1983. In the seismic analysis of Koyna Dam (India).0 − 3.
and a maximum of 80% dynamic magniﬁcation over the pseudostatic value was observed.5. 5.175and 0. the dynamic magniﬁcation criterion selected for tensile strength can be applied to the fracture energy of the nonlinear fracture mechanics model. That means. Briihwiler and Wittmann 1990). The test results reported in the literature are thus sketchy at the present time. and they are often in contradiction to one another. More complex relationships can also be established. Fracture energy values for the specimens subjected to compressive preloading were found considerably low. 1991) reported the increase of fracture energy with increasing aggregate size. The Gf parameter determined under simulated seismic loading rates showed substantial strain rate sensitivity. I 99 1). Experiments performed by Saouma et al. and another study by H. a study performed by H. δ f (Briihwiler 1990). 1991a) have shown that the fracture properties are not aﬀected by the specimen or aggregate sizes used.N. In contrary to this ﬁnding. The physical characteristic of fractured dam concrete. meaning that the Gf of dam concrete is two to three times larger than that of structural concrete. The Gf value under static loading condition was determined to be in the range of 0.5.4 Shear resistance of fractured concrete The fractured concrete derives shear resistance from interlocking of aggregates protruding out of the fracture plane. Briihwiler and Wittmann (1990) attributed the rate sensitivity of Gf mainly to the rate sensitivity of σ t . Linsbauer (Dungar et al. Biaxial and triaxial stressstrain eﬀects on fracture energy dissipation characteristic of concrete are also required to be investigated (Kreuzer et al. including the rate sensitivity of critical crack opening displacement. Mihashi (Dungar et al. Laboratory tests performed at the University of Colorado (Briihwiler and Saouma 1991) showed signiﬁcant reduction of the fracture properties of concrete with increased water pressure along the crack.310 N/mm. Numerous investigations have been conducted to determine the phenomenon of aggregate interlock in structural concrete where the aggregates are usually very strong. (1990.5. 1991) reported larger Gf values for larger specimen sizes. MATERIAL PARAMETERS FOR FRACTURE PROPAGATION ANALYSIS155 Limited results have been reported from experimental investigations on concrete collected from dam construction sites (Briihwiler 1990. The eﬀect of multiple cracks on fracture energy dissipation phenomenon is not known precisely. where the aggregate strength is .
156CHAPTER 5. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS substantially low, is not known clearly. The recent fracture investigation of dam concrete, performed by Briihwiler and Wittmann (1990), have shown that the crack does not travel around the aggregates; it goes straight through them. The roughness on resulting fracture plane, thus, would be mild with low shear resistance. No speciﬁc information is currently available to establish a shear resistance model. Rigorous parametric analyses are required to study the sensitivity of nonlinear seismic response of concrete dams to the,selected shear resistance model.
5.6
NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS
Micro cracking in concrete is believed to occur at relatively low levels of loading. Therefore, cracking progresses in a heterogeneous medium because of an increase in micro cracking, and because of the linking of various micro cracked zones. Experiments performed on cement paste as well as concrete show that micro cracking has an arbitrary orientation. When the load is increased, macroscopic cracks develop and the crack orientations follow the principal stress directions in the material. Proper understanding and mechanical modeling of the damage process of concrete, brought about by the internal defects, is of vital importance in discussing the mechanical eﬀects of the material deterioration on macroscopic behavior. Modeling of this phenomenon has triggered intensive research activities over the past 20 years or so. The main concept of this theory is to represent the damage state of material by an internal variable, which directly characterizes the distribution of micro cracks formed during the loading process. Each damage model established mechanical equations to describe the evolution of the internal variables and the mechanical behavior of damaged material. The damage mechanic model can be divided into isotropic and anisotropic damage models. The isotropic damage mechanics model uses a single scalar parameter and is based on Lenlaitre’s hypothesis of strain equivalence. It was used with some success to describe the damage of concrete. Yet, few practical applications to real structures were conducted using this approach. However, it was experimentally observed that crack growth in concrete structures signiﬁcantly
5.6. NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS157 depends on the direction of the applied stress and strain. Hence, the damage process in concrete is essentially orthotropic, the isotropic description is a mere simpliﬁcation. Compared with the fracture mechanics theory used in the context of discrete cracks the continuum models with ﬁxed mesh have the advantage of avoiding remeshing when ﬁnite elements are adopted. However, when compared with the smeared crack approach there seems to be a small advantage. Researchers have proved that diﬀerent mechanical phenomena can be formulated within the same framework of damage mechanics. The swelling problem is investigated in concrete dams using this concept. Finally, damage mechanics permits the easy implementation of any initial damage due to thermal stresses or any other phenomena, such as alkaliaggregate reactions, acting on an existing dam.
5.6.1
NUMERICAL PROBLEMS RELATED TO STRAIN SOFTENING
Damage process is associated with strain softening, which is usually accompanied by sudden transition from a smoothly varying deformation ﬁeld into a localized band. The treatment of strain softening and localization by ﬁnite elements could lead to serious diﬃculties, as reported by Simo (1989), Bazant and Belytschko (1985), and Wu and Freund (1984). Some of these diﬃculties are: (1) cracks tend to localize in a band that is generally the size of the element used in the discretization; (2) the fracture energy dissipated decreases as the mesh is reﬁned, in the limit it tends towards zero; (3) the solution obtained is extremely dependent on mesh size and orientation. To overcome these diﬃculties, the following solutions were proposed: 1 . The strain softening is related to the element size (Pietruszczak and Mroz 1981). which leads to the meshdependent hardening modulus technique (Bazant and Oh 1983. Pramono and Wiliam 1989; Simo 1989). 2. Nonlocal damage is formulated as proposed by Bazant and PijaudierCabot (1988), Mazars et at. (1991), Saouridis and Mazars (1992). 3. Introduction of viscoplasticity as suggested by Needleman (1988), Loret and Prevost (1991), Sluys and de Borst (1992).
158CHAPTER 5. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS 4. The use of the second gradient theory in the deﬁnition of the strain tensor (sluys et al. 1993). 5. Cosserat continuum approach (de Borst 1991; Vardoulakis 1989). From a purely mathematical viewpoint Simo (1989) showed that the ﬁrst three solutions can be interpreted as regularization procedures of the dissipation function. PijaudierCabot et al. (1988) performed a comparative study of these techniques for the propagation of waves in a longitudinal bar.
5.6.2
FUNDAMENTAL EQUATIONS OF DAMAGE MECHANICS
The concept of the damage mechanics model is based on the dissipation of energy by means of cracking and loss of rigidity of the material. Any induced plastic behavior is ignored. Equilibrium of the system is deﬁned by the thermodynamic potential, which is considered here as the strain energy W (ε, d) and is a function of the strain vector {ε} and the damage variable d. Thus, W (ε, d) can form the basis to formulate the following expressions: σi = ∂W ∂εi and Y =− ∂W ∂d
where σ i is the component of the stress vector in engineering notation; and Y is the thermodynamic force associated with d, interpreted as a damage energy release rate. The time derivative of the speciﬁc energy yields W = {σ}T {ε} − Y d The Clausius Duhem inequality gives a condition on the energy dissipated Wd by the damage process, which can be written as: Wd = Y d= {σ}T {ε}− W ≥ 0
. . . . . . .
(5.9)
The formulation of a damage model ﬁrst requires the deﬁnition of threshold of damage. which is the conditions that initiates the propagation of damage. Secondly, the evolution of the damage with loading must also be deﬁned, and it is a function of a measure of strains, stresses or energy.
5. [M] is a diagonal matrix equal to 1−d I. The eﬀective stresses are deﬁned as: {σ} = [M(d)]{σ} − (5. In equation 5. For isotropic damage model equation 5.10 is valid when the strains in the damaged material are assumed to be equivalent to strains in the virgin material.8C.[M] represents the damage matrix. The stress vector and the rate of energy dissipation can be easily obtained: {σ} = (1 − d)[C0 ]{ε} and 1 ∂W Y = {ε}T [C0 ]{ε} = − 2 ∂d . Figure 5.6. where [C(d)] is a second order matrix of material properties.12) 1 {σ} 1−d (5. it is necessary to relate the damage variable d to the other internal variables by some physical hypothesis.3 ISOTROPIC DAMAGE MODEL FOR CONCRETE To establish the damage constitutive equation. equation 5.10. In such a case. Here we try to brieﬂy describe an isotropic damage mechanics model.6. NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS159 5.10) where[M(d)] is in general a symmetric matrix of rank four.D shows the basis of this 1 assumption. in the context of strain equivalence. Thus. but possess a reduced modulus by a factor (1 − d). and I is identity matrix. The hypothesis of strain equivalence of Lemaitre and Chaboche (1978) is empirical in nature.10 reduces to: {σ} = − The strain energy for a damaged material is W = 1/2{ε}T [C(d)]{ε}. the evolution of damage decreases the material stiﬀness and the following expression can be written [C(d)] = (1 − d)[C0 ] (5. It states that any constitutive equation for a damaged material can be derived from the same potentials as that for a virgin material by for a damaged material by replacing the stresses by eﬀective stresses.11) where [C0 ] is initial matrix of undamaged material properties.
even though these problems are not addressed in this paper. d is always increasing as shown in equation 5. The basic assumption in behaviour of the damaged and equivalent undamaged element will results in diﬀerent model. the material itself is initially orthotropic. Diﬀerent hypothesis have been proposed for isotropic damage model. In the ﬁrst model. .4 ANISOTROPIC DAMAGE MODEL FOR CONCRETE A Gf type anisotropic damage model is described in this part for diﬀerent reasons: when roller compacted concrete is used in dams.160CHAPTER 5. The objectivity of the ﬁniteelement solution is ensured by introducing the nonlocal description of this measure by averaging it within an inﬂuence area in the ﬁniteelement mesh. In this case an extremely reﬁned mesh is required. a single damage variable independent of directions of stresses describes the behavior of concrete. Therefore. The damage evolution is deﬁned as a function of the elastic strain energy of an undamaged equivalent material. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS Since Y is a quadratic function of strain. Ju 1989. In addition. The size of the inﬂuence area is of the order of three times the aggregate size. While in the second concept.9. special attention will be given to the two models. They are mainly based on two concepts of the modelling. 5. the softening parameters are made meshdependent using the energy equivalence concept. EnergyBased Damage Model.6. Chow and Lu 1991). orthotropic damage models allow the modeling of joints in the dam and at the damfoundation interface. This is a characterization of the irreversibility of damage. This technique leads to reasonable mesh size and the model is suitable for the analysis of large concrete structures. it is positive and. This variable is calculated based on a certain measure of the material’s strain ﬁeld. StrainBased Damage Model. the damage is described by coupling the compression and tension eﬀects to deﬁne a single damage variable d. thus. Among several models.element solution. To ensure mesh objectivity of the ﬁnite. it becomes impractical to model real dams with a reasonable size mesh since at least three elements are required within the inﬂuence area of about 300900 mm. the isotropic model becomes a special case (Chow and Wang 1987. If the aggregate size varies between 100 and 300 mm. It will be shown that by developing an orthotropic damage model. the development of an anisotropic damage model is essential.
A)eﬀective areas for isotropic and anisotropic damages. D)stressstrain curve for equivaalence hypothesis.5. F)initial damage formulation .8: Material model in the damage mechanics concept.6. B)characteristic length. C)strain equivalence hypothesis. NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS161 Figure 5. E)closingopening criterion.
If the concept of net area is still considered. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS When isotropic damage is considered. 2. Chow and Lu (1991) and Valliappan et al.13) . (1990). They are based on the principle of elastic energy equivalence. the eﬀective stress {σ} (Lemaitre and Chaboche 1978) and the elastic stress {σ} are related by 5. The index i(1.8A. If damage is no longer isotropic. material anisotropy introduces diﬀerent terms on the diagonal of [M(d)]. 3) corresponds with the Cartesian axes x. This principle postulated that the elastic energy in the damaged material is equal to the energy of an equivalent undamaged material except that the stresses are replaced by eﬀective stresses. If the symmetrized eﬀective stress vector is deﬁned as: s ∗2 ∗2 σ 12 + σ 21 − − − − ∗ ∗ } {σ} = {σ 1 σ 2 σ 12 } = {σ 1 σ 2 2 It can be related to the real stresses by: σ1 − 0 σ2 = − 0 σ 12 − 1 1−d1 0 1 1−d2 0 0 0 r ³ 1 2 1 (1−d1 )2 + 1 (1−d2 ) σ1 ´ σ2 σ 12 2 (5. and Ω∗ is lost area i resulting from damage. In this case the ratio of the net area over the geometrical area may be diﬀerent for each direction.162CHAPTER 5. as shown in Figure 5. results in a nonsymmetric eﬀective stress vector.y and z. ∗ The relation between the eﬀective stresses {σ} and the elastic equivalent stresses {σ} becomes: ∗ 1 σ1 0 0 1−d1 ∗ 0 1 σ1 0 σ2 1−d2 σ2 ∗ = 1 0 0 σ 12 1−d2 σ 12 1 ∗ 0 0 σ 21 1−d1 di = − In this case the eﬀective stress tensor is no longer symmetric and an anisotropic damage model. because of cracking.11. based on equivalent strains. the deﬁnition for di becomes: Ωi − Ω∗ i Ωi where Ωi is tributary area of the surface in direction i. Various attempts to restore symmetry were proposed by.
6.. It can be written in a manner similar to that of the smeared crack approach. 3) in equation 5.16 represents the shear resistance of the damaged material. NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS163 This equation can be written as {σ} = [M]{σ}. 3) = βG in which β is the shear retention factor that is given here as a function of damage scalars d1 and d2 . Cd (3.14) Wd = {σ}T [Cd ]−1 {σ} 2 The elastic strain energy for the equivalent undamaged material is given by: 1 − − e (W0 )equivalent = {σ}T [C0 ]−1 {σ} 2 − − (5.16) The term Cd (3.15) Equating equations 5.5. β can be expressed as a function of strain in the principal direction. where[M] is damage matrix. the constitutive relation can be written as: .14 and 5. The elastic strain energy stored in the damaged material is equal to: 1 e (5.13 yields: [Cd ]−1 = [M]T [C0 ]−1 [M] which results in the following expression for the eﬀective plane stress material matrix: E (1 1−ν 2 Eν (1 1−ν 2 (5.15 and substituting for {σ} from equation 5. Transforming to the global axes. β= 2(1 − d1 )2 (1 − d2 )2 (1 − d1 )2 + (1 − d2 )2 [Cd ] = 0 − d1 )2 − d1 )(1 − d2 ) Eν (1 1−ν 2 E (1 1−ν 2 0 − d1 )(1 − d2 ) 0 − d2 )2 0 2G(1−d1 )2 (1−d2 )2 (1−d1 )2 +(1−d2 )2 If d2 is neglected as it assumed in the smeared crack approach.e. i. The constitutive law can be written as: {σ} = [Cd ]{ε} where the principal directions of damage are assumed to coincide with the principal stresses.
If d1 = d2 = d. the following relation is obtained: [Cd ] = (1 − d)2 [C0 ] which represents the damaged isotropic model.164CHAPTER 5.. If a point At on the stress strain curve (σ. The total dissipated energy is calculated as: Z Wd = dWd = Area(OA1 A2 O) The upper bound of Wd is total available energy of the material gt given by: . and [R] is transformation sin2 β sin βcosβ cos2 β − sin βcosβ 2 2 2 sin βcosβ cos β − sin β and β is angle of principal strain direction. Consider the elastic brittle uniaxial behavior as shown in Figure 5. the ﬁrst law of thermodynamics requires: dW = dWe + dWd where dWe is elastic energy variation.5 EVALUATION OF DAMAGE VARIABLE Now we try to relate the damage variable to the state of an element using uniaxial behaviour of a concrete specimen. a certain amount of energy is dissipated (dWd ). because of damage.6. Under conditions of inﬁnitesimal deformation and negligible thermal eﬀects. and dWd is energy dissipated by damage. ε) moves to A2 . It can be seen that after damage the strain will reach to its original strain at zero. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS [Cd ]G = [R]T [Cd ][R] Where[Cd ]G is material matrix in matrix deﬁned by: cos2 β [R] = sin2 β −2 sin βcosβ the global axes.12 by a square factor because of the energy equivalence.9. 5. This model diﬀers from equation 5.
6.9: Stressstrain curve for energy dissipation due to fracture . NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS165 Figure 5.5.
the eﬀective material modulus is: 2 E = (1 − d) E0 − 5. and the elastic energy in the virgin material W0 is: e e e Wd = (1 − d)2 W0 = (W0 )equivalent (5.20) .19) Therefore.18) and lch characteristic length of the element of volume representing the material’s average behavior. This volume will be characterized by its length which provides a measure of the region over which the damage is smeared so that the global response is reproduced by this volume. the relation between the strain energy stored in the e e damaged material Wd . a possible expression for damage is: s e Wd d=1− e W0 0 (5. For uniaxial loading.166CHAPTER 5. is called characteristic length and it should be measured in the direction normal to a potential crack plane ( Figure 5.17) 0 The fracture energy per unit surface Gf is deﬁned by: Gf = lch gt (5.19. To deﬁne the damage evolution of concrete. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS Z x gt = σdε (5.8B).6 Damage evolution for concrete subjected to tensile strain An element of volume of the material which can be representative of the global behaviour is now considered. The initial threshold is the strain beyond which damage can occur and is given by: f ε0 = t E0 Using equation 5.6. The principal strains are measured for an element to determine the state of the system. lch . This length.
8D can show that: s ¶ µ ε0 ε − ε0 ³ ε0 ´ − d=1− ε εf − ε0 ε d=1− The fundamental issue of this approach lies in the introduction of a geometrical factor.6. and therefore e d = 0. if 2EG lch ≥ f 02 f . 5.21. Condition in equation 5. This technique was proposed by Pietruszcak and Mroz and was employed by a number of authors.23) ε If a simple linear softening curve is assumed.20 and 5. In the limit of damage. which implies that Wd = W0 . When the ﬁnite element method is used. the evolutionary model for damage can be expressed as: ε0 [2 exp(−b(ε − ε0 )) − exp(−2b(ε − ε0 ))] (5. This leads therefore to a reasonable r .5. which implies d = 1. (1989): σ(ε) = ft [2 exp(−b(ε − ε0 )) − exp(−2b(ε − ε0 ))] 0 (5. the meshdependent hardening technique is the most practical for mass structures such as dams.22 should be interpreted as a localization limiter on the characteristic length of the volume representing the global behaviour of the material. A simple way to evaluate d is to adopt a postpeak stress function.18. When the deformations are less than the initial threshold (ε0 ) the material is elastic e e and all the energy is recoverable. Figure 5. and 5. In other words. a socalled meshdependent hardening modulus is obtained. the strainsoftening curve of concrete must be concave. which leads to: b= 3 2G E ε0 ( l ff 0 ch t − 1) ≥0 (5. NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS167 e where Wd is recoverable elastic energy of the damaged material. in the constitutive model. Using equation 5.17. Amongst many strategies used to ensure mesh objectivity.22) Using 5. Gf = 200 N/m and 0 ft = 2MPa yields a limit for lch = 3 m. lch .18 ensures conservation of the energy dissipated by the material.21. using average values for E = 30000 MPa. For t mass concrete. By considering an exponential function as proposed by Lubliner et al.21) The constant b can be evaluated using equations 5. it is not possible to develop strain softening in the volume. Wd → 0. According to Rotls (1991).
The damage model presented here is based on three parameters: the elastic modulus E.168CHAPTER 5. The value λ = 0. the fournode isoparametric element is preferred in the implementation of the local approach of fracturebased models and has .7 Opening and closing of the crack and initial damage When the strain is increasing. 5. The classical split of the total strain in a recoverable elastic part εe and an inelastic strain εin gives: ε = εe + εin = εe + λεmax where εmax is the maximum principal strain reached by the material and λ is a calibration factor varying from 0 to1.8 ANALYTICAL PROCEDURES IN A FINITE ELEMENT MODEL In a ﬁnite element model.24) When the principal strain is less than εin the crack is considered closed. During cyclic loading the strains are reversed and unloading will occur. Under compression the material recovers its stiﬀness. Alternatively. the crack will open when the principal strain is greater than λεmax .2 is selected. If the element of volume is initially damaged (d = do). the secant modulus and the fracture energy are reduced by (1 − do) so the eﬀective values are (Figure 5. Figure 5. damage will also increase. the unloadingreloading stiﬀness becomes : Eunl = E0 (1 − d)2 (1 − λ) (5. the initial strain threshold ε0 and the fracture energyGf . The characteristic length can be related to the Fracture Process Zone (FPZ) commonly used in ﬁctitious crack model for concrete. Experimental cyclic loading of concrete in tension shows evidence of permanent strain after unloading.8E illustrates this criterion. The introduction of such parameter is not for numerical convenience.8F): E 0 = (1 − d0 )E G0f = (1 − d0 )Gf 5.6. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS mesh size for dam models.6.
The stresses at each integration point are then computed using the matrix [Cd ] and the individual strains. ε2 set ε = Max(ε1 . Loop over elements e = 1.5. In the reloading regime the last damage calculated before unloading is used again. Iselect = e (d) {σ}e = [Cd ]{ε}e R (e) Compute the internal force vector re = Ae [B]T {σ}e dA f) Assemble the element contribution r ← re 4. Compute the average strains within the element and the corresponding principal strains: ε1 . For an eﬃcient control of the damage propagation in the ﬁnite element mesh. Initialization Iselect = 0. The average of the strain at the four Gauss points is obtained and the damage is evaluated from the corresponding principal strains of the element. The constitutive matrix [Cd ] is updated depending on the opening/closing and the damage state of the element. Finite element implementation of the damage mechanics 1. some adjustments have to be considered. The standard local deﬁnition of damage is modiﬁed such that it refers to the status of the complete element. Set damage parameter d = dold 2. using the square root of its total area. Check for loading/unloading and closing/opening of crack: (a) if (ε ≥ εmax ) the element is in a loading state . ε2 ) 3.24 until closing of the crack. When the crack closes the material recovers its initial properties. (New element damaged) (a) Correction of stresses due to damage (b) Update data base of damaged elements Subroutine UPDATE (Stress and damage update) 1. Emax = 0 3. nel (a) Compute deformations for each Gauss point : {²}gp = [B]gp {u}e e (b) If the element is already damaged call subroutine UPDATE (c) If not: (selection of element with largest energy density) Compute ε1 . The characteristic length of the element is calculated approximately. During the softening regime the stiﬀness is reduced as a consequence of damage . NONLINEAR MODELLING OF CONCRETE DAMS USING DAMAGE MECHANICS169 been used for the implementation of the described constitutive model.23.equation 5. if ε1 ≤ ε0 go to (d) Compute Ee = 1 {σ}e {ε}e 2 if Ee ≥ Emax then Emax = Ee . If (Iselect 6= 0). At the unloading stage the secant modulus is calculated using equation 5. Compute total displacement {u}i+1 = {u}i + {∆u} 2.6.
23 update if ²in .10a]. In the postpeak strain softening phase.7.7. A linear elastic relationship is assumed between compressive stresses and strains. and (iv) the softening. (ii) the criterion for softening initiation. 5. The tensile stresses and strains are referred to as positive quantities in the presentation. 5. and the ﬁnite element implementation of the formulations are presented in the following sections. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS Calculate d from equation 5. closing and reopening of cracks.2 Strain softening of concrete and the initiation criterion The stressstrain relationship for concrete becomes nonlinear near the peak strength [Figure 5. Compute[Cd ] using equation 5.16 5. E is the elastic modulus. and ν is the Poisson’s ratio. coalescence .1 Prefracture behaviour Stresses {σ} and strains {ε} in a linear elastic condition are related as: {σ} = [D]{ε} where [D] is the constitutive relation matrix deﬁned for an isotropic plane stress condition as: 1 ν 0 E ν 1 0 [D] = 1 − υ2 0 0 1−ν 2 Here. (iii) the fracture energy conservation.7 CONSTITUTIVE MODEL FOR SMEARED FRACTURE ANALYSIS The constitutive model deﬁning (i) the presoftening material behaviour.170CHAPTER 5. εmax = ε (b) else if ε ≥ εin ) d = dold (c) else d = 0 4.
5. In ﬁnite element analyses.10: Constitutive modelling for smeared fracture analysis. The area under the uniaxial stressstrain curve up to the peak is taken as the index for softening initiation: Z ε1 σ2 Eε2 σ i εi i U0 = = i = σ1 > 0 σdε = 2 2E 2 0 where σ i is the apparent tensile strength. 1 σ 1 ε1 (σ 1 . a)softening initiation criterion. and ε1 are the major principal 2 . that may be approximately taken 2530% higher than the true static strength.10a]. It is calibrated in such a way that a linear elastic uniaxial stressstrain relationship up to σ i will preserve the value U0 [Figure 5. d)closing and reopening of cracks of the microcracks causes a gradual reduction of the stress resistance.7. a linear elastic relationship is assumed until the tensile strain energy density. σ t . e)local axis system. CONSTITUTIVE MODEL FOR SMEARED FRACTURE ANALYSIS171 Figure 5. b)fracture energy conservation.
The strainrate eﬀect on the material parameter U0 is considered through a dynamic magniﬁcation factor. as shown in Figure 5. and the principal strain.10b].25 is re0 placed by the corresponding dynamic value. 5. At the instant of softening initiation under a dynamically applied load. respectively). Under dynamic loads. Under dynamic loading. the material parameter U0 in equation 5. are designated by σ 00 and ε00 .3 Fracture energy conservation The tensile resistance of the material is assumed to decrease linearly from the presoftening undamaged state to the fully damaged state of zero tensile resistance [Figure 5. U0 : 1 σ2 σ1 > 0 U0 = σ 1 ε1 = i (5. the biaxial eﬀect in the proposed strain softening initiation criterion is expressed as: r σ1 σ1 = σi Eε1 The above equation is a representative of a biaxial failure envelope. the principal stress. a 20 percent dynamic magniﬁcation of the apparent tensile strength is considered adequate for seismic analyses of concrete dams. as follows: 0 U0 = σ 02 i = (DMFe )2 U0 2E where the primed quantities correspond to the dynamic constitutive parameters. The slope of the softening curve is adjusted such that .10b]. respectively [Figure 5. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS stress and strain. at the instance of softening initiation. ε1 .172CHAPTER 5. and the principal strain. σ 1 . The principal stress σ 1 . DMFe . U0 .10a].7.25) 2 2E Taking the square roots of both sides. the prepeak nonlinearity decreases with increasing values for both σ t and εt [Figure 5.10b. and ε0 . becomes equal to the material parameter. The increased material resistance due to inertia and viscous eﬀects under dynamically applied loads has been explicitly considered in the dynamic equilibrium equations. respectively. Reviewing the literature on the dynamic fracture behaviour of concrete. ε1 . are designated by σ 0 .
The constitutive matrix. CONSTITUTIVE MODEL FOR SMEARED FRACTURE ANALYSIS173 the energy dissipation for a unit area of crack plane propagation. Gf . in the direction normal to a fracture plane and E. In ﬁnite element analyses. and ε2 .26) η= 2 1 − ηυ E 0 0 µ 1−ην 2(1+ν) 5. the ﬁnal strains of no tensile resistance for static and dynamic loading are deﬁned as: 2Gf εf = σ 0 lc ε0f 2G0f = 0 σ 0 lc where lc is the characteristic dimension deﬁned in the previous section.7. referred as the local axis system. The following options are considered with respect to the orientation of crack bands in ﬁnite element analyses.5 Coaxial Rotating Crack Model (CRCM) The local axis system n − p is always kept aligned with the directions of principal strains. respectively. ε1 . After the softening initiation. DMFf .10c. is conserved. In this model. the strains εn and εp are.5. The material reference axis system. a smeared band of microcracks is assumed to appear in the direction perpendicular to the principal tensile strain.4 Constitutive relationships during softening where the parameter η (0 ≤ η ≤ 1) is the ratio between the softening Young’s modulus En . relating the local stresses to local strains is deﬁned as: η ην 0 E En ην 1 0 [D]np = (5. . DMFf can therefore be assumed to be equal to DMFe . is aligned with the principal strain directions at that instant [directions np in Figure 5. is the initial isotropic elastic modulus. and µ is the shear resistance factor.10d]. The strainrate sensitivity of fracture energy is considered through a dynamic magniﬁcation factor. [Figure 5.7.7. applied to magnify the static fracture energy: G0f = DMFf Gf The strainrate sensitivity of fracture energy can mainly be attributed to that of tensile strength. 5.
10c]: sin2 θ sin θcosθ cos2 θ cos2 θ − sin θcosθ [T ] = sin2 θ 2 2 −2 sin θcosθ 2 sin θcosθ cos θ − sin θ . is diﬀerent from the usual formulations where only the crack normal strain is often considered as the damage index. µ. is similar to the formulation presented by BaZant and Oh (1983). the local reference axis system is ﬁrst aligned with the principal strain directions at the instance of softening initiation. maintains a backward compatibility with the presoftening isotropic elastic formulation when η = 1 and 2.174CHAPTER 5. that takes account of deformations in both lateral and normal directions to a fracture plane. NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS ε1 . the parameter µ is deﬁned for the CRCM as follows: 1+ν µ= 1 − ην 2 µ ηεn − εp − ην εn − εp ¶ 0≤µ≤1 (5.26. The local constitutive relationship matrix. respectively. The present formulation. The shear resistance factor.6 Fixed Crack Model With Variable Shear Resistance Factor (FCMVSRF) In this model. The deﬁnition of a variable shear resistance factor according to equation 5. with the degraded shear modulus term. The total stressstrain relationship matrix. except that they have not considered shear deformations in the constitutive relationship. 5.7. and ε2 at the newly oriented material reference state. [D]np .27. εn and εp is the normal strain components in the directions normal and parallel to the fracture plane. is derived using the strain components εn and εp corresponding to the ﬁxed local axis directions (which are not necessarily coaxial with the principal stress directions). can be transformed to the global coordinate directions as follows: [D]xy = [T ]T [D]np [T ] where [T ] is strain transformation matrix deﬁned as follows in terms of the inclination of the normal to it crack plane.27) Here. and kept nonrotational for the rest of an analysis. deﬁned in equation 5. Using an implicit deﬁnition of the softened shear modulus in cracked elements. θ [Figure 5.
For µc ≈ 1. remains unchanged. remains unchanged. With the reduction Of εn . εn . When εn > 0 in subsequent load steps. and hence the parameters η and µ decrease gradually. the residual strain upon closing of a crack is given by εn = νε2 . the value µ is determined by using the damaged value of η to determine the reopening of cracks. which is always kept aligned with the directions of principal strains to keep the principal stresses and strains coaxial. which is always kept aligned with the directions of coaxial principal stresses and strains. During unloading/reloading. the secant modulus. when the strain. The softened Young’s modulus in the direction n. E. is reused in equation 5.95.7 Closing and reopening of cracks Under reversible loading conditions. εmax [Figure 5. 5. Parametric analyses have shown that the seismic fracture response of concrete gravity dams is not aﬀected by a value of µc between 0.10f]. En . the shear resistance factor. CONSTITUTIVE MODEL FOR SMEARED FRACTURE ANALYSIS175 With an increasing strain softening. a change in the global constitutive relationship may also be caused by a rotation of the local axis system.90 and 0. may change during that process.26 is updated as the parameters η and µ change their values. εmax [Figure 5. gradually increases. in an element may alternatively increase and decrease. The constitutive matrix in equation 5. εn . En (which may have reached a zero value). A relatively ﬂexible tolerance. During unload ing/reloading. the secant modulus. En . The appropriate value of the damage modulus. can be used to minimize spurious stiﬀness changes during the closing of cracks. however. if the parameter µ is greater than a threshold value µc . the parameter µ. µc = 0. is replaced by the undamaged initial value.10d sshows the closingreopening behaviour for a special s . the tensile strain.9999. εn . If µ becomes less than µc . En . the element behaviour is determined by either the reloading or the reopening path depending on the ﬁnal state attained in previous tension cycles. is less than the previously attained maximum value. however.26 at that state.5. The change in global constitutive relations is also caused by a rotation of the local axis system. and may eventually reach zero values after the complete fracture (εn > εf or ε0f ). (Figure 5. the damaged Young’s modulus. Figure 5. is less than the previously attained maximum value.7. En .10d). µ. In the CRCM. when the strain. the parameter µ.7. changes during that process.10d]. The CRCM is very eﬀective in alleviating the stress locking generally observed in ﬁxed crack models.
NONLINEAR FRACTURE MODELS OF CONCRETE GRAVITY DAMS case when εn ≈ 0.176CHAPTER 5. .
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.