14

Physical modelling of
hydraulics
14.1 Introduction
Denition: the physical hydraulic model
A physical model is a scaled representation of a hydraulic ow situation. Both the
boundary conditions (e.g. channel bed, sidewalls), the upstream ow conditions
and the ow eld must be scaled in an appropriate manner (Fig. 14.1).
Physical hydraulic models are commonly used during design stages to optimize a
structure and to ensure a safe operation of the structure. They have an important
further role to assist non-engineering people during the `decision-making' process.
A hydraulic model may help the decision-makers to visualize and to picture the
ow eld, before selecting a `suitable' design.
In civil engineering applications, a physical hydraulic model is usually a smaller-
size representation of the prototype (i.e. the full-scale structure) (e.g. Fig. 14.2).
Other applications of model studies (e.g. water treatment plant, otation column)
may require the use of models larger than the prototype. In any case the model is
investigated in a laboratory under controlled conditions.

Discussion
Hydraulic modelling cannot be disassociated from the basic theory of uid
mechanics. To be ecient and useful, experimental investigations require theoretical
guidance which derives primarily from the basic principles (see Chapter 13) and the
theory of similarity (see the next subsection).
In the present section, we will consider the physical modelling of hydraulic ows:
i.e. the use of laboratory models (with controlled ow conditions) to predict the
behaviour of prototype ow situations.

14.2 Basic principles

In a physical model, the ow conditions are said to be similar to those in the prototype
if the model displays similarity of form (geometric similarity), similarity of motion
(kinematic similarity) and similarity of forces (dynamic similarity).

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
262 Physical modelling of hydraulics

Fig. 14.1 Basic ow parameters.

Basic scale ratios

Geometric similarity implies that the ratios of prototype characteristic lengths to
model lengths are equal:
lp dp Hp
Lr Length 14:1
l m dm H m
where the subscripts p and m refer to prototype (full-scale) and model parameters
respectively, and the subscript r indicates the ratio of prototype-to-model quantity.
Length, area and volume are the parameters involved in geometric similitude.
Kinematic similarity implies that the ratios of prototype characteristic velocities to
model velocities are the same:
Vp V1 p V2 p
Vr Velocity 14:2
Vm V1 m V2 m
Dynamic similarity implies that the ratios of prototype forces to model forces are
equal:
F1p F2p
Fr Force 14:3
F1m F2m

Work and power are other parameters involved in dynamic similitude.

Notes
1. Geometric similarity is not enough to ensure that the ow patterns are similar
in both model and prototype (i.e. kinematic similarity).
2. The combined geometric and kinematic similarities give the prototype-to-
model ratios of time, acceleration, discharge, angular velocity.

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.2 Basic principles 263

(a)

(b)

Fig. 14.2 Example of physical model: breakwater jetty for the Changhua Reclamation area, along the north-
west coastline of Taiwan, Republic of China (January 1994). (a) Prototype breakwater jetty. (b) Model break-
water jetty in a wave ume (Tainan Hydraulic Laboratory).

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
264 Physical modelling of hydraulics

Subsequent scale ratios

The quantities Lr , Vr and Fr (dened in equations (14.1) to (14.3)) are the basic scale
ratios. Several scale ratios can be deduced from equations (14.1) to (14.3):

Mr r L3r Mass 14:4

Lr
tr Time 14:5
Vr
Qr Vr L2r Discharge 14:6
Fr
Pr Pressure 14:7
L2r
where  is the uid density. Further scale ratios may be deduced in particular ow
situations.

Application
In open channel ows, the presence of the free-surface
p means that gravity eects
are important. The Froude number Fr V= gL is always signicant. Second-
ary scale ratios can be derived from the constancy of the Froude number1 which
implies:
p
Vr Lr Velocity
Other scale ratios are derived from the Froude similarity (e.g. Henderson 1966):
5=2
Qr Vr L2r Lr Discharge
Mr L r
Fr r L3r Force
Tr2
F
Pr 2r r Lr Pressure
Lr

14.3 Dimensional analysis

14.3.1 Basic parameters

The basic relevant parameters needed for any dimensional analysis (Fig. 14.1) may be
grouped into the following groups.
(a) Fluid properties and physical constants (see Appendix A1.1). These consist of the
density of water  (kg/m3 ), the dynamic viscosity of water  (N s/m2 ), the surface
tension of air and water  (N/m), the bulk modulus of elasticity of water Eb (Pa),
and the acceleration of gravity g (m/s2 ).

1
It is assumed that the gravity acceleration is identical in both the model and the prototype.

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.3 Dimensional analysis 265

(b) Channel (or ow) geometry. These may consist of the characteristic length(s) L (m).
(c) Flow properties. These consist of the velocity(ies) V (m/s) and the pressure
dierence(s)
P (Pa).

14.3.2 Dimensional analysis

Taking into account all basic parameters, dimensional analysis yields:

F1 ; ; ; Eb ; g; L; V;
P 0 14:8
There are eight basic parameters and the dimensions of these can be grouped into
three categories: mass (M), length (L) and time (T). The Buckingham -theorem
(Buckingham 1915) implies that the quantities can be grouped into ve (5 8 3)
independent dimensionless parameters:
0 1
B V V 2 VL V V C
F2 B
@p ; ; ; r ; rC 14:9a
gL
P   Eb A
L 
F2 Fr; Eu; Re; We; Ma 14:9b
The rst ratio is the Froude number Fr, characterizing the ratio of the inertial force to
gravity force. Eu is the Euler number, proportional to the ratio of inertial force to
pressure force. The third dimensionless parameter is the Reynolds number Re
which characterizes the ratio of inertial force to viscous force. The Weber number
We is proportional to the ratio of inertial force to capillary force (i.e. surface tension).
The last parameter is the SarrauMach number, characterizing the ratio of inertial
force to elasticity force.

Notes
1. The Froude number is used generally for scaling free surface ows, open
channels and hydraulic structures. Although the dimensionless number was
named after William Froude (18101879), several French researchers used
it before: e.g. Belanger (1828), Dupuit (1848), Bresse (1860), Bazin (1865a).
Ferdinand Reech (18051880) introduced the dimensionless number for test-
ing ships and propellers in 1852, and the number should really be called the
ReechFroude number.
2. Leonhard Euler (17071783) was a Swiss mathematician and physicist, and a
close friend of Daniel Bernoulli.
3. Osborne Reynolds (18421912) was a British physicist and mathematician
who expressed rst the `Reynolds number' (Reynolds 1883).
4. The Weber number characterizing the ratio of inertial force over surface
tension force was named after Moritz Weber (18711951), German Professor
at the Polytechnic Institute of Berlin.
5. The SarrauMach number is named after Professor Sarrau who rst high-
lighted the signicance of the number (Sarrau 1884) and E. Mach who

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
266 Physical modelling of hydraulics

introduced it in 1887. The SarrauMach number was once called the Cauchy
number as a tribute to Cauchy's contribution to wave motion analysis.

Discussion
Any combination of the dimensionless numbers involved in equation (14.9) is
also dimensionless and may be used to replace one of the combinations. It can
be shown that one parameter can be replaced by the Morton number
Mo g4 =3 , also called the liquid parameter, since:
We3
Mo 14:10
Fr2 Re4
The Morton number is a function only of uid properties and the gravity
constant. For the same uids (air and water) in both model and prototype,
Mo is a constant (i.e. Mop Mom ).

14.3.3 Dynamic similarity

Traditionally model studies are performed using geometrically similar models. In a

geometrically similar model, true dynamic similarity is achieved if and only if each
dimensionless parameter (or -terms) has the same value in both model and proto-
type:
Frp Frm ; Eup Eum ; Rep Rem ; Wep Wem ; Map Mam 14:11
Scale eects will exist when one or more -terms have dierent values in the model
and prototype.

Practical considerations
In practice, hydraulic model tests are performed under controlled ow conditions.
The pressure dierence
P may usually be controlled. This enables
P to be treated
as a dependent parameter. Further compressibility eects are small in clear-water
ows2 and the SarrauMach number is usually very small in both model and proto-
type. Hence, dynamic similarity in most hydraulic models is governed by:
0 1

P B V VL V C
F3 B
@p ; ; rC 14:12a
V 2 gL  A
L
Eu F3 Fr; Re; We Hydraulic model tests 14:12b
There are a multitude of phenomena that might be important in hydraulic ow
situations: e.g. viscous eects, surface tension, gravity eect. The use of the same
uid on both prototype and model prohibits simultaneously satisfying the Froude,
Reynolds and Weber number scaling criteria (equation (14.12)) because the Froude
2
This statement is not true in airwater ows (e.g. free-surface aerated ows) as the sound celerity may
decrease to about 20 m/s for 50% volume air content (e.g. Cain 1978, Chanson 1997).

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.3 Dimensional analysis 267
p
number similarity requires Vr Lr , the Reynolds numberpscaling implies that
Vr 1=Lr and the Weber number similarity requires: Vr 1= Lr .
In most cases, only the most dominant mechanism is modelled. Hydraulic models
commonly use water and/or air as owing uid(s). In fully-enclosed ows (e.g. pipe
ows), the pressure losses are basically related to the Reynolds number Re. Hence,
a Reynolds number scaling is used: i.e. the Reynolds number is the same in both
model and prototype. In free-surface ows (i.e. ows with a free surface), gravity
eects are always important and a Froude number modelling is used (i.e.
Frm Frp ) (e.g. Fig. 14.2).

Discussion
When inertial and surface tension forces are dominant, a Weber number similar-
ity must be selected. Studies involving air entrainment in owing waters (i.e.
white waters), de-aeration in shaft or bubble plumes are often based upon a
Weber number scaling.
The Euler number is used in practice for the scaling of models using air rather
than water: e.g. hydraulic models in wind tunnels, or a manifold system with
water ow which is scaled at a smaller size with an air ow system.

14.3.4 Scale effects

Scale eects may be dened as the distortions introduced by eects (e.g. viscosity,
surface tension) other than the dominant one (e.g. gravity in free-surface ows).
They take place when one or more dimensionless parameters (see Section 14.3.3)
dier between model and prototype.
Scale eects are often small but they are not always negligible altogether. Consid-
ering an overow above a weir, the uid is subjected to some viscous resistance along
the invert. However the ow above the crest is not signicantly aected by resistance,
the viscous eects are small and the dischargehead relationship can be deduced as
for ideal-uid ow.
In free-surface ows, the gravity eect is dominant. If the same uid (i.e. water) is
used in both the model and the prototype, it is impossible to keep both the Froude
and Reynolds numbers in the model and full-scale. Indeed it is elementary to show
3=2
that a Froude similitude implies Rer Lr , and the Reynolds number becomes
much smaller in the model than in the prototype (if Lr < 1).
Note that dierent uids may be used to have the same Reynolds and Froude
numbers in both the model and prototype, but this expedient is often not practical
nor economical.

Some examples of scale effects

Example No. 1
Considering the drag exerted on two-dimensional bodies, Fig. 14.3 shows the
eects of the Reynolds number on the drag coecient. Dynamic similarity

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
268 Physical modelling of hydraulics

Fig. 14.3 Drag coefcient on two-dimensional bodies.

(equation (14.12)) requires the drag coecient to be the same in the model
and prototype. If the Reynolds number is smaller in the model than at full-
scale (in most practical cases), Fig. 14.3 suggests that the model drag
coecient would be larger than that of the prototype and dynamic similarity
could not be achieved. Moreover, the drag force comprises the form drag
and the surface drag (i.e. skin friction). In small-size models, the surface
drag might become predominant, particularly if the model ow is not fully-
rough turbulent or the geometrical scaling of roughness height is not
achievable.
In practice, an important rule, in model studies is that the model Reynolds
number Rem should be kept as large as possible, so that the model ow is
fully-rough turbulent (if prototype ow conditions are fully-rough turbulent).

Example No. 2
Another example is the eect of the corner radius on the drag force on two-
dimensional bodies (Fig. 14.4). Figure 14.4 shows signicant dierences in the
Reynolds numberdrag coecient relationships depending upon the relative
radius r=D. When the corner radius on the prototype is small and Lr is
large, it is impossible to have the same ratio of corner radius to body size
in the model and prototype because the model cannot be manufactured
with the required accuracy. In such cases, the drag force is not scaled
adequately.

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.3 Dimensional analysis 269

Fig. 14.4 Effect of corner radius and surface roughness on the drag coefcient of two-dimensional
bodies.

Example No. 3
A dierent example is the ow resistance of bridge piers. Henderson (1966)
showed that the resistance to ow of normal bridge pier shapes is such that
the drag coecient is about or over unity, implying that the form drag is a
signicant component of the total drag. In such a case, the viscous eects are
relatively small, and dynamic similarity is achievable, provided that model
viscous eects remain negligible.

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
270 Physical modelling of hydraulics

Discussion
If scale eects would become signicant in a model, a smaller prototype-to-model
scale ratio Lr should be considered to minimize the scale eects. For example, in a
100: 1 scale model of an open channel, the gravity eect is predominant but viscous
eects might be signicant. A geometric scale ratio of 50 : 1 or 25: 1 may be considered
to reduce or eliminate viscous scale eects.
Another example is the entrainment of air bubbles in free-surface ows. Gravity
eects are predominant but it is recognized that surface tension scale eects can
take place for Lr > 10 to 20 (or Lr < 0:05 to 0.1) (e.g. Wood 1991, Chanson 1997).
At the limit, no scale eect is observed at full-scale (i.e. Lr 1) as all the -terms
(equation (14.11)) have the same values in the prototype and model when Lr 1.

14.4 Modelling fully-enclosed ows

14.4.1 Reynolds models
Fully-enclosed ow situations include pipe ows, turbomachines and valves. For such
ow situations, viscosity eects on the solid boundaries are important. Physical
modelling is usually performed with a Reynolds similitude: i.e. the Reynolds
number is kept identical in both the model and prototype:
Rep Rem 14:13
If the same uid is used in both the model and prototype, equation (14.13) implies:
Vr 1=Lr (Reynolds similitude)
For Lr > 1, the model velocity must be larger than that in the prototype.

Discussion
For example, if the model scale is 10 :1 (i.e. Lr 10), the velocity in the model
must be ten times that in the prototype. By using a dierent uid in the
model, the ratio r =r becomes dierent from unity and Vm can be reduced.

14.4.2 Discussion
Flow resistance in pipe ows
For pipe ows, the Darcy equation relates the pressure losses to the pipe geometry
(diameter D, length L) and to the ow velocity V:
L V 2

P f 14:14
D 2
where f is the DarcyWeisbach friction factor. After transformation and combining
with equation (14.10), it leads:
fL
Eu F4 Fr; Re; We; Ma; . . . 14:15
2D

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.4 Modelling fully-enclosed ows 271

Fig. 14.5 Friction factor versus Reynolds number in pipe ows.

In pipe ows, gravity and surface tension have no eect on the pressure losses.
For steady liquid ows, the compressibility eects are negligible. The roughness
height ks is, however, an additional characteristic length. For a uniformly distribu-
ted-roughness, equation (14.15) becomes:
 
fL ks
F4 Re; 14:16
2D D

Equation (14.16) expresses the dimensionless relationship between friction losses in

pipes, the Reynolds number and relative roughness. An illustration is the Moody
diagram (Fig. 14.5).

Skin friction and form drag

Considering the drag on a body (e.g. Figs 14.3 and 14.4), the pressure losses
associated with the modication of the ow eld caused by the presence of
the body are usually expressed in terms of the drag force Fd on the body. The
Euler number is rewritten as: Eu Fd =V 2 A, where A is the projection of the
body in the plane normal to the ow direction. Fd =A is equivalent to the pressure
loss
P.
Equations (14.10) and (14.15) may be combined to relate the drag coecient Cd to
the -terms:
Eu F
Cd  d F5 Fr; Re; We; Ma; . . . 14:17
2 V 2A
2
In equation (14.17), the Reynolds number Re is related to the skin friction drag due to
viscous shear as well as to form drag resulting from the separation of the ow stream-
lines from the body.

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
272 Physical modelling of hydraulics

14.4.3 Practical considerations in modelling fully-enclosed ows

The ow regime in pipes is either laminar or turbulent. In industrial applications, it is

commonly accepted that the ow becomes turbulent for Reynolds numbers larger
than 2000 to 3000, the Reynolds number being dened in terms of the equivalent
pipe diameter D and of the mean ow velocity V.
For turbulent ows, the ow regime can be sub-divided into three categories:
smooth, transition and fully rough. Each category can be distinguished as a function
of a shear Reynolds number dened as:

V ks
Re  14:18


where V is the shear velocity. The transition between smooth turbulence and fully-
rough turbulence is approximately dened as:

Flow situation Open channel ow Pipe ow

Ref. (Henderson 1966) (Schlichting 1979)

Smooth turbulent Re < 4 Re < 5

Transition 4 < Re < 100 5 < Re < 75
Fully rough turbulent 100 < Re 75 < Re

Dynamic similarity of fully-enclosed ows implies the same resistance coecient in

both the model and the prototype. This can be achieved with the Reynolds number
being the same in the model and prototype, or with both ows in the model and
prototype being fully-rough turbulent (Fig. 14.5).
If the full-scale ow is turbulent, it is extremely important to ensure that the model
ow is also turbulent. Laminar and turbulent ows exhibit very important basic
dierences. In most cases, turbulent ows should not be scaled with laminar ow
models.

Note
The Reynolds number can be kept constant by changing the owing uid. For
example the atmospheric wind ow past a tall building could be modelled in a
small-size water tunnel at the same Reynolds number.

14.5 Modelling free-surface ows

14.5.1 Presentation

In free-surface ows (e.g. rivers, wave motion), gravity eects are predominant.
Model-prototype similarity is performed usually with a Froude similitude:

Frp Frm 14:19

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.5 Modelling free-surface ows 273

If the gravity acceleration is the same in both the model and prototype, a Froude
number modelling implies:
p
Vr Lr (Froude similitude)

pvelocity is less than that in the prototype for Lr > 1 and the time
Note that the model
scale equals tr Lr .

Remarks
Froude number modelling is typically used when friction losses are small and the
ow is highly turbulent: e.g. spillways, overow weirs, ow past bridge piers. It is
also used in studies involving large waves: e.g. breakwater or ship models.3
A main concern is the potential for scale eects induced by viscous forces.
Scale eects caused by surface tension eects are another concern, in particular
when free-surface aeration (i.e. air entrainment) takes place.

14.5.2 Modelling hydraulic structures and wave motion

In hydraulic structures and for wave motion studies (Fig. 14.2), the gravity eect is
usually predominant in the prototype. The ow is turbulent, and hence viscous and
surface tension eects are negligible in the prototype if the ow velocity is reasonably
small. In such cases a Froude similitude must be selected.
The most economical strategy is:
1. to choose a geometric scale ratio Lr such as to keep the model dimensions small,
and
2. to ensure that the model Reynolds number Rem is large enough to make the ow
turbulent at the smallest test ows.

14.5.3 Modelling rivers and ood plains

In river modelling, gravity eects and viscous eects are basically of the same order of
magnitude. For example, in uniform equilibrium ows (i.e. normal ows), the gravity
force component counterbalances exactly the ow resistance and the ow conditions
are deduced from the continuity and momentum equations.
In practice, river models are scaled with a Froude similitude (equation (14.19)) and
viscous scale eects must be minimized. The model ow must be turbulent, and
possibly fully-rough turbulent with the same relative roughness as for the prototype:
Rem > 5000 14:20
ks r Lr 14:21

3
The testing of ship models is very specialized. Interestingly, F. Reech and W. Froude were among the rst
to use the Froude similitude for ship modelling.

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
274 Physical modelling of hydraulics

where the Reynolds number is dened in terms of the hydraulic diameter (i.e.
Re VDH =).

Distorted models
A distorted model is a physical model in which the geometric scale is dierent between
each main direction. For example, river models are usually designed with a larger
scaling ratio in the horizontal directions than in the vertical direction: Xr > Zr . The
scale distortion does not distort seriously the ow pattern and it usually gives good
results.
A classical example of a distorted model is that of the Mississippi river, built by the
US Army Corps of Engineers. The Mississippi basin is about 3 100 000 km2 and the
river is nearly 3800 km long. An outdoor model was built with a scale of 2000: 1. If
the same scaling ratio was applied to both the vertical and horizontal dimensions,
prototype depths of about 6 m would imply model depths of about 3 mm. With
such small ow depths, surface tension and viscous eects would be signicant.
The Mississippi model was built, in fact, with a distorted scale: Zr 100 and
Xr 2000. Altogether the model size is about 1.5 km per 2 km!
A distorted model of rivers is designed with a Froude similitude:
Frp Frm 14:19
where the Froude number scaling ratio is related to the vertical scale ratio:
Vr
Frr p 14:22
Zr
As for an undistorted model, the distorted model ow must be turbulent (equation
(14.20)), and preferably fully-rough turbulent with the same relative roughness as
for the prototype:
ks r Zr 14:23
The Froude similitude (equation (14.22)) implies:
p
Vr Zr Velocity 14:24
3=2
Qr Vr Xr Zr Zr Xr Discharge 14:25
Xr Xr
tr p Time 14:26
Vr Zr
Zr
tan r Longitudinal bed slope 14:27
Xr
where  is the angle between the channel bed and the horizontal.
With a distorted scale model, it is possible to select small physical models (i.e. Xr
large). In addition to the economical and practical benets, distorted models also
have the following advantages compared with non-distorted models:
. the ow velocities and turbulence in the model are larger (equation (14.24)),
. the time scale is reduced (equation (14.26)),
. the model Reynolds number is larger, improving the prototype-to-model dynamic
similarity, and

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.5 Modelling free-surface ows 275

. the larger vertical scale (i.e. Zr < Xr ) allows a greater accuracy on the ow depth
measurements.

Discussion
Practically it is recommended that the model distortion (i.e. the ratio Xr =Zr )
should be less than 5 to 10. Some disadvantages of distorted models may be
mentioned for completeness: the velocity directions are not always reproduced
correctly, and some observers might be distracted unfavourably by the model
distortion leading to inaccurate or incorrect judgements.

Movable-bed models
Movable-bed hydraulic models are some of the most dicult types of models and they
often give unsatisfactory results.
The primary diculty is to scale both the sediment movement and the uid motion.
Furthermore, the bed roughness becomes a function of the bed geometry and of the
sediment transport. Early movable bed model studies on the River Mersey (England)
and Seine River (France) in the 1880s showed that the time scale governing the uid
ow diers from the time scale governing sediment motion (see Appendix A3.1).
A detailed analysis of sediment transport modelling is developed in Appendix A3.1.
Several authors (e.g. Henderson 1996, pp. 497508, Graf 1971, pp. 392398) also
discussed various methods for `designing' a movable-bed model.
The most important point is the need to verify and to calibrate a movable-bed
model before using it as a prediction tool.

14.5.4 Resistance scaling

The modelling of ow resistance is not a simple matter. Often the geometric similarity
of roughness height and spacing is not enough. For example, it is observed sometimes
that the model does not reproduce the ow patterns in the prototype because the
model is too `smooth' or too `rough'. In some cases (particularly with a large scale
ratio Lr ), the model ow is not as turbulent as the prototype ow. A solution is to
use roughness elements (e.g. mesh, wire, vertical rods) to enhance the model ow
turbulence, hence to simulate more satisfactorily the prototype ow pattern.
Another aspect is the scaling of the resistance coecient. The ow resistance can be
described in terms of the Darcy friction factor or an empirical resistance coecient
(e.g. Chezy or GaucklerManning coecients).
In uniform equilibrium ows, the momentum equation implies:
s
p DH r sin r
Vr Lr 14:28
fr
For an undistorted model, a Froude similitude (equation (14.19) and (14.28)) implies
that the model ow resistance will be similar to that in the prototype:
fr 1 14:29

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
276 Physical modelling of hydraulics

Most prototype ows are fully-rough turbulent and the Darcy friction factor is
primarily a function of the relative roughness.
Another approach is based upon the GaucklerManning coecient. The Chezy
equation implies that, in gradually-varied and uniform equilibrium ows, the follow-
ing scaling relationship holds:
p 1 p
Vr Lr DH r 2=3 sin r 14:30
nManning r

For an undistorted scale model, equation (14.30) becomes:

1=6
nManning r Lr 14:31

Equation (14.31) indicates that the notion of complete similarity is applied both to the
texture of the surface and to the shape of its general outline (Henderson 1966). In
practice, the lowest achievable value of nManning is about 0.009 to 0.010 s/m1=3 (i.e.
for glass). With such a value, the prototype resistance coecient nManning p and
the GaucklerManning coecient similarity nManning r could limit the maximum
geometrical similarity ratio Lr . If Lr is too small (typically less than 40), the physical
model might not be economical nor convenient.
In summary, a physical model (based upon a Froude similitude) has proportion-
ally more resistance than the prototype. If the resistance losses are small (e.g. at a
weir crest), the resistance scale eects are not considered. In the cases of river
and harbour modelling, resistance is signicant. The matter may be solved using
distorted models.

Distorted models
With a distorted scale model, equations (14.28) and (14.30) become respectively:
s
p DH r sin r
Vr Zr 14:32
fr
p 1 p
Vr Zr DH r 2=3 sin r 14:33
nManning r

For a wide channel (i.e. DH r Zr ) and a at slope (i.e. sin r tan r ), the
scaling of ow resistance in distorted models implies:
Zr
fr wide channel and flat slope 14:34
Xr
2=3
Zr
nManning r p wide channel and flat slope 14:35
Xr

Discussion
In practice Zr =Xr < 1 and equation (14.34) would predict a model friction factor
lower than that in the prototype. But equation (14.35) could imply a model

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.6 Design of physical models 277

resistance coecient larger or smaller than that in the prototype depending upon
2=3 p
the ratio Zr = Xr !

14.6 Design of physical models

14.6.1 Introduction

Before building a physical model, engineers must have the appropriate topographic
and hydrological eld information. The type of model must then be selected, and a
question arises:
Which is the dominant effect: e.g. viscosity, gravity or surface tension?

14.6.2 General case

In the general case, the engineer must choose a proper geometric scale. The selection
procedure is an iterative process.

Step 1. Select the smallest geometric scale ratio Lr to t within the constraints of the
laboratory.

Step 2. For Lr , and for the similitude criterion (e.g. Froude or Reynolds), check if the
discharge can be scaled properly in the model, based upon the maximum
model discharge Qm max .

For Lr and the similitude criterion, is the maximum model discharge large
enough to model the prototype ow conditions?

Step 3. Check if the ow resistance scaling is achievable in the model.

Is it possible to achieve the required fm or nManning m in the model?

Step 4. Check the model Reynolds number Rem for the smallest test ow rate.

For Qp min , what are the ow conditions in the model: e.g. laminar or
turbulent, smooth-turbulent or fully-rough-turbulent? If the prototype ow is
turbulent, model ow conditions must be turbulent (i.e. typically Rem > 5000).

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
278 Physical modelling of hydraulics

Step 5. Choose the convenient scale.

When a simple physical model is not feasible, more advanced modelling techniques
can be used: e.g. a two-dimensional model (e.g. spillway ow), a distorted scale
model (e.g. river ow).

14.6.3 Distorted scale models

For a distorted scale model, the engineer must select two (or three) geometric scales.
The model design procedure is again an iterative process:
Step 1. Select the smallest horizontal scale ratio Xr to t within the constraints of the
laboratory.
Step 2. Determine the possible range of vertical scale Zr such as:
the smallest scale Zr 1 is that which gives the limit of the discharge scaling
ratio, based upon the maximum model discharge Qm max ,
the largest scale Zr 2 is that which gives the feasible ow resistance coe-
cient (i.e. feasible fm or nManning m , and
check the distortion ratio Xr =Zr .
(Xr =Zr should be less than 5 to 10.)
Step 3. Check the model Reynolds number Rem for the smallest test ow rate. This
might provide a new largest vertical scale ratio Zr 3 .
check the distortion ratio Xr =Zr .
Step 4. Select a vertical scale ratio which satises: Zr 1 < Zr < MinZr 2 ; Zr 3 . If
this condition cannot be satised, a smaller horizontal scale ratio must be
chosen.
check the distortion ratio Xr =Zr .
In practice it is recommended that Xr =Zr should be less than 5 to 10.
Step 5. Choose the convenient scales (Xr , Zr ).

14.7 Summary
Physical hydraulic modelling is a design technique used by engineers to optimize the
structure design, to ensure the safe operation of the structure and/or to facilitate the
decision-making process.
In practice, most hydraulic models are scaled with either a Froude or a Reynolds
similitude: i.e. the selected dimensionless number is the same in the model and in
the prototype (i.e. full-scale).
The most common uids are air and water. Free-surface ow modelling is most
often performed with the same uid (e.g. water) in full-scale and the model. Fully-
enclosed ow modelling might be performed with water in the prototype and air in
the model. The selection of uid in the model and the prototype xes the density
scale ratio r .
Table 14.1 summarizes the scaling ratios for the Froude and Reynolds
similitudes.

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.8 Exercises 279

Table 14.1 Scaling ratios for Froude and Reynolds similitudes (undistorted model)

Parameter Unit Scale ratio with

a
Froude law Froude lawa Reynolds law
(distorted model)
(1) (2) (3) (4) (5)

Geometric properties
Length m Lr Xr , Zr Lr
Area m2 L2r L2r
Kinematic properties p p
Velocity m/s Lr Z 1=Lr  r =r
3=2 pr
Discharge per unit width m2 /s Lr Zr Xr r =r
5=2 3=2
Discharge m3 /s Lr Zr Xr Lr r =r
p p
Time s Lr X r = Zr L2r r =r
Dynamic properties
Force N r L3r 2r =r
Pressure Pa r Lr  r Zr 2r =r  1=L2r
Density kg/m3 r r r
3=2 p
Dynamic viscosity Pa s Lr r r
Surface tension N/m L2r 2r =r  1=Lr
Note: a assuming identical gravity acceleration in model and prototype.

14.8 Exercises
Numerical solutions to some of these exercises are available from the Web at
www.arnoldpublishers.com/support/chanson
A buttery valve is to be tested in a laboratory to determine the discharge coecient
for various openings of the disc. The prototype size will be 2.2 m in diameter and it
will be manufactured from cast steel with machined inside surfaces (roughness
height estimated to be about 0.5 mm). The maximum discharge to be controlled by
the valve is 15 m3 /s. The laboratory model is a 5 : 1 scale model.
(a) What surface condition is required in the model? What model discharge is
required to achieve complete similarity with the prototype, if water is used in both?
(b) Can these conditions be achieved? (c) If the maximum ow available for model
tests is 200 L/s, could you accurately predict prototype discharge coecients from
the results of the model tests?
Summary sheet:

(a) ks m Qm

(b) Yes/No Reasons:

(c) Rep Rem

Discussion:

The inlet of a Francis turbine is to be tested in a laboratory to determine the per-

formances for various discharges. The prototype size of the radial ow rotor will be:

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
280 Physical modelling of hydraulics

inlet diameter 0.6 m, width 0.08 m, inlet crossow area   diameter  width. It
will be manufactured from cast steel with machined inside surfaces (roughness height
estimated to be about 0.3 mm). The maximum discharge to be turbined (by the
Francis wheel) is 1.4 m3 /s. The laboratory model is a 5:1 scale model.
(a) What surface condition is required in the model? What model discharge is required
to achieve complete similarity with the prototype, if water is used in both?
(b) Can these conditions be achieved? (Compute the minimum required model total
head and ow rate. Compare these with the pump performances of a typical
hydraulic laboratory: H  10 m, Q  100 L/s.)
(c) If the maximum ow available for model tests is 150 L/s, would you be able to
predict accurately prototype performances from the results of the model tests?
(Justify your answer.)

Summary sheet:

(a) ks m Qm

(b) Yes/No Reasons:

(c) Rep Rem

Discussion:

An overow spillway is to be designed with an uncontrolled crest followed by a stepped

chute and a hydraulic jump dissipator. The maximum spillway capacity will be
4300 m3 /s. The width of the crest, chute and dissipation basin will be 55 m. A 50:1
scale model of the spillway is to be built. Discharges ranging between the maximum
ow rate and 10% of the maximum ow rate are to be reproduced in the model.
(a) Determine the maximum model discharge required. (b) Determine the mini-
mum prototype discharge for which negligible scale eects occur in the model. (Com-
ment on your result.) (c) What will be the scale for the force ratio?
Laboratory tests indicate that operation of the basin may result in unsteady
wave propagation downstream of the stilling basin with a model wave amplitude of
about 0.05 m and model wave period of 47 seconds. Calculate: (d) the prototype
wave amplitude and (e) the prototype wave period.

Summary sheet:

(a) Maximum Qm

(b) Minimum Qp Why?

(c) Forcer

(d) Ap

(e) tp

A 35.5:1 scale model of a concrete overfall spillway and stilling basin is to be built. The
prototype discharge will be 200 m3 /s and the spillway crest length is 62 m. (a) Determine
the maximum model discharge required and the minimum prototype discharge for

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.8 Exercises 281

which negligible scale eects occur in the model. (b) In tests involving bae blocks for
stabilizing the hydraulic jump in the stilling basin, the force measured on each block was
9.3 N. What is the corresponding prototype force? (c) The channel downstream of the
stilling basin is to be lined with rip-rap (angular blocks of rock) approximately
650 mm in size. The velocity measured near the rip-rap is as low as 0.2 m/s. Check
whether the model Reynolds number is large enough for the drag coecient of the
model rocks to be the same as in the prototype. What will be the scale for the force ratio?
Summary sheet:
(a) Maximum Qm
Minimum Qp Why?

(b) Forcep

(c) Rem Comment:

Force ratio

A sluice gate will be built across a 25 m wide rectangular channel. The maximum
prototype discharge will be 275 m3 /s and the channel bed will be horizontal and con-
crete-lined (i.e. smooth). A 35:1 scale model of the gate is to be built for laboratory tests.
(a) What similitude should be used? Calculate: (b) model width and (c) maximum
model ow rate.
For one particular gate opening and ow rate, the laboratory ow conditions are:
upstream ow depth of 0.2856 m, downstream ow depth of 0.0233 m. (d) Compute
the model discharge. State the basic principle(s) involved. (e) Compute the model
force acting on the sluice gate. State the basic principle(s) involved. (f ) What will be
the corresponding prototype discharge and force on the gate? (g) What will be the
scale for the force ratio?
Gate operation may result in unsteady ow situations. If a prototype gate
operation has the following characteristics: gate opening duration 15 minutes,
initial discharge 180 m3 /s, new discharge 275 m3 /s, calculate: (h) gate opening
duration and (i) discharges to be used in the model tests.
Summary sheet:
(a) Similitude: Why?

(b) Bm

(c) Qm max

(d) Qm Principle(s):

(e) Fm Principle(s):

(f ) Qp
Forcep

(g) Forcer

(h) tm

(i) Qm (before) Qm (after)

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
282 Physical modelling of hydraulics

A hydraulic jump stilling basin, equipped with bae blocks, is to be tested in a labora-
tory to determine the dissipation characteristics for various ow rates. The maximum
prototype discharge will be 220 m3 /s and the rectangular channel will be 10 m wide.
(Assume the channel bed to be horizontal and concrete-lined, i.e. smooth.) A 40: 1
scale model of the stilling basin is to be built. Discharges ranging between the
maximum ow rate and 10% of the maximum ow rate are to be reproduced in
the model.
(a) What similitude should be used? (Justify your selection.) (b) Determine the
maximum model discharge required. (c) Determine the minimum prototype
discharge for which negligible scale eects occur in the model. (Comment on your
result.)
For one particular inow condition, the laboratory ow conditions are: upstream
ow depth of 0.019 m, upstream ow velocity of 2.38 m/s, downstream ow depth
of 0.122 m. (d) Compute the model force exerted on the bae blocks. (State the
basic principle(s) involved.) (e) What is the direction of force in (d): i.e. upstream
or downstream? (f ) What will be the corresponding prototype force acting on the
blocks? (g) Compute the prototype head loss.
Operation of the basin may result in unsteady wave propagation downstream of the
stilling basin. (h) What will be the scale for the time ratio?

Tests will be made on a model sea wall of 1/18 prototype size. (a) If the prototype
wave climate is: wave period 12 seconds, wave length 20 m, wave amplitude
2.1 m, what wave period, wave length and wave amplitude should be used in the
model tests? (b) If the maximum force exerted by a wave on the model sea wall is
95 N, what corresponding force will be exerted on the prototype?

Summary sheet:

(a) tm Lm

Am

(b) Forcep

A xed bed model is to be built of a certain section of a river. The maximum full-scale
discharge is 2750 m3 /s, the average width of the river is 220 m and the bed slope is
0.16 m per kilometre. The GaucklerManning coecient for the prototype is
estimated at 0.035 s/m1=3 . Laboratory facilities limit the scale ratio to 200 : 1 and
maximum model discharge is 45 L/s. Note that the smoothest model surface feasible
has a GaucklerManning coecient of about 0.014 s/m1=3 . Discharges ranging
between the maximum ow rate and 15% of the maximum ow rate are to be
reproduced in the model.
Determine the acceptable maximum and minimum values of the vertical scale ratio
Zr . Select a suitable scale for practical use, and calculate the corresponding model
values of the GaucklerManning coecient, maximum discharge and normal
depth (at maximum discharge). (It can be assumed that the river channel is wide
(i.e. DH  4d) in both the model and prototype for all ows. Assume that uniform
equilibrium ow is achieved in the model and prototype.)

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.
 14.8 Exercises 283

Summary sheet:

Minimum Zr Why?

Maximum Zr Why?

Alternative Max. Zr Why?

Allowable range for Zr

Your choice for Zr Why?

Corresponding values of Qr nManning r

nManning r Maximum Qm

dm at maximum ow rate

An articial concrete channel model is to be built. Laboratory facilities limit the scale
ratio to 50 : 1 and the maximum model discharge is 50 L/s. The maximum full-scale
discharge is 150 m3 /s, the cross-section of the channel is approximately rectangular
(50 m bottom width) and the bed slope is 0.14 m per kilometre. (Note: The roughness
height of the prototype is estimated as 3 mm while the smoothest model surface
feasible has a Darcy friction factor of about f 0:03.) Discharges ranging between
the maximum ow rate and 10% of the maximum ow rate are to be reproduced
in the model.
For an undistorted model: (a) what would be the model discharge at maximum full-
scale discharge? (b) what would be the Darcy coecient of the model ow? (c) what
would be the Darcy coecient of the prototype channel? (d) comment and discuss
your ndings. (Assume normal ow conditions.)
A distorted model is to be built. (e) Determine the acceptable maximum and mini-
mum values of the vertical scale ratio Zr . (f ) Select a suitable scale for practical use.
Calculate the corresponding model values of: (g) Darcy coecient, (h) maximum
discharge and (i) normal depth (at maximum discharge).

A xed bed model is to be made of a river with a surface width of 80 m. The

GaucklerManning coecient for the river is estimated at 0.026 s/m1=3 . Scale ratios
of Xr 150 and Zr 25 have been selected. (a) Find the required model values of
the GaucklerManning coecient corresponding to prototype depths of water of
2.0 and 5.0 m, if the cross-sectional shape is assumed to be rectangular. (b) What
material would you recommend to use in the laboratory model for a prototype
depth of 2.0 m?

The Hydraulics of Open Channel Flow by Hubert Chanson.

Published in 1999 by Arnold, 338 Euston Road, London NW1 3BH, UK.
# Hubert Chanson. All rights reserved.