TOPIC T2: FLOW IN PIPES AND CHANNELS

AUTUMN 2013

Objectives
(1)
(2)
(3)
(4)

Calculate the friction factor for a pipe using the Colebrook-White equation.
Undertake head loss, discharge and sizing calculations for single pipelines.
Use head-loss vs discharge relationships to calculate flow in pipe networks.
Relate normal depth to discharge for uniform flow in open channels.

1. Pipe flow
1.1 Introduction
1.2 Governing equations for circular pipes
1.3 Laminar pipe flow
1.4 Turbulent pipe flow
1.5 Expressions for the Darcy friction factor,
1.6 Other losses
1.7 Pipeline calculations
1.8 Energy and hydraulic grade lines
1.9 Simple pipe networks
1.10 Complex pipe networks (optional)
2. Open-channel flow
2.1 Normal flow
2.2 Hydraulic radius and the drag law
2.3 Friction laws Chzy and Mannings formulae
2.4 Open-channel flow calculations
2.5 Conveyance
2.6 Optimal shape of cross-section
Appendix

References
Chadwick and Morfett (2013) Chapters 4, 5
Hamill (2011) Chapters 6, 8
White (2011) Chapters 6, 10 (note: uses f = 4cf = for friction factor)
Massey (2011) Chapters 6, 7 (note: uses f = cf = /4 for friction factor)

Hydraulics 2

T2-1

David Apsley

1. PIPE FLOW
1.1 Introduction
The flow of water, oil, air and gas in pipes is of great importance to engineers. In particular,
the design of distribution systems depends on the relationship between discharge (Q),
diameter (D) and available head (h).
Flow Regimes: Laminar or Turbulent

laminar
In 1883, Osborne Reynolds demonstrated the
occurrence of two regimes of flow laminar or
turbulent according to the size of a dimensionless
parameter later named the Reynolds number. The
conventional definition for round pipes is
VD
(1)
Re

where
V = average velocity (= Q/A)
D = diameter
= kinematic viscosity (= /)

turbulent

For smooth-walled pipes the critical Reynolds number at which transition between laminar
and turbulent regimes occurs is usually taken as
Recrit 2300
(2)
In practice, transition from intermittent to fully-turbulent flow occurs over 2000 < Re < 4000.

Development Length
At inflow, the velocity profile is often nearly
uniform. A boundary layer develops on the
pipe wall because of friction. This grows
with distance until it fills the cross-section. Beyond this distance the velocity profile becomes
fully-developed (i.e., doesnt change any further with downstream distance). Typical
correlations for this development length are (from White, 2011):
Ldev 0.06 Re (laminar )

(3)
1/6
(turbulent)
D
4.4 Re
The kinematic viscosity of air and water is such that most pipe flows in civil engineering
have high Reynolds numbers, are fully turbulent, and have a negligible development length.
Example.
water = 1.0106 m2 s1. Calculate the Reynolds numbers for average velocity 0.5 m s1 in
pipes of inside diameter 12 mm and 0.3 m. Estimate the development length in each case.
Answer: Re = 6000 and 1.5105; Ldevelop = 0.23 m and 9.6 m.

Hydraulics 2

T2-2

David Apsley

1.2 Governing Equations For Circular Pipes

Fully-developed pipe flow is determined by a balance between three forces:

pressure;

weight (component along the pipe axis);

friction.
For a circular pipe of radius R, consider the forces with components along the pipe axis for an
internal cylindrical fluid element of radius r < R and length l.

l
p

direct
ion of
flow

z
mg

p+p

Note:
(1)
p is the average pressure over a cross-section; for circular pipes this is equal to the
centreline pressure, with equal and opposite hydrostatic variations above and below.
(2)
The arrow drawn for stress indicates its conventional positive direction,
corresponding to the stress exerted by the outer on the inner fluid. In this instance the
inner fluid moves faster so that, if V is positive, will actually be negative.
Balancing forces along the pipe axis:
p( r 2 ) ( p p)(r 2 ) mg
(2r l ) 0
sin

weight

net pressure force

friction

From the geometry,

m r 2 l ,

sin

z
l

Hence:

p(r 2 ) r 2 g z (2r l ) 0
Dividing by the volume, r2 l,
( p gz )

2 0
l
r
Writing p* = p + gz for the piezometric pressure and rearranging for the shear stress,

1 p *
r
2 l

(4)

Since the flow is fully-developed the shear stress and the gradient of the piezometric pressure
are independent of distance. For convenience write G for the streamwise pressure gradient:
p *
dp*
G

(constant )
(5)
l
dl
(The negative sign is included because we expect p* to drop along the pipe.) Hence, from (4),

Hydraulics 2

T2-3

David Apsley

 12 Gr

(6)

where

gh f
dp *
pressure drop

dl
length
L
G is the piezometric pressure gradient and hf is the head lost (by friction) over length L.
G

(7)

(6) applies to any fully-developed pipe flow, irrespective of whether it is laminar or turbulent.
For laminar flow it can be used to establish the velocity profile, because can be related to
the velocity gradient du/dr (Section 1.3). For turbulent flow an analytical velocity profile is
not available, but gross parameters such as quantity of flow and head loss may be obtained if
the wall shear stress w can be related empirically to the dynamic pressure V2 (Section 1.4).
1.3 Laminar Pipe Flow
Laminar flow through a circular pipe is called Poiseuille1 flow or Hagen2-Poiseuille flow.
In laminar flow the shear stress is related to the velocity gradient:
du

dr
Hence, from (6) and (8),
du
1G

r
dr
2

R
(8)

Integrating and applying the no-slip condition at the wall (u = 0 on r = R),

Laminar pipe-flow velocity profile
G 2
u
(R r 2 )
4

(9)

Example. Find, from the velocity distribution given above,

(a) the centreline velocity u0 ;
(b) the average velocity V;
(c) the volumetric flow rate Q, in terms of head loss and pipe diameter;
L V2
(d) the friction factor , defined by h f ( ) , as a function of Reynolds number, Re.
D 2g
4
GR 2
GR 2
64
gh f D
1
Answer: (a) u 0 u (0)
; (b) V 2 u 0
; (c) Q
; (d)
Re
4
8
128 L

Part (d) of this exercise demonstrates that the friction factor is not constant.
1

J.L.M Poiseuille (1799-1869); French physician who was interested in flow in blood vessels.
G.L.H. Hagen; German engineer who, in 1839, measured water flow in long brass pipes and reported that there
appeared to be two regimes of flow.
2

Hydraulics 2

T2-4

David Apsley

1.4 Turbulent Pipe Flow

In turbulent flow one is usually interested in time-averaged quantities. Velocity usually
implies time-averaged velocity and the shear stress is the time-averaged rate of transport of
momentum per unit area; it is dominated by turbulent mixing rather than viscous stresses.
In turbulent flow there is no longer an explicit relationship between mean stress and mean
velocity gradient du/dr because a far greater transfer of momentum arises from the net effect
of turbulent eddies than the relatively small viscous forces. Hence, to relate quantity of flow
to head loss we require an empirical relation connecting the wall shear stress and the average
velocity in the pipe. As a first step define a skin friction coefficient cf by

wall shear stress

cf
1 w2
(10)
dynamic pressure

V
2
(Later, cf will be absorbed into a friction factor to simplify the expression for head loss.)
For the length of pipe shown, the balance of forces along the axis in fully-developed flow is:
D 2
p
mg
w DL 0
sin

weight
net pressure force

wall friction

From the geometry,

D 2
m
L
4
z
sin
L

L
w
p

Substituting these gives:

D 2
D 2
p
gz
w DL
4
4
D 2

( p gz )
w DL
4

direct
ion of
flow

p+p

mg

Dividing by the cross-sectional area (D2/4),

L
p* 4 w
D
Write:

w c f ( 12 V 2 )

(definition of skin-friction coefficient)

Substituting and rearranging gives for the drop in piezometric pressure:

L
p * 4c f ( 12 V 2 )
D
The quantity 4cf is known as the Darcy friction factor and is denoted by .

Hydraulics 2

T2-5

David Apsley

Darcy3-Weisbach4 Equation
p *

L 1
( 2 V 2 )
D

pressure loss due to friction

(11)

L
dynamic pressure
D

Dividing by g this can equally well be written in terms of head rather than pressure:
L V2
hf ( )
D 2g
L
head loss due to friction dynamic head
D

(12)

*** Very important ***

There is considerable disagreement about what is meant by friction factor and what symbol
should be used to denote it. What is represented here by is also denoted f by some authors
and 4f by others (including ourselves in past years and exam papers). Be very wary of the
definition. You can usually distinguish it by the expression for friction factor in laminar flow:
64/Re with the notation here; 16/Re with the next-most-common alternative.
It remains to specify for a turbulent pipe flow. Methods for doing so are discussed in
Section 1.5 and lead to the Colebrook-White equation. Since depends on both the relative
roughness of the pipe (ks/D) and the flow velocity itself (through the Reynolds number
Re VD/) either an iterative solution or a chart-based solution is usually required.
Although the bulk velocity V appears in the head-loss equation the more important quantity is
the quantity of flow, Q. These two variables are related, for circular pipes, by
D 2
Q VA V
4
where D is the pipe diameter.
At high Reynolds numbers tends to a constant (determined by surface roughness) for any
particular pipe. In this regime compare:
Q2
h f 5 (turbulent)
D
Q
h f 4 (laminar)
D

3
4

Henri Darcy (1803-1858); French engineer; conducted experiments on pipe flow.

Julius Weisbach; German professor who, in 1850, published the first modern textbook on hydrodynamics.

Hydraulics 2

T2-6

David Apsley

1.5 Expressions for the Darcy Friction Factor

Laminar Flow (theory)
64

Re

Turbulent Flow (smooth or rough pipes)

Nikuradse5 (1933) used sand grains to roughen pipe surfaces. He defined a relative roughness
ks/D, where ks is the sand-grain size and D the diameter of the pipe. His experimental curves
for friction factor (see, e.g., Whites textbook) showed 5 regions:
1. laminar flow (Re < Recrit 2000; roughness irrelevant)
2. laminar-to-turbulent transition (approximately 2000 < Re < 4000)
3. smooth-wall turbulent flow ( is a function of Reynolds number only)
4. fully-rough-wall turbulent flow ( a function of relative roughness only)
5. intermediate roughness ( is a function of both Re and ks/D)
In the smooth- or rough-wall limits, Prandtl6 and Von Krmn7 gave, respectively:
1
Re

Smooth-wall turbulence:
2.0 log 10
2.51

1
3.7 D

Rough-wall turbulence:
2.0 log 10
ks

Unfortunately, in practice, many commercial pipes lie in the region where both roughness
and Reynolds number are important, so that the friction factor is not constant for any
particular pipe but depends on the flow rate. Colebrook and White (1937) combined smoothand rough-wall turbulence laws into a single formula, the Colebrook-White equation.
Colebrook-White Equation
2.51
k
2.0 log 10 s

3.7 D Re

(13)

This is the main formula for the friction factor in turbulent flow. The main difficulty is that it
is implicit ( appears on both sides of the equation) and so must be solved iteratively. There
are several explicit approximations to (13), accurate to within a few percent for realistic
ranges of Reynolds number see the references in Massey and Whites textbooks.

Equivalent Sand Roughness

For commercial pipes the pattern of surface roughness may be very different to that in the
artificially-roughened surfaces of Nikuradse. Colebrook (1939) and Moody (1944) gathered
5

Johann Nikuradse (1894-1979); PhD student of Prandtl.

Ludwig Prandtl (1875-1953); German engineer; introduced boundary-layer theory.
7
Theodore von Krmn (1881-1963); Hungarian mathematician and aeronautical engineer; gave his name to the
double row of vortices shed from a 2-d bluff body and now known as a Krmn vortex street.
6

Hydraulics 2

T2-7

David Apsley

data to establish effective roughness for typical pipe materials. Typical values of ks are given
in the Appendix.

Moody Chart
Graphical solutions of (13) exist. The most well known is the Moody chart ( versus Re for
various values of relative roughness ks/D). The curves are just solutions of the ColebrookWhite equation. A home-produced version is shown below.

0.10
0.09
0.08

ks/D
0.05

0.06
0.05

0.04

Laminar
l = 64/Re

Transition

0.07

0.03
0.02
0.01
0.005

0.03

0.002

0.001

0.02

0.0004

smooth-walled limit

0.00001

0.01
1.0E02

1.0E03

1.0E04

1.0E05

1.0E06

0.0002
0.0001
0.00005

1.0E07

Re = VD/n

Hydraulics 2

T2-8

David Apsley

1.6 Other Losses

Pipeline systems are subject to two sorts of losses:

wall-friction, contributing a continuous fall in head over a large distance;

minor losses due to abrupt changes in geometry; e.g. pipe junctions, valves, etc.
Each type of loss can be quantified using a loss coefficient K, the ratio of pressure loss to
dynamic pressure (or head loss to dynamic head):
V2
or
(14)
pressure loss K ( 12 V 2 )
head loss K ( )
2g
Typical values of K are given below.
Commercial pipe fittings (approximate)
Fitting
K
Globe valve
10
Gate valve wide open
0.2
Gate valve open
5.6
0.9
90 elbow
Side outlet of T-junction
1.8

Entry/exit losses
Configuration
Bell-mouthed entry
Abrupt entry
Protruding entry
Bell-mouthed exit
Abrupt enlargement

K
0
0.5
1.0
0.2
1.0

Minor losses are a one-off loss, occurring at a single point. Frictional losses are
proportional to the length of pipe L and, in the grand scheme of things, usually dominate. For
long pipelines minor losses are often ignored.

Hydraulics 2

T2-9

David Apsley

1.7 Pipeline Calculations

The objective is to establish the relationship between available head and quantity of flow.
Available head, H1 H2 = sum of head losses along the pipe

(15)

Available head is the overall drop in head from start to end of the pipe, usually determined
by still-water levels, sometimes supplemented by additional pumping head. Head losses are
proportional to the dynamic head V 2/2g. Fluid then flows through the pipe at precisely the
right velocity V (or discharge Q) that (15) is satisfied.
Pipe parameters are illustrated below. Although a reservoir is indicated at each end of the
pipe, this is simply a diagrammatic way of saying a point at which the total head is known.

D
Q

Typical pipeline problems are: given two of the following parameters, find the third.
Head loss:
h
Quantity of flow:
Q
Diameter:
D
Other parameters: length L, roughness ks, kinematic viscosity and minor loss coefficient K.
Calculations involve:
(1) Head losses
e.g. with friction factor and minor-loss coefficient K:
L
V2
h ( K )( )
D
2g
(2) Expressions for loss coefficients
e.g. the Colebrook-White equation for friction factor:
1
2.51
k
2.0 log 10 s

3.7 D Re

(16)

(17)

In most problems (16) and/or (17) must be solved iteratively. The exception is the calculation
of Q when h and D are known (and minor losses neglected) because in this special case (Type
1 in the examples which follow) the Reynolds number can be expanded to give:
k
1
2.51

(18)
2.0 log 10 s
2
3
.
7
D

D V

If minor-loss coefficient K = 0 then the combination V2 can be found from (16) and hence
can be found. Knowledge of both V2 and gives V and thence Q.

Hydraulics 2

T2-10

David Apsley

Overall Head Loss

If there is a free surface in still water then both gauge pressure and velocity there are zero and
so the total head equals the surface elevation: H = z. If, however, the pipe discharges to
atmosphere as a free jet then the total head includes the dynamic head, V2/2g.
If the discharge is to another reservoir (a) then
(with a well-rounded exit):
H1 z1 , H 2 z 2
and the loss in head is just the difference in
still-water levels:
h z1 z 2

z1

Alternatively, if the discharge is a free jet to

atmosphere, then
H1 z1 , H 2 z 2 V22 / 2 g
and the loss in head is
h z1 z 2 V22 / 2 g
(In terms of piezometric head this is equivalent
to having an exit loss coefficient 1.0.)

z1

z2

z2

The second case also applies if there is an abrupt exit into a tank, since flow separation
means that the pressure in the jet leaving the pipe is essentially the hydrostatic pressure in the
tank (piezometric head z2) but there is still a dynamic head V22 / 2 g whose energy is ultimately
dissipated in the receiving tank. Again, this is equivalent to an exit loss coefficient 1.0. For
long pipelines, however, this is usually neglected.

Hydraulics 2

T2-11

David Apsley

V2

Type 1: Diameter D and head difference h known; find the quantity of flow Q.
Example. A pipeline 10 km long, 300 mm diameter and with roughness 0.03 mm, conveys
water from a reservoir (top water level 850 m AOD) to a water treatment plant (700 m AOD).
Assuming that the reservoir remains full, and neglecting minor losses, estimate the quantity
of flow. Take = 1.0106 m2 s1.

Solution.
List known parameters:
L = 10000 m
D = 0.3 m
h = 150 m
ks = 3105 m
= 1.0106 m2 s1
Since D and h are known, the head-loss equation enables us to find V2:
L V2
2 gDh 2 9.81 0.3 150
h ( )

V 2

0.08829 m 2 s 2
D 2g
L
10000
Rewriting the Colebrook-White equation,
1
2.51
k
2.0 log 10 s

3.7 D Re
k
2.51

2.0 log 10 s
2
3
.
7
D
D V

5
3 10
2.51 1.0 10 6

2.0 log 10

0.3 0.08829
3.7 0.3
8.516
Hence,
1

0.01379
8.516 2

Knowledge of both V2 and gives

V

V 2

0.08829
0.01379

2.530 m s 1

Finally, the quantity of flow may be computed as velocity area:

D 2
0.32
Q VA V (
) 2.530
0.1788 m 3 s 1
4
4
Answer: quantity of flow = 0.18 m3 s1.

Hydraulics 2

T2-12

David Apsley

Type 2: Diameter D and quantity of flow Q known; find the head difference h.
Example. The outflow from a pipeline is 30 L s1. The pipe diameter is 150 mm, length
500 m and roughness estimated at 0.06 mm. Find the head loss along the pipe.
Solution.
List known parameters:
Q = 0.03 m3 s1
L = 500 m
D = 0.15 m
ks = 6105 m
= 1.0106 m2 s1
Inspect the head-loss equation:
L V2
h ( )
D 2g
We can get V from Q and D, but to find h we will require the friction factor.
First V:
V

Q
A

Q
D 2 / 4

0.03
0.15 2 / 4

1.698 m s 1

Inspect the Colebrook-White equation:

1
2.51
k
2.0 log 10 s

3.7 D Re
To use this we require the Reynolds number:
VD
1.698 0.15
Re

254700

1.0 10 6
Substituting values for ks, D and Re in the Colebrook-White equation and rearranging for :
1

9.854 10 6
4
)
2.0 log 10 (1.081 10

Iterating from an initial guess, with successive values substituted into the RHS:
Initial guess:
= 0.01
First iteration
= 0.01841
Second iteration
= 0.01784
Third iteration
= 0.01787
Fourth iteration
= 0.01787
can then be substituted in the head-loss equation to derive h:
L V2
500 1.698 2
h ( ) 0.01787

8.753 m
D 2g
0.15 2 9.81
Answer: head loss = 8.8 m.

Hydraulics 2

T2-13

David Apsley

Type 3 (Sizing problem): Quantity of flow Q and available head h known; find the required
diameter D.
Example. A flow of 0.4 m3 s1 is to be conveyed from a headworks at 1050 m AOD to a
treatment plant at 1000 m AOD. The length of the pipeline is 5 km. Estimate the required
diameter, assuming that ks = 0.03 mm.
Solution.
List known parameters:
Q = 0.4 m3 s1
h = 50 m
L = 5000 m
ks = 3105 m
= 1.0106 m2 s1
Before iterating, try to write D in terms of . From the head-loss equation:

L V2
L Q
( )

D 2g
2 gD A

L Q

2 gD D 2 / 4

8LQ 2
2 gD 5

1/ 5

8LQ 2

D 2
gh
Substituting values of Q, L and h gives a working expression (with D in metres):
D (1.322 )1 / 5

The Colebrook-White equation for is:

1
2.51
k
2.0 log 10 s

3.7 D Re
The Reynolds number can be written in terms of the diameter D:
VD
4Q
5.093 10 5
Q D
Re

D
D
D / 4
Substituting this expression for Re we obtain an iterative formula for :
1

8.108 10 6 4.928 10 6 D

)
2.0 log 10 (
D

(*)

(**)

Iterate (*) and (**) in turn, until convergence.

Guess:
= 0.01

D = 0.4210 m
Iteration 1:
= 0.01293

D = 0.4432 m
Iteration 2:
= 0.01276

D = 0.4420 m
Iteration 3:
= 0.01277

D = 0.4421 m
Iteration 4:
= 0.01277

D = 0.4421 m
Answer: required diameter = 0.44 m.
In practice, commercial pipes are only made with certain standard diameters and the next
available larger diameter should be chosen.

Hydraulics 2

T2-14

David Apsley

Both Pipe Friction and Minor Losses

In some circumstances minor losses (including exit losses) actually contribute a significant
proportion of the total head loss and must be included in the head-loss equation
L
V2
h ( K )( )
D
2g
An iterative solution in conjunction with the Colebrook-White equation is then inevitable,
irrespective of whether the problem is of type 1, 2 or 3.

Example.
A reservoir is to be used to supply water to a factory 5 km away. The water level in the
reservoir is 60 m above the factory. The pipe lining has roughness 0.5 mm. Minor losses due
to valves and pipe fittings can be accommodated by a loss coefficient K = 80. Calculate the
minimum diameter of pipe required to convey a discharge of 0.3 m3 s1.
Answer: 0.44 m

Hydraulics 2

T2-15

David Apsley

1.8 Energy and Hydraulic Grade Lines

Energy grade lines and hydraulic grade lines are graphical means of portraying the energy
changes along a pipeline.
Three elevations may be drawn:
pipe centreline:

geometric height

hydraulic grade line (HGL):

p
z
g

piezometric head

energy grade line (EGL):

p
V2
z
g
2g

total head

p is the gauge pressure (i.e. pressure difference from atmospheric pressure).

Illustrations

reservoir

energy
grade li
ne
hydrauli
c grade
line

Pipe friction only

pipeline

Pipe friction with minor

losses (exaggerated),
including change in pipe
diameter.

V2/2g
p/g

entry loss
EGL
HGL

reservoir

exit loss

pipeline

Pumped system

EGL
HGL

pipeline

pump

Hydraulics 2

T2-16

David Apsley

Energy Grade Line (EGL)

Shows the change in total head along the pipeline.

Starts and ends at still-water levels.

Steady downward slope reflects pipe friction (slope change if pipe radius changes);
small discontinuities correspond to minor losses;
large discontinuities correspond to turbines (loss of head) or pumps (gain of head).

The EGL represents the maximum height to which water may be delivered.

Hydraulic Grade Line (HGL)

Shows the change in piezometric head along the pipeline.

For pipe flow the HGL lies a distance p/g above the pipe centreline. Thus, the
difference between pipe elevation and hydraulic grade line gives the static pressure p.
If the HGL drops below pipe elevation this means negative gauge pressures (i.e. less
than atmospheric). This is generally undesirable since:
extraneous matter may be sucked into the pipe through any leaks;
for very negative gauge pressures, dissolved gases may come out of solution and
cause cavitation damage.
An HGL more than patm/g ( 10 m of water) below the pipeline is impossible.

The HGL is the height to which the liquid would rise in a piezometer tube.

For open-channel flows, pressure is atmospheric (i.e. p = 0) at the surface; the HGL is
then the height of the free surface.

The EGL is always higher than the HGL by an amount equal to the dynamic head V 2 / 2 g .
For uniform pipes (constant V), the two grade lines are parallel.

Hydraulics 2

T2-17

David Apsley

1.9 Simple Pipe Networks

For all pipe networks the following basic principles apply:
(1)
continuity at junctions (total flow in = total flow out);
(2)
the head is uniquely defined at any point;
(3)
each pipe satisfies its individual resistance law (i.e. head-loss vs discharge relation):
h = Q2
The last of these comes from the proportionality between head loss and dynamic head, i.e.
L
V2
4Q
where
V
h ( K )
D
2g
D 2
is the friction factor and K is the sum of minor loss coefficients.
For hand calculations, is often taken as a constant for each pipe (although, in reality, it will
vary slightly with flow rate).
There is a close analogy with electrical networks:
head H

potential V
discharge Q

current I
However, the hydraulic equivalent of Ohms law is usually non-linear:
head loss H Q2
potential difference V I

1.9.1 Pipes in Series

1
Q1 = Q2
H = H1 + H2

same flow
add the head changes

1.9.2 Pipes in Parallel

1
2
H1 = H2
Q = Q1 + Q2

Hydraulics 2

same head change

add the flows

T2-18

David Apsley

1.9.3 Branched Pipes Single Junction

The simplest case is three pipes meeting at a single junction.
If the flows are known then the heads can be determined (relative to the head at one point) by
calculating the head losses along each pipe.

If, however, the heads HA, HB and HC are known

(for example, from the water levels in reservoirs)
then we have a classic problem known as the threereservoir problem.

B
J ?

The head at J is adjusted (iteratively) to satisfy:

(a) the loss equation ( h Q 2 ) for each pipe; i.e:

2
H J H A JA QJA
2
H J H B JB QJB
2
H J H C JC QJC
(b) continuity at the junction J:
net flow out of junction QJA QJB QJC 0

Note the sign convention: QJA is the flow from J to A; it will be negative if the flow actually
goes from A to J. The direction of flow in any pipe is always from high head to low head.
Although we consider only 3 reservoirs, the problem and its solution method clearly
generalise to any number of reservoirs.
Solution Procedure
(0) Establish the head-loss vs discharge equations for each pipe;
(1) Guess an initial head at the junction, HJ;
(2) Calculate flow rates in all pipes (from the head differences);
(3) Calculate net flow out of J;
(4) As necessary, adjust HJ to reduce any flow imbalance and repeat from (2).
If the direction of flow in a pipe, say JB, is not obvious then a good initial guess is to set
HJ = HB so that there is initially no flow in this pipe. The first flow-rate calculation will then
establish whether HJ should be lowered or raised and hence the direction of flow in this pipe.
Example.
Reservoirs A, B and C have constant water levels of 150, 120 and 90 m respectively above
datum and are connected by pipes to a single junction J at elevation 125 m. The length (L),
diameter (D), friction factor () and minor-loss coefficient (K) of each pipe are given below.
Pipe
JA
JB
JC

L (m)
1600
1600
2400

D (m)
0.3
0.2
0.25

0.015
0.015
0.025

K
40
25
50

(a) Calculate the flow in each pipe.

(b) Calculate the reading of a Bourdon pressure gauge attached to the junction J.
Hydraulics 2

T2-19

David Apsley

Solution.
First, prepare head-loss vs discharge relations for each pipe:
L
V2
Q
Q
where
V

h ( K )
A
D
2g
D 2 / 4
L
8

h ( K ) 2 4 Q 2
D
gD
Substituting L, D, and K for each pipe we obtain the head-loss vs discharge relationships:
H J 150
2
H J H A 1224QJA
Pipe AJ:
or
QJA
1224
Pipe JB:

2
H J H B 7488QJB

or

QJB

Pipe JC:

2
H J H C 6134QJB

or

QJC

H J 120
7488
H J 90
6134

The value of HJ is varied until the net flow out of J is 0.

If there is net flow into the junction then HJ needs to be raised.

If there is net flow out of the junction then HJ needs to be lowered.

After the first two guesses at HJ, subsequent iterations are guided by interpolation.
The working is conveniently set out in a table.
HJ (m)
QJA (m3 s1)
QJB (m3 s1)

H J 150
1224

0.1566
0.0904
0.1196
0.1219

120
140
132.5
131.8

H J 120
7488

0.0000
0.0517
0.0409
0.0397

QJC (m3 s1)

H J 90
6134

Net flow out of J

(m3 s1)
= QJA + QJB+ QJC

0.0699
0.0903
0.0832
0.0825

0.0867
0.0516
0.0045
0.0003

This is sufficient accuracy (0.0003/0.1219 or about 0.25%). The quantity of flow in each pipe
is given in the bottom row of the table, with the direction implied by the sign.
(b) A Bourdon gauge measures absolute pressure. From the piezometric head at the junction:
p
HJ
z
g
where p is the gauge pressure. Hence,
p g ( H J z ) 1000 9.81 (131.8 125)
67000 Pa

Taking atmospheric pressure as 101000 Pa, the absolute pressure is then

101000 + 67000 = 168000 Pa
Answer: 1.68 bar.

Hydraulics 2

T2-20

David Apsley

1.10 Complex Pipe Networks (Optional)

1.10.1 Loop Method (Hardy-Cross, 1936)

loop 1

Used for networks made up of a series of closed

loops, where the external flows are known.

loop 2

Basic Idea
Start with any flow satisfying continuity. Apply iterative flow corrections Q until the net
head change round each loop is 0.

Adopt a suitable sign convention (e.g. Q positive if clockwise) in each loop. The signed head
loss for any particular pipe is then
(19)
h sQ 2
where s = +1 if Q is positive and 1 if Q is negative.
Initially, the net head loss round a closed loop probably wont be 0. To try to achieve this
after perturbing the flow in all pipes of a loop by Q we require
s(Q Q) 2 0
loop

where Q is the same for every pipe in the loop. Expanding:

sQ 2 2 sQQ s Q 2 0
Neglecting the Q2 term and noting that sQ Q leads to a flow correction for this loop of

sQ
2 Q

(20)

This is applied to every pipe in the loop.

Algorithm
Divide the network into closed loops.
Start with any flow satisfying continuity.
For each loop in turn:

sQ
calculate Q
2 Q

update all pipes in this loop by Q;

Repeat until the net head change around all loops is sufficiently small.

An example with two loops is given on the Example Sheet.

Hydraulics 2

T2-21

David Apsley

1.10.2 Nodal Method (Cornish, 1939)

Used for loops or branches where the external heads are known.

H1

H2

Basic Idea
Start with guessed heads Hi at each internal junction and
calculate the resulting flow in each pipe. Apply iterative head
corrections Hi so as to satisfy continuity at each junction.
(The 3-reservoir problem is the simplest case, with a single
junction).

H3

H4

As in the previous subsection, head changes at junctions i and j cause a change in the flow
between them:
Hi H j
H i H j 2sij ij Qij Qij 2
Qij
(21)
Qij
where Qij is the flow rate from the ith node to the jth node, with appropriate sign.
Initially, the net outflow at the ith junction wont be 0; to try to achieve this we aim to perturb
the flow so that
(Qij Qij ) 0
j

or

Q 2( H

Qij

(H i H j ) 0,
i 1, 2, 3,
Hj)
Taken over all junctions i this gives a set of simultaneous equations for the Hi.
ij

(22)

A 2-junction example is given on the Example Sheet.

Hydraulics 2

T2-22

David Apsley

2. OPEN-CHANNEL FLOW
Flow in open channels (e.g. rivers, canals, guttering, ...) and partially-full conduits (e.g.
sewers) is characterised by the presence of a free surface where the pressure is atmospheric.
Unlike pipe flow, open-channel flow is always driven by gravity, not pressure.
PIPE FLOW

OPEN-CHANNEL FLOW

LIQUIDS or GASES

LIQUIDS (free surface)

Driven by: PRESSURE, GRAVITY or BOTH

GRAVITY (down slope)

Fluid:

Size:

DIAMETER

HYDRAULIC RADIUS

Volume:

FILLS pipe

Depends on DEPTH

Equations: DARCY-WEISBACH (head loss)

COLEBROOK-WHITE (friction factor)

MANNINGS FORMULA

2.1 Normal Flow

The flow is uniform if the velocity profile does not change along the channel. (This is at best
an approximation for natural channels like rivers where the channel cross-section changes.)
The flow is steady if it does not change with time.
Steady uniform flow is called normal flow and the depth of water is called the normal depth.
The normal depth h depends on the discharge Q.

EGL
V 2 /2g

HGL (fr
ee surfac
e): p = 0

hf

L
In normal flow equal hydrostatic pressure forces at any cross-section mean no net pressure
force. Hence, the downslope component of weight balances bed friction;
Note.
The following assumes the slope to be sufficiently small for there to be negligible difference
between the depth h measured vertically (which determines the energy level) and that
perpendicular to the bed of the channel (which determines the flow rate).
Hydraulics 2

T2-23

David Apsley

2.2 Hydraulic Radius and the Drag Law

In both open channels and partially-full pipes, wall friction
occurs only along the wetted perimeter.

A
P

Let A be the cross-sectional area occupied by fluid and P the

wetted perimeter.

For steady, uniform flow, the component of

weight down the slope balances bed friction:
(AL) g sin b PL
where b is the average wall friction. Hence,
A
b g ( ) sin
P

mg

Define:
Hydraulic radius

A cross - sectional area

P
wetted perimeter
Note that, in general, the hydraulic radius depends on depth.
Rh

(23)

Hence, for normal flow,

Normal flow relationship

b gRh S

(24)

where S (= drop length) is the slope. (We have assumed tan sin for small angles.)

Examples.
(1)
For a circular pipe running full,
A
R 2
Rh

12 R
(25)
P
2R
i.e. for a full circular pipe, the hydraulic radius is half the geometric radius. (Sorry
folks, this is just one of those things!). As a result, it is common to define a hydraulic
diameter Dh by
Dh 4Rh
(2)

For a rectangular channel of width b with water depth h,

A
bh
h
Rh

P
b 2h
1 2h / b
For a very wide channel h/b << 1, and hence
Rh h
i.e. Rh is equal to the depth of flow.

To progress we need an expression for the average bed stress b.

Hydraulics 2

T2-24

David Apsley

2.3 Friction Laws Chzy and Mannings Formulae

From the balance of forces above:
b gRh S
In principle, a (skin-)friction coefficient can be used to relate the (average) bed shear stress to
the dynamic pressure. Hence,
c f ( 12 V 2 ) gRh S
(26)
Friction factors = 4cf based on the Colebrook-White equation (using 4Rh as the hydraulic
diameter) are unsatisfactory for open conduits because the shear stress is not constant on the
wetted perimeter. Instead, engineers tend to use simpler empirical formulae due to Chzy8
and Manning9.
Rearranging equation (26) gives:
2g
V2
Rh S
cf
whence
Chzys Formula
V C Rh S

(27)

C ( 2 g/c f ) is Chzys coefficient. This gives the variation with slope for a particular
channel, but it is not a helpful design equation because C varies with channel roughness and
hydraulic radius.
The most popular correlation for C is that of Manning who proposed, on the basis of a review
of experimental data, that
C Rh1 / 6 function of roughness
which he chose to write as
R1 / 6
C h
n
Combined with Chzys formula (27), this yields:
Mannings Formula
V

1 2 / 3 1/ 2
Rh S
n

(28)

Very important.
Both Chzys C and Mannings n are dimensional and depend on the units used. Typical
values of n in metre-second units are given in the Appendix. Typical figures for artificiallylined channels and natural water courses are 0.015 m1/3 s and 0.035 m1/3 s respectively.

Antoine Chzy (1718-1798); French engineer who carried out experiments on the Seine and on the Courpalet
Canal.
9
Robert Manning (1816-1897); Irish engineer. Actually, if you live on the wrong side of the English Channel
then what we call Mannings equation is variously ascribed to Gauckler and/or Strickler.

Hydraulics 2

T2-25

David Apsley

2.4 Uniform-Flow Calculations

Assuming that the channel slope, shape and lining material are known, there are two main
classes of problem:
(Type A - easy) Given the depth (h) determine the quantity of flow (Q)
Calculate:
(1)
cross-sectional area A and wetted perimeter P from geometry of channel;
A
(2)
hydraulic radius Rh .
P
1
(3)
average velocity from Mannings formula: V Rh2 / 3 S 1 / 2
n
(4)
quantity of flow as velocity area: Q = VA.

(Type B - harder) Given the quantity of flow (Q) determine the depth (h)
(1)

Follow the steps for Type A above to write algebraic expressions for,
successively, A and P, Rh, V, Q in terms of depth h.

(2)

Invert the Q vs h relationship graphically or numerically.

Example. A smooth concrete-lined channel has trapezoidal cross-section with base width
6 m and sides of slope 1V:2H. If the bed slope is 1 in 500 and the normal depth is 2 m
calculate the quantity of flow.

Solution.
We are given slope S = 0.002. From the Appendix, Mannings n is 0.012 m1/3 s.

2m
6m

4m

Break the trapezoidal section into rectangular and triangular elements to obtain, successively:
Area:
A 6 2 2 12 4 2 20 m 2
Wetted perimeter:
Hydraulic radius:
Average velocity:
Quantity of flow:

P 6 2 4 2 2 2 14.94 m
A
20
Rh

1.339 m
P
14.94
1
V Rh2 / 3 S 1 / 2 3.727 Rh2 / 3 4.528 m s 1
n
Q VA 4.528 20 90.56 m 3 s 1

Answer: 91 m3 s1.

Hydraulics 2

T2-26

David Apsley

Example. For the channel above, if the quantity of flow is 40 m3 s1, what is the normal
depth?

Solution. This time we need to leave all quantities as functions of height h. In metre-second
units we have the following.
Area:
A 6h 2h 2
Wetted perimeter:
P 62 5 h
A
Hydraulic radius:
Rh
P
Average velocity:
V 3.727 Rh2 / 3
Quantity of flow:
Q VA
Now try a few values of h.
h (m)
2
1
1.23
1.28
1.30

A (m s1)
20.00
8.00
10.41
10.96
11.18

P (m)
14.94
10.47
11.50
11.72
11.81

Rh (m)
1.339
0.7641
0.9052
0.9352
0.9467

V (m s1)
4.528
3.115
3.488
3.564
3.593

Q (m3 s1)
90.56
24.92
36.31
39.06
40.17

After the first two guesses, subsequent choices of h home in on the solution by
interpolating/extrapolating from previous results.
Answer: h = 1.30 m
If desired, the sequence of calculations could be assembled into a single formula for Q(h):
2/3

1 A
S A5 / 3
Q VA S 1 / 2 A
nP
n P2/3
With the given values of S and n, and the formulae for A and P:
(6 h 2 h 2 ) 5 / 3
Q 3.727
(6 2 5h) 2 / 3
A sequence of trial values of h is then used as above. (Small numerical differences are due to
rounding).
h (m)
Q (m3 s1)
2
90.52
1
24.92
1.23
36.28
1.28
39.03
1.30
40.16

Note. Microsoft Excel is good for inverting Q(h) (use the Goal-Seek or Solver tools).

Hydraulics 2

T2-27

David Apsley

2.5 Conveyance
Combining Mannings formula for average velocity ( V (1 / n) Rh2 / 3 S 1 / 2 ) with expressions for
hydraulic radius ( Rh A / P ) and discharge ( Q VA ) we obtain:

Q KS 1 / 2

(29)

where

1 A5 / 3
n P2/3

(30)

K is a function of the channel geometry and the roughness of its lining. It is called the
conveyance of the channel and is a measure of the channels discharge-carrying capacity.

1
The primary use of K is in determining the
discharge capacity of compound channels for
flood plain
example river and flood plain. By adding the
contribution to total discharge from individual
components with different roughness:
Q Qi K i S 1 / 2 K eff S 1 / 2
the total conveyance is simply the sum of the separate conveyances:
K eff K i

river

flood plain

2.6 Optimal Shape of Cross-Section

Expressions for A, P and Rh for important channel shapes are given below.
rectangle

trapezoid

h2
tan
2h
b
sin
b h/ tan
h
b 2h/ sin

bh

wetted perimeter P

b 2h

hydraulic radius Rh

h
1 2h/b

R
h

cross-sectional
area A

circle

bh

R 2 ( 12 sin 2)
2R

R sin 2
1

2
2

The most hydraulically-efficient shape of channel is the one which can pass the greatest
quantity of flow for any given area. This occurs for the minimum hydraulic radius or,
equivalently, for the minimum wetted perimeter corresponding to the given area.
A semi-circle is the most hydraulically-efficient of all channel cross-sections. However,
hydraulic efficiency is not the only consideration and one must also consider, for example,

Hydraulics 2

T2-28

David Apsley

fabrication costs, excavation and, for loose granular linings, the maximum slope of the sides.
Many applications favour trapezoidal channels.

Trapezoidal Channels
For a trapezoidal channel:
cross-sectional area:
wetted perimeter:

h2
tan
2h
P b
sin
A bh

What depth of flow and what angle of side give maximum hydraulic efficiency?
To minimise the wetted perimeter for maximum hydraulic efficiency, we substitute for b in
terms of the fixed area A:
2h
A
h
2h
A
2
1
(31)
P b
(
)
h(

)
sin
h tan sin
h
sin tan
To minimise P with respect to water depth we set
P
A
2
1
2 (

)0
h
sin tan
h
and, on substituting the bracketed term into the expression (31) for P, we obtain
2A
P
h
The hydraulic radius is then
A
h
Rh

P
2
In other words, for maximum hydraulic efficiency, a trapezoidal channel should be so
proportioned that its hydraulic radius is half the depth of flow.
Similarly, to minimise P with respect to the angle of slope of the sides, , we set
P
2
1
h
h( 2 cos
sec 2 )
(1 2 cos ) 0
2

sin
tan
sin 2
This occurs when cos = . The most efficient side angle for a trapezoidal channel is 60.
Substituting these results for h and into the general expression for Rh one obtains
h/b 3/ 2 ; i.e. the most hydraulically-efficient trapezoidal channel shape is half a
regular hexagon.

Circular Ducts
In similar fashion it can be shown that the maximum quantity of flow for a circular duct
actually occurs when the duct is not full in fact for a depth about 94% of the diameter
(Exercise. Prove it; then try to explain in words why you might expect this).

Hydraulics 2

T2-29

David Apsley

Appendix
Material
Riveted steel
Concrete
Wood stave
Cast iron
Galvanised iron
Asphalted cast iron
Commercial steel or wrought iron
Drawn tubing
Glass

ks (mm)
0.9 9.0
0.3 3.0
0.18 0.9
0.26
0.15
0.12
0.046
0.0015
0 (smooth)

Table 1. Typical roughness for commercial pipes (from White, 2011).

n (m1/3 s)
Artificial lined channels:
Glass
Brass
Steel, smooth
painted
riveted
Cast iron
Concrete, finished
unfinished
Planed wood
Clay tile
Brickwork
Asphalt
Corrugated metal
Rubble masonry
Excavated earth channels:
Clean
Gravelly
Weedy
Stony, cobbles
Natural channels:
Clean and straight
Sluggish, deep pools
Major rivers
Floodplains:
Pasture, farmland
Light brush
Heavy brush
Trees

0.01
0.011
0.012
0.014
0.015
0.013
0.012
0.014
0.012
0.014
0.015
0.016
0.022
0.025
0.022
0.025
0.03
0.035
0.03
0.04
0.035
0.035
0.05
0.075
0.15

Table 2. Typical values of Mannings n (from White, 2011).

Hydraulics 2

T2-30

David Apsley