Hydraulic transients in pressurized pipe flow

(Advanced Water Transport and Distribution)

David Ferràs

Departament of Environmental Engineering & Water Technology,

Water Supply Engineering group

IHE-Delft

March 14, 2018

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Definitions and basic concepts . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Time scales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Classification of flow regimes . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Water-hammer wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Rigid column theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 Derivation of equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4 Water-hammer theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.1 Derivation of equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.2 Simplified equations and numerical approaches . . . . . . . . . . . . . . . . 14
5 The method of characteristics (MOC) . . . . . . . . . . . . . . . . . . . . . 15
5.1 Characteristic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6 Surge protection measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.1 Surge tank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.2 Hydropneumatic vessel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.3 Valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
6.4 By-passes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

1
 2

6.5 Unconventional devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

6.5.1 Flywheel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
6.5.2 Breakable fittings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.5.3 Intermediate flexible pipes . . . . . . . . . . . . . . . . . . . . . . . . 25
7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1 Introduction 3

1 Introduction

A transient flow is the intermediate-stage flow, when the flow conditions are changed, from
one steady-state condition to another steady-state (Chaudhry, 2014). Valve manoeuvres,
starting or stopping of pumps, load acceptance or rejection of turbines, and a number
of other operations and processes occurring in pipe systems may induce fluid transients
which analysis is not possible by the steady or quasi-steady state approaches. When fluid
compressibility and pipe-wall elasticity become relevant, more advanced tools are required
for accurately describing unsteady flows in pipe conduits.
The present chapter aims at introducing the fundamentals of the classic water-hammer
theory. First basic definitions are described and a classification of flow regimes is provided.
Then, generalized forms of mass and momentum conservation equations are derived by
means of the Reynolds transport theorem, a simplified version is reached and expressions
for the wave speed are presented. After a brief explanation of the different numerical
approaches, the well-known Method of Characteristics (MOC) is explained and used to
solve the water-hammer equations. A section is devoted for surge protection methods,
which are explained by means of practical examples. Finally, a number of exercises are
suggested with the goal to better understanding the presented concepts.

2 Definitions and basic concepts

2.1 Time scales

The unsteadiness of pipe flow depends on: (1) the time scale of the transient event, which is
the time lag between the initial and the final steady states; (2) the time scale of the system
response, related to the period of the system vibration; (3) the time scale of the transient
excitation, which refers to the duration of the disturbance that causes the transient event.

2.2 Classification of flow regimes

Different flow regimes can be assumed according to the three previous time scales:

• Steady flow: velocity is a constant value during all the considered period (i.e.,
V (t) = k). The representative time scale is as well t = ∞.

• Quasi-steady flow: velocity varies slowly enough to assume the conditions are
identical to those of a steady flow with the same instantaneous mean velocity. Rep-
resentative time scale: 2D/(f V )  t < ∞.

• Unsteady flow: two main approaches can be applied considering fluid inertia,
compressibility, and pipe-wall elasticity.

– Rigid water-column theory: in this type of flow inertia significantly impedes

changes in velocity, but compressibility of the fluid can be ignored. The velocity
2 Definitions and basic concepts 4

along the pipe can be assumed uniform (as long as the pipe cross-section is
constant). Representative time scale: t ∼ 2D/(f V ).

– Elastic water-column theory: for highly accelerated flow, fluid compress-

ibility may not be negligible and non-uniformity of the flow velocity along the
pipe must be considered. These conditions occur for representative time scales
of t . L/a, where a is the wave speed of the transient wave and L the rep-
resentative length scale which frequently corresponds to the distance between
extreme boundaries or the pipe length. Elastic water-column theory is broadly
known as the water-hammer theory.

2.3 Water-hammer wave

The physical phenomenon
Water-hammer waves are generated when the fluid at a certain steady-state is forced to
stop suddenly. They represent a extreme case of a transient event. In a simple scheme
of a reservoir-pipe-valve system, the sequence of events following valve closure may be
divided into four parts (Fig. 1): (1) first the flow velocity at the valve is reduced to zero as
soon as the valve is completely closed, this increases the pressure at the valve; (2) as the
wave reaches the upstream reservoir, pressure at a section on the reservoir side is constant
this generates a negative flow re-establishing the steady state pressure and a negative
wave travels towards the valve; (3) however, as the valve is closed, a negative velocity
cannot be maintained, this produces a pressure drop, hence a negative pressure wave is
sent out towards the reservoir; (4) as soon as this negative wave reaches the reservoir, an
unbalanced condition opposite to the stage-2 is created again at the upstream end closing
the iterative cycle. Water-hammer waves are, therefore, stationary.
2 Definitions and basic concepts 5

Fig. 1: Representation of the main stages of a water-hammer wave propagating throughout

a reservoir-pipe-valve system.

Wave speed
Wave speed, or wave celerity, is the velocity in which water-hammer waves propagate. This
propagation velocity depends on the bulk modulus of compressibility, the elastic properties
of the pipe-wall and the external constraints of the conduit. The general expression for
the water-hammer wave velocity is presented in Halliwell’s equation:
s
K
a= (1)
ρ[1 + (K/E)ψ]
in which ψ is a dimensionless parameter that depends upon the elastic properties of the
conduit. Following are presented the expressions for thin-walled conduits (e < D/10)

• Rigid conduits: ψ = 0
D
• Conduit anchored against longitudinal movement: ψ = e (1 − ν 2)
D
• Conduit anchored against longitudinal movement at the upper end: ψ = e (1−0.5ν)
D
• Conduit with frequent expansion joints: ψ = e

Wave period
The wave period depends on the celerity and the travelling distance of the propagating
water-hammer wave. For instance, in reservoir-pipe-valve systems the wave have to travel
3 Rigid column theory 6

two times back and forth the pipe to close a cycle (v.s. Fig. 1), while in valve-pipe-valve
systems it is the half. Hence, the wave period in a reservoir-pipe-valve system is:

4L
T = (2)
a

Wave amplitude
First presented in shock waves theory and later adapted for pressurized conduits, the
well-known Joukowsky equation relates pressure to velocity changes:

∆p = ρa∆V (3)

Which in meter of head is

a∆V
∆H = (4)
g
Equation 3 has been extensively used in pipe system design due to the easy applicability
and good accuracy for estimating maximum pressures for fast hydraulic transients. How-
ever, the description of the propagating transient wave throughout the hydraulic system is
often required either at the design stage, operation or even for post-accident analyses. For
an efficient protection of pipe systems against pressure surges accurate numerical models
are required.
Figure 2 depicts the pressure history measured at the downstream section of the ideal
frictionless water-hammer wave propagating throughout a reservoir-pipe-valve system.

Fig. 2: Ideal frictionless water-hammer wave.

3 Rigid column theory

3.1 Assumptions
As mentioned in the previous section, rigid water-column theory assumes that the pipe-
walls are rigid and the liquid is incompressible. The theory solves the final steady state
3 Rigid column theory 7

for a perturbation caused in a pipe system, hence it is useful to determine the effects
of slow transient excitations (t ∼ 2D/(f V )). For faster transient excitations (e.g. fast
valve manoeuvres), elastic water-column theory (water-hammer theory) is required. For
an extended explanation of the rigid water-column theory see Parmakian (1963).

3.2 Derivation of equations

Momentum conservation equation
Consider the reservoir-pipe-valve (R-P-V) system depicted in Fig.3.

Fig. 3: Forces acting on a rigid water-column of a reservoir-pipe-valve system.

The momentum equation is derived by considering that the water-column contained

in the R-P-V system as a rigid-solid, where balance of forces and Newton’s law of motion
are applied. Hence, with the balance of forces at the left-hand-side of the equation, Eq. 5
shows the Newton’s law of motion applied to the water-column.

dV
ρgA (H0 + Ha ) − ρgA (H0 − Z1 ) − ρgAL sin α = −ρAL (5)
dt
Since L sin α = Z1 , rearranging Eq. 5 the momentum conservation equation of rigid
water-column theory is obtained (Eq. 6)

L dV
Ha = − (6)
g dt

Mass conservation equation

Steady-state flow through a valve discharging into air may be written as:
p
Q0 = (Cd Av )0 2gH0 (7)
4 Water-hammer theory 8

in which subscript 0 indicates steady-state conditions, Cd is the coefficient of discharge,

and Av area of the valve opening. Equation 7 can written also as:
p
V0 = B0 H0 (8)
where, √
(Cd Av )0 2g
B0 = (9)
A
Then at any instant during the valve manoeuvre,
p
V = B H0 + Ha (10)
and r
V B Ha
= 1+ (11)
V0 B0 H0
Defining τ = B/B0 ,
r
Ha
V = τ V0 1+ (12)
H0
where τ is a function of time which defines the ratio of the effective gate opening at any
time during the gate closure to the effective gate opening during the initial steady-state.
Equations 6 and 12 are the basic transient equations for valve closure as defined by
the rigid water-column theory.

4 Water-hammer theory

4.1 Derivation of equations

In water-hammer theory, or elastic water-column theory, the fluid inertia and its compress-
ibility, together with the pipe-wall elasticity is taken into account in order to describe the
unsteady flow occurring during transient events. Mass and momentum conservation prin-
ciples are applied to derive the governing equations of the 1D water-hammer theory.

Reynolds transport theorem

In Reynolds transport theorem the flow variables for a specified quantity of fluid mass,
called system, are related to that of a specified region, called control volume. In the present
analysis the control volume is assumed to remain at a location (Eulerian approach), and
volume changes are allowed with respect to time.

Fig. 4: Schematic of control volume.

4 Water-hammer theory 9

An intensive property β of a fluid is defined in terms of its extensive property B as

follows:
∆B
β = lim (13)
∆m→0 ∆m

And the total amount of the extensive property B in the control volume is:
Z
Bcv = βρd∀ (14)
cv

in which m = mass, ρ = mass density, and f ∀ = differential volume of the fluid.

Based in fig 4 and assuming 1D flow the following balance can be written:

Bsys (t) = Bcv (t) + ∆Bin

Bsys (t + ∆t) = Bcv (t + ∆t) + ∆Bout (15)

The time rate of change of property B of the system is

dBsys Bsys (t + ∆t) − Bsys (t)

= lim (16)
dt ∆t→0 ∆t
Substituting expressions for Bsys from Eq. 15 into Eq. 16 and rearranging yields

dBsys Bcv (t + ∆t) − Bcv (t) ∆Bout ∆Bin

= lim + lim − lim (17)
dt ∆t→0 ∆t ∆t→0 ∆t ∆t→0 ∆t

The first term on the right hand side of Eq. 17 represents the time rate of change of
property B in the control volume, which taking into account Eq. 14 can be written as:

Bcv (t + ∆t) − Bcv (t)

Z
dBcv d
lim = = βρd∀ (18)
∆t→0 ∆t dt dt cv
And the second and third terms on the right hand side of Eq. 17 are:
∆Bout
lim = [βρA(V2 − W2 )]
∆t→0 ∆t
∆Bin
lim = [βρA(V1 − W1 )] (19)
∆t→0 ∆t

where A = cross-sectional area of the conduit V = flow velocity and W = velocity of the
control surface (see fig. 5).
Substituting Eqs. 18 and 19 into Eq. 17 the Reynolds transport theorem takes the
following form:
Z
dBsys d
= βρd∀ + [βρA(V2 − W2 )] − [βρA(V1 − W1 )] (20)
dt dt cv
4 Water-hammer theory 10

Mass conservation equation (continuity equation)

The extensive property for mass conservation analysis is obviously mass, hence the inten-
sive property is

∆m
β = lim =1 (21)
∆m→0 ∆m

Fig. 5: Control volume sketch.

And since dMsys /dt = 0 Eq. 20 becomes

d x2
Z
ρAdx + ρ2 A2 (V2 − W2 ) − ρ1 A1 (V1 − W1 ) = 0 (22)
dt x1

Applying Leibnitz’s rule to the first term on the left gives

Z x2
∂ dx2 dx1
(ρA)dx + ρ2 A2 − ρ1 A1 + ρ2 A2 (V2 − W2 ) − ρ1 A1 (V1 − W1 ) = 0 (23)
x1 ∂t dt dt

Noting that dx2 /dt = W2 and dx1 /dt = W1 , the previous equation simplify to
Z x2
∂
(ρA)dx + (ρAV )2 − (ρAV )1 = 0 (24)
x1 ∂t

And based on the mean value theorem

∂
(ρA)∆x + (ρAV )2 − (ρAV )1 = 0 (25)
∂t
where ∆x = x2 − x1 . Dividing by ∆x and letting ∆x approach zero
∂ ∂
(ρA) + (ρAV ) = 0 (26)
∂t ∂x
Expanding terms
∂ρ ∂A ∂V ∂A ∂ρ
A +ρ + ρA + ρV + AV =0 (27)
∂t ∂t ∂x ∂x ∂x
rearranging in total derivatives and dividing by ρA
4 Water-hammer theory 11

1 dρ 1 dA ∂V
+ + =0 (28)
ρ dt A dt ∂x
The previous Eq. 28 is already the continuity equation, though, the derivatives of ρ
and A must be expressed in terms of the variables of interest, which are V and P .
For this purpose, first we define the bulk modulus of elasticity, K, of a fluid by

dP
K= (29)
dρ/ρ
so the derivative of ρ in Eq. 28 can be written as
dρ ρ dP
= (30)
dt K dt
On the other side, for a circular conduit having radius R,
dA dR
= 2πR (31)
dt dt
In terms of circumferential strain 

dA 1 dR
= 2πR2 (32)
dt R dt
or
1 dA d
=2 (33)
A dt dt
Assuming the pipe material behaviour is linearly elastic and that there are expansion
joints throughout the pipe, the circumferential strain  in terms of circumferential stress
σ is
σ
= (34)
E
And the circumferential stress in a thin-walled conduit for inner pressure P load is

Pr
σ= (35)
e
where e = thickness of the conduit. By taking time derivative
dσ P dr r dP
= + (36)
dt e dt e dt
Based on Eq. 35 the previous expression can be written as
d P r d r dP
E = + (37)
dt e dt e dt
which can be simplified to
r dP
d
= e dtP r (38)
dt E− e
4 Water-hammer theory 12

and from Eq. 33

1 dA 2 r dP
= e dtP r (39)
A dt E− e
Substituting Eqs. 30 and 39 into Eq. 28
!
∂V 1 1 dP
+ + eE P =0 (40)
∂x K 2r − 2
dt
Since normally P/2 << eE/2r the previous equation can be written as
!
∂V 1 1 dP
+ 1 + eE =0 (41)
∂x K 2rK
dt
Finally defining the wave celerity as
v
u
u K
ρ
a=t (42)
1 + 2rK
eE

Substituting Eq. 42 into Eq. 41 and in terms of partial derivatives the continuity
equation takes the following form:
∂P ∂P ∂V
+V + ρa2 =0 (43)
∂t ∂x ∂x

Momentum conservation equation

The extensive property of the fluid for the momentum conservation principle is momen-
tum = mV , hence the intensive property
V ∆m
β = lim =V (44)
∆m→0 ∆m
According to Newton’s second law of motion
dMsys X
= F (45)
dt
Where Msys is the momentum of a system and F the forces exerted on the system.

Fig. 6: Notation for momentum equation.

4 Water-hammer theory 13

Substituting Eqs. 44 and 45 into Reynolds transport theorem Eq. 20

Z
d X
V ρd∀ + [ρA(V − W )V ]2 − [ρA(V − W )V ]1 = F (46)
dt cv
Applying Leibnitz’s rule to the first term on the left-hand side of the previous equation
Z x2
∂ dx2 dx1 X
(ρAV )dx + (ρAV )2 − (ρAV )1 + [ρA(V − W )V ]2 − [ρA(V − W )V ]1 = F
x1 ∂t dt dt
(47)
Taking into account that dx1 /dt = W1 and dx2 /dt = W2
Z x2
∂ X
(ρAV )dx + (ρAV 2 )2 − (ρAV 2 )1 = F (48)
x1 ∂t

Applying mean-value theorem and dividing by ∆x

(ρAV 2 )2 − (ρAV 2 )1
P
∂ F
(ρAV ) + = (49)
∂t ∆x ∆x
Forces acting on the system in the pipe axial direction (Fig. 6)

• Pressure force at section-1: FP1 = P1 A1

• Pressure force at section-2: FP2 = P2 A2

• Pressure force on the converging or diverging sides: FP12 = 21 (P1 + P2 )(A1 − A2 )

• Fluid weight: Fwx = ρgA(x2 − x1 ) sin θ

• Shear force: Fs = τ0 πD(x2 − x1 )

where θ = is the complementary angle between gravity direction and pipe axis, and
τ0 = shear stress between the conduit and the fluid.
The balance of forces considering downstream direction as positive is
X 1
F = P1 A1 − P2 A2 − (P1 + P2 )(A1 − A2 ) − ρgA(x2 − x1 ) sin θ − τ0 πD(x2 − x1 ) =
2
1
= (P1 − P2 )(A1 + A2 ) − ρgA(x2 − x1 ) sin θ − τ0 πD(x2 − x1 )
2
(50)

Dividing the previous equation by ∆x

P
F (P1 − P2 )(A1 + A2 )
= − ρgA sin θ − τ0 πD (51)
∆x 2∆x
Substituting Eq. 51 into Eq. 49 and letting ∆x approach zero.
∂ ∂ ∂P
(ρAV ) + (ρAV 2 ) + A + ρgA sin θ + τ0 πD = 0 (52)
∂t ∂x ∂x
4 Water-hammer theory 14

In the present derivation steady friction head losses are assumed and Darcy-Weisbach
friction formula is used
1
τ0 = ρf V |V | (53)
8
where f = Darcy-Weisbach friction factor.
Substituting Eq. 53 into Eq. 52 and expanding terms
∂ ∂V ∂ ∂V ∂P ρAf V |V |
V (ρA) + ρA +V (ρAV ) + ρAV +A + ρgA sin θ + = 0 (54)
∂t ∂t ∂x ∂x ∂x 2D
Rearranging gives
 
∂ ∂ ∂v ∂V ∂P ρAf V |V |
V (ρA) + (ρAV ) + ρA + ρAV +A + ρgA sin θ + = 0 (55)
∂t ∂x ∂t ∂x ∂x 2D
According to the continuity equation Eq.26 the sum of the terms in the squared brack-
ets is zero. Hence dropping the terms and dividing by ρA, the momentum conservation
equation is reached
∂V ∂V 1 ∂P f V |V |
+V + + g sin θ + =0 (56)
∂t ∂x ρ ∂x 2D

4.2 Simplified equations and numerical approaches

Eqs. 43 and 56 are a set of hyperbolic differential equations valid for unsteady, non-
uniform flow of compressible fluids in elastic conduits. In most engineering applications,
the convective acceleration terms, V (∂P/∂x) and V (∂V /∂x), are very small compared
to the other terms. This approximation allows a solution of the conservation equations
over a regular grid as explained in section 5. Similarly, the slope term is usually small
and may be neglected. Therefore, dropping these terms from the governing equations,
and substituting the dependant variables (P and V ) in terms of piezometric head H and
discharge Q the resultant conservation equations take their traditional form as
∂Q ∂H
+ gA + RQ|Q| = 0 (57)
∂t ∂x
∂Q ∂H
a2 + gA =0 (58)
∂x ∂t
in which R = f /(2DA).
Several numerical methods can be used to solve the previous system of equations:
• method of characteristics
• finite-difference methods
• finite element method
• spectral method
• boundary-integram method
The method of characteristics (MOC) is the most popular and extensively used ap-
proach due to its ease programming, computational efficiency and accuracy on the results.
5 The method of characteristics (MOC) 15

5 The method of characteristics (MOC)

5.1 Characteristic equations

In this subsection the simplified hyperbolic partial differential equations Eqs. 57 and 58
are converted into ordinary differential equations, enabling the solution in their integral
form over a regular grid.
Let’s consider a linear combination between mass and momentum conservation equa-
tions
   
∂Q ∂H ∂Q ∂H
+ gA + RQ|Q| + λ a2 + gA =0 (59)
∂t ∂x ∂x ∂t
or    
∂Q 2 ∂Q ∂H 1 ∂H
+ λa + λgA + + RQ|Q| = 0 (60)
∂t ∂x ∂t λ ∂x
For H and Q in function of the independent variables x and t the total derivative is
dQ ∂Q ∂Q dx
= + (61)
dt ∂t ∂x dt
and
dH ∂H ∂H dx
= + (62)
dt ∂t ∂x dt
By defining the unknown multiplier λ as
1 dx
= = λa2 (63)
λ dt
i.e.,
1
λ=± (64)
a
and by using Eqs. 61 and 62, Eq. 60 can be written as

dQ gA dH
± + RQ|Q| = 0 (65)
dt a dt
Note that the compatibility equations, Eqs. 65, are only valid for dx dt = ±a. Hence
delimiting the solution over straight lines with slope ±1/a one independent variable can
be cancelled out and the partial differential equations are converted to ordinary differential
equations. The lines over the solution is possible are called characteristic lines (see Fig.7)
and physically they represent the path over the xt ˆ plane traversed by a disturbance.
5 The method of characteristics (MOC) 16

ˆ plane.
Fig. 7: Characteristic lines in the xt

Based on Fig.7 Eqs. 65 can be solved by multiplying the left hand side by dt and
integrating Z P
gA P
Z Z P
dQ + dH + R Q|Q|dt = 0 (66)
A a A A
and Z P Z P Z P
gA
dQ − dH + R Q|Q|dt = 0 (67)
B a B B
Although the first two terms in the previous equations are easy to evaluate, an ap-
proximation must be carried out for the friction losses term. Following, a first order
approximation (Eq. 68) and a second order approximation (Eq. 69) are presented.
Z P
R Q|Q|dt ≈ RQA |QA |∆t (68)
A
Z P
R Q|Q|dt ≈ R∆t|QA |QP (69)
A
For the sake of simplicity in the present analysis a first-order approximation is applied
allowing an explicit solution of the system. The first-order approximation usually yields
satisfactory results for most engineering applications.
Finally the characteristic equations are reached
gA
QP − QA + (HP − HA ) + R∆tQA |QA | = 0 (70)
a
gA
QP − QB − (HP − HB ) + R∆tQB |QB | = 0 (71)
a
By combining the known variables together Eqs. 70 and 71 the positive and negative
characteristic equations are reached:

QP = Cp − Ca HP (72)
QP = Cn + Ca HP (73)
5 The method of characteristics (MOC) 17

being

Cp = QA + Ca HA − R∆tQA |QA | (74)

Cn = QB − Ca HB − R∆tQB |QB | (75)

and
gA
Ca = (76)
a
Solving explicitly the system of equations (Eqs. 72 and 73) HP and QP can be de-
termined in each time step. Though, notice that the characteristic equations are only
valid along the characteristic lines AP and BP respectively. The constants Cp and Cn
are known values, they represent the information transferred from time step to time step
through the positive and negative characteristics lines. And Ca depends upon the conduit
properties.

5.2 Boundary conditions

The characteristic equations Eqs. 72 and 73 are used to solve the dependant variables of
the system H and Q at the inner nodes (grid points) of the pipe. However, at the boundary
nodes only one of the equations is valid, as only one characteristic line reaches the boundary
grid points, the positive characteristic line in case of the downstream boundary and the
negative for the upstream boundary. Therefore extra information is required in order to
determine the dependant variables at the boundary nodes. Boundary conditions provides
such information. Some basic boundary conditions are developed in the present section.

Constant-level upstream
In many situations the changes in the reservoir level are negligible. If the entrance losses
as well as the velocity head are negligible, then

HP1,j = Hres (77)

in which Hres = height of the reservoir water surface. Eq. 73 for the upstream end
thus becomes
QP1,j = Cnj + Caj Hres (78)

Dead end at downstream end

At the dead end discharge is zero and from the positive characteristic equation the following
boundary conditions are set

QPn,j = 0 (79)
Cp j
HPn,j = (80)
Caj
5 The method of characteristics (MOC) 18

Valve at downstream end

The steady-state flow through a valve can be described as
q
Q0n,j = (Cd Av )0 2gH0n,j (81)

where the subscript 0 indicates steady-state conditions, Cd is the coefficient of discharge,

and Av the area of the valve opening.
Assuming that the transient-state flow through a valve may be described by an equiv-
alent expression p
Qn,j = (Cd Av ) 2gHn,j (82)
Dividing Eq. 82 by Eq. 81, taking the square of both sides and defining the relative
opening τ = (Cd Av )/(Cd Av )0

(Q0n,j τ )2
Q2Pn,j = HPn,j (83)
H0n,j

By means of the positive characteristic equation (Eq. 72) the previous equation becomes

Q2Pn,j + Cv QPn,j − Cpi Cv = 0 (84)

in which Cv = (τ Q0n,j )2 /(Ca H0n,j ). Solving for QPn,j and neglecting the negative sign in
the radical term q
QPn,j = 0.5(−Cv + Cv2 + 4Cpi Cv ) (85)

And finally HPn,j is obtained applying the positive characteristic equation (Eq. 72). Notice
that τ is the parameter that defines the valve manoeuvre. τ = 1 corresponds to the initial
state valve opening and τ = 0 corresponds to a closed valve. Values of τ during the
manoeuvre transition may be specified by an algebraic expression or in a tabular form.

Branching junctions
Branching junctions are solved by applying mass and energy conservation principles at
the branching node. Considering Fig. 8, mass conservation equation can be expressed as

QPi,n+1 = QPi+1,1 + QPi+2,1 (86)

5 The method of characteristics (MOC) 19

Fig. 8: Schematic of branching junction.

While the respective characteristic equations from each branch are:

QPi,n+1 = Cpi − Cai HPi,n+1 (87)

QPi+1,1 = Cni+1 − Cai+1 HPi+1,n (88)

QPi+2,1 = Cni+2 − Cai+2 HPi+2,n (89)

If head losses and different velocity heads are neglected at the branching junction, from
energy conservation it follows that

HPi,n+1 = HPi+1,1 = HPi+2,1 (90)

Hence, combining Eqs. 8 to 90 the solution for hydraulic head is:

Cpi − Cni+1 − Cni+2

HPi,n+1 = (91)
Cai + Cai+1 + Cai+2
Finally HPi+1,1 and HPi+2,1 are obtained by substituting HPi,n+1 value at Eq. 90 and
then resepctctive discharges by applying Eqs. 87, 88 and 89.

Other boundary conditions

Hydraulic machinery (e.g. pumps, turbines) or surge protection devices (e.g. surge tanks,
hydropneumatic vessels, relief valves, etc) are usually modelled by breaking apart the pipe
and considering the effect of such devices as boundary conditions. The way to model these
mechanisms is explained inStreeter & Wylie (1967) and Chaudhry (2014).
6 Surge protection measures 20

6 Surge protection measures

6.1 Surge tank

Aiming at dampening sudden variations of pressures, surge tanks are water storage struc-
tures, open or closed, connected to the piping system. Also named surge shafts or surge
chambers, surge tanks reduce the amplitude of pressure fluctuations by reflecting the in-
coming pressure waves or by storing or providing water. Figure 9 depicts a surge tank
used for protecting a water transport system.

Fig. 9: Picture of a surge tank protecting a water transport system, Tortosa (Spain).

Depending on their design and site constraints they may take several forms:

• Simple tank: open shaft or standpipe connected to the pipeline with negligible losses
at its junction with the pipeline (v.i. Fig. 10-a).

• Orifice tank: simple tank with an orifice restricting the entrance to the surge tank
(v.i. Fig. 10-b).

• Differential tank: two composed surge tanks, a simple surge tank and an orifice surge
tank (v.i. Fig. 10-c).

• One-way tank: the water is allowed to flow only from the tank to the main pipe with
the goal to protect the system for pressure drops (v.i. Fig. 10-d).

• Closed tank: with its top closed and partially filled by compressed air (v.i. Fig. 10-e).

• Tank with galleries: with upper or lower galleries, providing larger cross-sectional
area and storage capacity (v.i. Fig. 10-f).
6 Surge protection measures 21

Fig. 10: Schematics of surge tank types.

The period of mass oscillation in a reservoir-pipe-tank system can be computed, ne-

glecting friction, following Eq. 92.
s
LAs
T = 2π (92)
gAp

Where L i the pipe length, Ap the cross-sectional area of the pipe and As the cross-sectional
area of the surge tank.
Generally, the time-scale of a water-hammer event is much shorter than the time-scale
of the mass oscillation produced by surge tanks, hence the two problems use to be treated
separately. In Streeter & Wylie (1967) and Chaudhry (2014) methods for modelling and
sizing surge tanks are described.

6.2 Hydropneumatic vessel

The purpose of hydropneumatic vessels is to protect pipe systems from high and low
pressures and prevent them from column separation during hydraulic transients. They
are composed of a closed chamber filled partially with compressed air and water. The
connection between the chamber and the pipeline is designed with an orifice that restricts
the flow, normally giving preference to the outflow (from the chamber to the pipeline) by
means of a by-pass or a differential orifice. To avoid air intrusion into the pipeline some
hydropneumatic vessels are deassigned with a bladder so the two fluids, air and water,
remain separate. Figure 11 depicts a typical set-up of a hydropneumatic vessel protecting
a pumping system.
6 Surge protection measures 22

Fig. 11: Schematic of hydropneumatic vessel.

6.3 Valves
Hydraulic transients can be controlled by means of auxiliary valving and flow control
valves. The aim is mainly to allow a rapid outflow for a certain upsurge pressure or an air
inflow for a downsurge pressure. Several types of valves may be used for transient control:

• Safety valves (Fig. 13): they open when the pressure inside the pipe exceeds the
pressure limit by a spring or weight-loaded mechanism. They close rapidly when the
pressure drops bellow the limit.

Fig. 12: Schematic of pressure-safety valve.

• Pressure-relief valves (Fig. 13): also loaded by a spring or by weights, pressure-relief

valves are used for controlling excess pressures by opening the valve when a certain
pressure is exceeded, the closure may be damped allowing longer closure times.
6 Surge protection measures 23

Fig. 13: Schematic of pressure-relief valve.

• Pressure-regulating valves (Fig. 14): PRV’s are pilot-controlled throttling valves,

which are opened or closed by a servomotor with the opening and closing times set
individually.

Fig. 14: Schematic of pressure-regulating valve.

• Air-inlet valves (Fig. 15): they let air flow in the pipe avoiding cavitation and the
high pressures derived from column separation.

Fig. 15: Schematic of air-inlet valve.

6 Surge protection measures 24

• Check valves (Fig. 16): used to prevent reverse flow through a pump and to prevent
inflow into a one-way surge tank from the pipeline.

Fig. 16: Schematic of check valve.

6.4 By-passes
By-pass valving is a low-cost efficient technique for avoiding pressure surges (up and down)
in pumping stations. Several configurations are possible. Essentially the aim is to give
continuity to the water flow in case of pump failure. Figure 17 depicts a typical by-pass
set-up in a pumping station.

Fig. 17: Schematic of a by-pass in a pumping station.

6.5 Unconventional devices

6.5.1 Flywheel
The aim of the flywheel is to avoid or reduce the transient flow caused by hydraulic
machinery by means of increasing the inertia of their runners. A disks or wheel of a
specific mass and shape is added in the runner shaft, so when there is a pump trip or a
sudden turbine load variation, their increased inertia slows down the transient excitation.
The size of the flywheel depends of the pipe set-up. For long pipes heavy flywheels may be
required incrementing excessively the initial power of the hydraulic machine to start-up.
Figure 18 depicts a flywheel assembled on a Francis turbine.
6 Surge protection measures 25

Fig. 18: Flywheel on a Francis turbine.

6.5.2 Breakable fittings

Also called burst or rupture disks, breakable fittings are an alternative to safety valves.
They consists of an opening in the pipe covered by a diaphragm that breaks and relieves
pressure beyond a certain limit. Afterwards the fitting must be replaced. Figure 19 depicts
a burst disk before and after rupture.

Fig. 19: Example of burst disk before and after rupture.

6.5.3 Intermediate flexible pipes

Hyper-flexible pipe sections can be used for adding damping to water-hammer waves. The
increased hydraulic section due to a highly deformable pipe-wall and the effect of pipe-wall
viscoelasticity contribute to reduce (but not to suppress) transient pressures. Figure 19
depicts a pumping station where intermediate flexible pipes are used to mitigate the effect
of hydraulic transients.
6 Surge protection measures 26

Fig. 20: Example of a pumping station protected by intermediate flexible pipes.

7 Exercises 27

7 Exercises

Exercise-1: steady-state equations

Derive the steady-state equations from the water-hammer equations.

Exercise-2: wave speed and Joukowsky pressure rise

Wave speed: Consider a pipe with frequent expansion joints, inner diameter D = 0.02 m,
pipe-wall thickness e = 0.001 and a bulk modulus of the water K = 2.2 GP a. Plot the
wave celerity in function of Young’s modulus of elasticity E using Halliwell’s formula.
Mark in the plot the celerity corresponding to the following pipe-wall materials: High-
Density Polyethylene (E = 0.8 GP a), Polyvinyl chloride (E = 2.9GP a), Glass reinforced
polyester matrix (E = 17 GP a), copper (E = 117 GP a), steel (E = 200 GP a). Compute
the celerity and give an explanation of the obtained value when the Young’s modulus of
elasticity E → ∞

Joukowsky pressure rise: Joukowsky equation gives the piezometric head rise resulting
from an instantaneous closure of the valve in a frictionless system for an initial steady flow
aV0
∆H = (93)
g
Notice that using MOC positive characteristic equation and forcing QP = 0, the previous
expression is easily obtained. Derive it and compute Joukowsky pressure according to
the celerities obtained in the previous exercise and considering an initial flow velocity
V0 = 0.3 m/s.

Exercise-3: reservoir-pipe valve system

Consider a reservoir-pipe-valve system like the one in Fig. 21, composed by a thin-walled
conduit of length of L = 100 m and with frequent expansion joints, inner diameter D =
0.02 m, pipe-wall thickness e = 0.001 and a bulk modulus of the water K = 2.2 GP a.

Fig. 21: Reservoir-pipe-valve system

Assuming copper is the material of the reservoir-pipe-valve system, use the excel
spreadsheet from the lecture material and run three different simulations for a linear
closure of the valve considering the following closing times: Tf = 0, Tf = 2L/a and
Tf = 4L/a. The initial flow conditions before closing the valve are: piezometric head
in the reservoir Hres = 40 m, piezometric head in the valve Hv = 39 m and initial flow
7 Exercises 28

velocity V0 = 0.3 m/s. Simulation parameters: time-step dt = 5 ms, simulation period

T = 5 s and the valve starts closing at the time 1s. Determine the piezometric head
variation next to the valve (downstream section) for the three simulations.

Exercise-4: water-hammer code

Based on the VBA code from the Excel spreadsheet provided, implement a second order
scheme for head losses computation.

Exercise-5: surge protection devices

Consider the following reservoir-pipe-valve-pipe-reservoir system (v.i. Fig. 22) and carry
out the following simulations based on the input data presented in Tables 1 and 2. Assume
the pipe-wall is made of steel, its absolute roughness is 0.002 mm and the valve manoeuvre
is linear and takes 2 s.

1. Compare the performance upstream the valve (N8) with: no protection devices, a
surge tank (D = 0.5m) and a hydropneumatic vessel (V = 0.785m3 ).

2. Do the same at the section downstream the valve (N5).

Discuss the different results obtained, what are the advantages of hydropneumatic vessels
vs. surge tanks?

Fig. 22: Schematic of reservoir-pipe-valve-pipe-reservoir system

Tab. 1: Input table for pipes

Element Ni Zi (m) Nf Zf (m) Dint (mm) L(m) e(mm)
T1 N1 0 N8 0 100 100 10
T2 N8 0 N3 0 100 100 10
T3 N4 0 N5 0 100 100 10
T5 N4 0 N6 0 100 100 10

Tab. 2: Input table for reservoirs

Element Zb (m) ZSW (m)
D1 0 20
D2 0 18
7 Exercises 29

Exercise-6: branching junction

When a numerical model is implemented, an important stage is testing the model by
means of benchmark problems that will show if the model performs as expected from
theory. The following problem is useful to test branching junctions. The piping system
to be modelled consists of two reservoirs located upstream and connected by two conduits
that merge in a Y junction which is followed by another conduit with a valve in the other
pipe-end (v.i. Fig. 23). Downstream the valve there is a fourth conduit that leads to a
lower reservoir. The upstream reservoirs have a constant water level at 10 m, while the
downstream reservoir at 0 m. Pipe-4 has a length L = 100 m, a diameter D = 0.1 m,
pipe-wall thickness e = 0.001 m. Assume the pipe-wall is made of steel and its absolute
roughness is 0.002 mm.
Tasks:

1. Simulation-1: design Subsystem-1 in order to make it hydraulically equivalent to

Subsystem-2. Build up the numerical model and test if they are equivalent, for both
steady and transient conditions. For the pipe design consider: pipe-cross section
(D), Darcy-Weisbach friction coefficient (f), and pipe-wall thickness (e).

2. Simulation-2: change the diameter of pipe-1 to D = 0.1m.

3. Simulation-3: change the pipe-wall thickness of pipe-1 to e = 0.001m.

4. Simulation-4: change the water surface elevation of one of the upstream reservoirs
to Hres = 3m.

Compare and discuss the pressure output from the four tests at the downstream section
of pipe-3.

Fig. 23: Schematic of the pipe system with branching junction

7 Exercises 30

Exercise-7: pumping scheme

Protect against undesired hydraulic transients the pumping scheme depicted below (v.i.
Fig. 24). Input parameters are presented in Tables 3, 4, 5 and 6.
1. Assess first the required valve manoeuvre to avoid cavitation.
2. Propose a solution to enable 10 s valve closures with no cavitation.

Fig. 24: Schematic of the pumping system

Tab. 3: Input table for pipes

Element Ni Zi (m) Nf Zf (m) Dint (mm) L(m) e(mm)
T1 N1 0 N4 0 1000 100 10
T3 N6 100 N7 100 500 10000 10
T4 N9 0 N6 100 500 500 10
T5 N5 0 N8 0 500 10 10

Tab. 4: Input table for reservoirs

Element Zb (m) ZSW (m)
D1 0 10
D2 100 110

Tab. 5: Input table for pumps

Element Ni Nf Q(l/s)
B1 N4 N5 1000

Tab. 6: Input table for reservoirs

Element Ni Nf DN(mm)
Rg1 N8 N9 500
7 Exercises 31

Exercise-8: micro-hydropower scheme

The aim of this exercise is to assess an accident that occurred in a real micro-hydropower
scheme (Göksenli & Eryürek, 2015). The scheme is composed of two in-line waterfalls
(v.i. Fig. 25). The downstream one is where the accident occurred and is the one assessed
hereby. The system is not protected to pressure surges. With an initial flow rate of 11
m3 /s, a fast closure generated a water-hammer event, that combined with a deficient
welding of the flanges caused circumferential cracks on the penstock joints. Based on the
in-put data provided in Table 7 and Fig. 26, design the best surge protection to avoid
pressures above 30 bars.
→ Hint: simplify the scheme to a reservoir-pipe-valve system.

Fig. 25: Schematic of the micro-hydropower system

Tab. 7: Input table for pipes

Element Zi (mm) Zf (m) Dint (mm) L(m) e(mm) a(m/s)
Tunnel 500 450 2000 6500 84 1000
Penstock 450 300 2000 400 – 1200
7 Exercises 32

Fig. 26: Valve manoeuvre

Exercise-9: large-hydropower scheme

The schematic shown in Fig.27 is based on a real hydropower scheme (Purcell, 2015)
which experienced an incident caused by a transient event due to a fast manoeuvre of the
upstream gate. Build up the hydraulic model according to the data provided in Tables 8,
9, 10 and 11, and carry out the following simulations:

1. Downstream water-hammer: simulate an instantaneous closure of the down-

stream valve. Is the system well designed to resist such transient loads?

2. Incident: simulate a valve trip of the upstream gate for a closing time of 20 s
and the downstream closing after 244 s. This is actually the incident described in
(Purcell, 2015). Why did the hydropower scheme failed?

3. Solution: design a solution to avoid the problems observed at the previous simula-
tion.
7 Exercises 33

Fig. 27: Schematic of the hydropower system

Tab. 8: Input table for pipes

Element Ni Zi (m) Nf Zf (m) Dint (mm) L(m) e(mm)
T6 N4 167.3 N5 0 2500 200 75
T7 N6 0 N9 0 2500 20 75
T8 N10 0 N2 0 3500 200 100
T9 N13 167.3 N17 167.3 4880 20 150
T10 N1 167.3 N12 167.3 4880 20 150
T13 N17 167.3 N20 167.3 4880 200 150
T14 N20 167.3 N21 167.3 4880 200 150
T15 N21 167.3 N4 167.3 2500 20 75

Tab. 9: Input table for reservoirs

Element Zb (m) ZSW (m)
D1 167.3 186.3
D2 0 5

Tab. 10: Input table for turbines

Element Ni Nf Q(l/s) H(m)
Tb1 N9 N10 39000 130

Tab. 11: Input table for reservoirs

Element Ni Nf DN(mm)
Rg1 N5 N6 1500
Rg2 N12 N13 4880
7 Exercises 34

References

Chaudhry, M. H. (2014). Applied Hydraulic Transients, ISBN: 978-1-4614-8537-7. New

York, NY: Springer.

Göksenli, A. & Eryürek, B. (2015). Failure analysis of pipe system at a hydroelectric

power plant. World Academy of Science, Engineering and Technology, International
Journal of Mechanical, Aerospace, Industrial, Mechatronic and Manufacturing Engi-
neering 9(9), 1643–1646.

Parmakian, J. (1963). Waterhammer analysis. Prentice-Hall, Inc., Englewood Cliffs,

NJ, (Dover Reprint).

Purcell, P. J. (2015). Modelling a transient event at an hydroelectric scheme. In:

International Conference on Pressure Surges (Dublin, Ireland).

Streeter, V. L. & Wylie, E. B. (1967). Hydraulic transients, vol. 967. McGraw-Hill

New York.